MORAL HUMILITY: AN ANTIDOTE TO THE DARK SIDE OF MORALITY
By
Shree Vallabha
A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Psychology — Doctor of Philosophy
2025
ABSTRACT
Morality has a dark side. Our moral tendencies breed rigidity, conflict, extremism, hate, intolerance, and
violence. In this project, I proposed moral humility as one antidote to these dark features of our morality. Given that
moral humility is a largely empirically unexplored phenomenon, across ten studies (N = 10, 978) in this project, I
investigated its measurement and structure, its nature and correlates, and its implications in a context characterized
by the dark features of our morality.
To these ends, I first developed a moral humility scale using psychometric factor analytic methods (EFA
and CFA), which comprised of three factors—moral fallibility, moral openness/learning, and moral superiority.
Second, I probed its nomological network, specifically testing its relationship to personality, religiosity, ideology,
political extremity, moral grandstanding, and other relevant constructs. Moral humility was associated with other
constructs in expected and sensible ways, such as higher openness to experience and modesty, and lower moral
grandstanding and political extremity. Third and importantly, I tested its predictive validity in the context of political
polarization in the US. Polarization was chosen as the context because research has suggested that it is a moralized
context with features associated with the dark aspects of morality. Across correlational studies, I found that moral
humility was associated with lower levels of polarization across a range of outcomes. It was associated with lower
antipathy and antagonism towards outgroup, lower rigidity in one’s own political views, and lower rejection of
compromise and contact, amongst other outcomes. These findings provided strong evidence to suggest that moral
humility could be ameliorative in contexts that bring out our dark moral nature. These studies also established moral
humility’s incremental validity over related and important constructs like moral relativism and intellectual humility,
indicating that moral humility is likely a distinct construct and has unique explanatory value.
Finally, I designed five moral humility interventions and tested its impact on moral humility and
polarization using experimental design. I found that some interventions worked more consistently in increasing
moral humility – providing evidence that moral humility can be targeted and increased using interventions. These
interventions also lowered polarization, such as decreasing antipathy against political outgroup and increasing
willingness toward cross-cutting exposure, thus further supporting moral humility’s predictive and causal role in
moralized and high-conflict situations. The experimental studies also established moral humility’s discriminant
validity from close constructs like moral relativism and intellectual humility, reinforcing that moral humility offers a
distinct and meaningful mechanism through which moralized conflicts and contexts can be understood and
addressed. This project also highlighted gaps in the current understanding of moral humility and laid the
groundwork for future inquiry into the nature and significance, such as its impact on moral behaviors and moral
improvement, mechanism underlying change and development of moral humility, and contextual and cultural
variations in moral humility.
Taken together, this project is the first comprehensive investigation into moral humility. By developing a
validated measure, mapping its conceptual network, and demonstrating its predictive and causal role in moralized
context like political polarization, this research establishes moral humility as a theoretically meaningful and
practically consequential construct. It provides insight into how moral humility could help counter the dark and ugly
aspects of our morality and quell the rigidity, intolerance, and antagonism that often accompany moralized
intergroup conflicts. Finally, it opens doors for a deeper understanding of moral humility across outcomes,
individuals, context, and cultures.
Dedicated to the realization of
Saccidānanda
iv
ACKNOWLEDGEMENTS
With this dissertation, a very consequential and meaningful chapter of my life comes to a close. I want to
take a moment to acknowledge all those who have profoundly enriched my life and journey so far, and have
especially made it possible for me to reach this significant milestone.
Foremost, my advisor, Mark Brandt, who has poured so much energy, time, and resources into
transforming me into a careful and rigorous scholar, nothing I will say here will be enough to fully express my
gratitude to him. For whatever reason, he took a chance on me and my potential – someone from thousands of miles
away, from a different culture and educational context – and allowed me the privilege of learning how to do good
research from one of the best in the field. For five years, a very substantial chunk of his life and mine – he has
generously, earnestly, and to a great extent selflessly, invested in my development as a social scientist, a
psychologist, a researcher, a theorist, a methodologist, an academic, a teacher, a mentor, a public speaker, a writer, a
professional, and a person with a meaningful and rewarding life. Very few people have an inordinate impact on your
life—Mark is one of those rare people whose impact on my life is and will always be immeasurable and life-
changing. Everything that I have accomplished so far academically and will accomplish academically in the future, I
owe and dedicate to him, and even that feels insignificant compared all that he has done for me. I remember him
telling me during my first semester as his mentee that he wanted to mentor me in a way that would prepare me to
stand confidently as an independent scholar by the end of my PhD. I believe that not only has he nurtured me into
becoming that but also very importantly exemplified for me the value of being an ethical, kind, and responsible
scholar. I will always be grateful for the privilege of having been his student. I hope to take all that I have learnt
from him into the next phase of my life, and carry forward at least some of his extraordinary creativity, rigor,
prolificacy, integrity, generosity, and goodness through my own future pursuits. I look forward to many more years
of collaboration and I hope I will make him proud.
I am also immensely thankful to other outstanding mentors and academics whom I have known closely and
who have enriched my intellectual and academic development. This includes Rich Lucas, Jennifer Wolak, and Bill
Chopik — who I am fortunate to have as my dissertation committee. Taking Rich’s class in grad school and
engaging with his intellectual process through brownbags and meetings fundamentally changed my outlook to
research and teaching. It has moved me to be attracted to the big picture in science and to be ambitious in pushing
the boundaries of your field. Jenny’s class was the first one that I took in grad school that spoke to my own
v
substantive interests. It was in this class, that through her very invested and systematic mentorship, I “birthed” the
idea and research on moral humility, and simultaneously learnt how to go about methodically formulating research
questions in psychology. I am deeply grateful for that. I never formally had a secondary advisor during graduate
school, but I felt that Bill always had my back and that someone apart from my advisor was always looking out for
my best interests. I am always in awe of his creative thinking as a scientist, and proactive and supportive spirit as a
mentor—qualities that I hope to emulate in my own academic career. I am grateful to late Debby Kashy for making
me love statistics again. I am also very grateful to my mentors from my previous academic institution, CBCS,
especially Prof Narayanan Srinivasan who embodies extraordinary scientific excellence and cultivated in me deep
intellectual curiosity and critical thinking — which forms the bedrock of academic identity. Without his and Prof
Bhoomika Kar’s support and encouragement, I could not have gotten here.
My family is the core of my meaning, existence, and bliss. They have always been a reminder to me of the
beauty and privilege of life, which sometimes got obscured by the stress and isolation of graduate school. All my
pursuits, intellectual and otherwise, spring from the strength and meaning I derive from the immensity of their love,
support, and wisdom. I am eternally thankful to them for everything and could never get to where I am today
without them. My special thanks a few family members. My parents who have been unimaginably and
unconditionally supportive of me through my whole life, and especially when I chose a field that wasn’t very
respected in their social milieu, and when I further wanted to go to a faraway land to pursue it. Honestly, thanking
them using mere words is the hardest and feels the most inadequate. I will spend my whole life thanking them
through my actions and devotion. My little brother, Mukunda, who has been my best friend and always challenged
my worldviews, routines, and habits — pushing me to continuously grow and never become stagnant or dogmatic.
Amruta, Omkar, and the kids, for being a home away from home; I found so much warmth in your abode. And
finally, the love of my life, my husband Pranav, who first inspired me to listen to my intellectual instincts over ten
years ago, who through his extraordinary philosophical aptitude helped me develop clearer conceptual thinking
about my research during graduate school, who has elevated my life spiritually and aesthetically, who fills my life
with so much hope, joy, love, courage, and adventure that I feel like a hero(ine) in a great story, and with whom I
see a future so thrilling that I am eager to leap forward into the next chapter of life.
My immense gratitude to all my friends, colleagues, and lab members who have so beautifully filled the
interstices of these last five years that I shudder to imagine a life without them; the picture would have been indeed
vi
very dreary if they were not in there. I especially appreciate all the wonderful, smart, and kind graduate students I
had the privilege of interacting and/or working with. Special thanks to my lab sister Abby for being such a big part
of my time here in USA. I have been so fortunate to have met her, worked with her, laughed with her, travelled with
her, been vulnerable with her, discussed life, research, and politics with her—it has truly been wholesome. I have
learnt so much from her and deeply admire her for her warmth, resilience, and intellect. I look forward to our life-
long friendship and collaborations. Shoutout to Kenya and Hyewon who have been very sharp, interesting, and
amazing friends and colleagues. I have had such a heartwarming and intellectually enriching time with them, I could
not have asked for better companions on this shared journey. Thanks to Jeewon for being the first colleague to make
Lansing feel warm and welcoming, and for being absolutely inspirational. Thanks to Brian and Lindsay for often
being a social glue in our program and hosting wonderful get togethers (and allowing me the joy of being around
their beautiful kids). Thanks to Rebekka and Mariah for their help and guidance during the job market. Thanks also
to all the undergraduate research assistants and students I worked with during my time in graduate school—for
improving me and my research.
Finally, thanks to the Psychology Department at Michigan State University for supporting me and
providing me with multiple opportunities to learn and grow academically. Thanks also to Michigan State University
and USA for a life-changing and memorable experience. I will definitely look back at these five years as one of the
best times of my life.
vii
TABLE OF CONTENTS
Chapter I: Introduction ................................................................................................................................................... 1
Chapter II: Scale Development, Nomological Associations, Predictive Validity ....................................................... 11
Chapter III: Intervention Development and Causal Assessment ................................................................................. 55
Chapter IV: General Discussion, Future Directions, and Conclusion ....................................................................... 105
REFERENCES ........................................................................................................................................................... 117
viii
Chapter I: Introduction
Morality plays a unique, central, and multifaceted role in human psychological experience. Our moral
tendencies motivate noble and inspiring acts of heroism, altruism, kindness, sacrifice, cooperation, and collective
action (Curry et al., 2019; Haidt, 2012; Tomasello & Vaish, 2013; Van Zomeren et al., 2012). For these reasons,
people have always considered our moral sense as one of our chief human sensibilities. This sentiment is echoed in
Charles Darwin’s writing on human morality (1871), “I fully subscribe to the judgment of those writers who
maintain that of all the differences between man and the lower animals, the moral sense or conscience is by far the
most important …it is summed up in that short but imperious word ought, so full of high significance. It is the most
noble of all the attributes of man.”
However, morality is a double-edged sword. Apart from its noble side, our capacity for morality also has a
dark side. Decades of psychological work has uncovered how our moral tendencies breed rigidity, conflict, hate,
intolerance, and violence (Baumeister, 1999; Fiske & Rai, 2014; Kovacheff et al., 2018; Skitka et al., 2021; Pretus et
al., 2023). This is because when people process something as a moral concern, i.e., as a matter of right and wrong, or
good and bad, people consider their moral stance on the matter as objective and universal. They further imbue their
moral stance with intense emotions and prescriptive motivations. Consequently, they become rigid, extreme,
intolerant of disagreement, and reject compromise (Garrett, 2019; Ryan, 2014; Yoder & Decety, 2022; Clifford,
2019; Delton et al., 2020; Goodwin & Darley, 2008; Jung & Clifford, 2024; Kodapanakkal et al., 2022; Pretus et al.,
2023; Ryan, 2017; Skitka et al., 2021; Van Bavel et al., 2012). In other words, their thinking and approach becomes
black-and-white. People consider themselves to be on the “good” side and those who think or act differently are
perceived as bad or evil. People become willing to adopt violent solutions to conflicts and use morality to legitimize
such actions and tendencies (Baumeister, 1999; Fiske & Rai, 2014; Haidt, 2012; Pinker, 2008; Skitka et al., 2021).
Indeed, many perpetrators of violence and atrocities are found to motivated by morality, i.e., doing what is morally
good (Fiske & Rai, 2014). They believe their violent and oppressive actions are justifiable and necessary means to a
noble, moral end (Baumeister, 1999; Fiske & Rai, 2014; Skitka & Mullen, 2002; Effron & Miller, 2012; Haidt,
2012; Reicher et al., 2008; Sedikides et al., 2014; Kovacheff et al., 2018). Our moral sense therefore “has the nasty
habit of always putting the self on the side of the angels.” (Pinker, 2008)
Considering how our morality traps us into a rigid mindset, making us quick to demonize others, and
paradoxically, blinds us to our own potentially harmful tendencies, I suggest that it might be fruitful to find
1
countervailing forces that can counter the dark side of our morality such as moral disdain, moral righteousness,
moral supremacy, and moral rigidity. In the current project, building on the psychological literature on the virtue of
humility (McElroy et al., 2019; Wright et al., 2017; Davis at el., 2016; Van Tongeren et al., 2019), I advance a form
of humility, moral humility, as one such antidote to our morally dark tendencies. Thus, in this project, I first explore
the idea of moral humility and why it’s a suitable antidote to the dark aspects of our morality. Second, I develop a
psychometrically valid measure for it. Third, I test moral humility’s association with other constructs. Fourth, I test
if moral humility has positive implications in the context of a present-day moral conflict i.e., political polarization in
the USA. Fifth, I develop and validate interventions of moral humility, test if it can be changed, and examine if it
has downstream outcomes on conflicts like political polarization.
Humility
Humility is understood as an attribute or quality that lowers self-focus and increases other-focus (Wright et
al., 2017). That is, it “involves a shift from the narrow preoccupation with self…into the broader consideration of
self and other” (Wright et al., 2017, p. 4). Specifically, having humility has been associated with having an accurate
view of the self, which involves attributes like acknowledging one’s limitations and weaknesses. It is also associated
with having open-mindedness and flexibility which involves attributes like a general desire to learn and correct
mistakes. And with having an other-oriented interpersonal stance which involves attributes like restraint of egotism,
being modest in self-presentation, and a respectful attitude towards others’ ideas, skills, and abilities (McElroy et al.,
2019; Wright et al., 2017; Davis at el., 2016; Van Tongeren et al., 2019). A humble person would be someone who
understands that they are not perfect, that they have scope to learn and improve, and holds an attitude of respect and
openness, and not superiority and disdain towards others.
The work on virtues has highlighted that humility occurs in different types of domains or contexts and can
accordingly vary across these. For example, a politician might be high in humility about their athletic abilities, but
low in humility about their moral traits, or an engineer might be high in humility about their social skills, but low in
humility about their technical abilities. Thus, the construct of humility may be too general. For these reasons,
identifying important domains which can make the idea of humility more specific and increase its predictive
efficacy has been considered fruitful (Davis at el., 2016; McElroy et al., 2014; Hoyle et al., 2016; Leary et al., 2017;
Ballantyne, 2023; Van Tongeren et al., 2019).
2
Moral Humility
In line with this, I propose that morality may be an important domain for enriching our understanding of
humility. Thus, humility in the domain of morality, or what I call moral humility, forms the center of the
investigation in this project. The reason why I consider morality an important domain for understanding humility is
multifaceted. Before going into these reasons, I first conceptualize moral humility.
Definition
Building on the idea of humility and initial work on moral humility, I conceptualize moral humility as the
awareness of one’s moral limitations or fallibility, having moral openness or moral learning orientation, and an
other-focused moral orientation (Owens et al., 2019, Smith & Kouchaki, 2018; Vallabha et al., 2024). Like humility,
it contains both self-oriented and other-oriented dimensions of fallibility, learning, openness, and respect, but
specifically about our own and others’ morality. Thus, a morally humble person acknowledges their moral
imperfections, strives towards moral improvement and learning, acknowledges others’ moral strengths, and
expresses understanding and openness towards moral differences.
Importance of the moral domain
Why should we study humility in the moral domain? Like previously mentioned, morality plays a unique
and central role in human psychological experience. For instance, people experience morality as a basic
psychological need (Prentice et al., 2019) — they have a fundamental need to feel moral. They consider moral
aspects of the self as most central to personal identity (Strohminger & Nichols 2014; Stanley et al., 2020), and moral
self-identity predicts well-being and meaning in life (Goering et al., 2024). Other people and groups are judged
primarily on their morality, which is associated with how people interact and affiliate with them (Brambilla et al.,
2021; Goodwin et al., 2014; Nicolas et al., 2022; Leach et al., 2015; Halevy et al., 2015). Morality is associated with
intense and specific emotions (Skitka et al., 2005; Schein & Gray, 2018; Brady et al., 2017; Garrett, 2019; Ryan,
2014; Yoder & Decety, 2022), and moralized attitudes and judgements are more rigid and extreme than non-moral
ones (Luttrell, et al., 2016; Van Bavel et al., 2012). Morality binds (us in groups, like religious, political) and blinds
us (to the humanity of moral outgroups; Haidt & Kesebir, 2010; Haidt, 2012). Thus, morality is central to our ego
and self-concept, making our emotions run hot, our minds closed, and serves as the basis of social affiliation,
cooperation, exclusion, and dehumanization. Together this suggests that the moral domain is at the front and center
3
of our social psychological experience, influencing our cognitions, emotions, and behaviors. These self and other
concerning aspects together make morality a relevant and important domain for the study of a virtue like humility.
The most important reason for studying moral humility is that if we look at the dark aspects of our moral
nature, like the ones highlighted before, we will notice that moral humility is a virtue that is very well positioned to
counter the negative aspects of morality. If morality traps us into rigid and closed mindsets of moral correctness,
moral humility opens us to moral growth and learning; if morality traps us into thinking that we are always on the
side of angels and makes a villain of those we disapprove, then moral humility enables us to see the other side with
respect, as morally worthy and valuable; if morality makes us susceptible to moral superiority, righteousness, and
sanctimony, then moral humility helps us become aware of our own moral flaws and weaknesses.
From a measurement perspective, domain-specific psychological traits are valuable as they are narrower in
scope and help enhance fidelity and criterion-related validity. However, too much narrowness can come at the cost
of predictive bandwidth and tautological tests with criteria (Salgado, 2017). In my view, morality is a domain that is
placed well between too much specificity or generality, making the study of moral humility as a domain specific
form of humility reasonable. It is suited to narrowly predict criteria that pertains to morality, but at the same time,
given the relatively wide range of things the moral domain touches (e.g., politics, religion, cultural norms,
leadership, relationships, conflicts), it is well positioned to predict a broader range of outcomes.
Moral humility in religion, culture, and literature
Does the idea of moral humility have any grounding in human experience and common understanding?
We can indeed find the idea of moral humility contained in various religions and cultures, scholarly and literary
works, and public discourse, highlighting its practical importance to humans. For instance, in Christianity, one is
advised to reflect and work on one’s own moral flaws before being quick in moral condemnation of others.
“Why do you see the speck that is in your brother’s eye, but do not notice the log that is in your own eye?
Or how can you say to your brother, ‘Let me take the speck out of your eye,’ when there is the log in your own eye?
You hypocrite, first take the log out of your own eye, and then you will see clearly to take the speck out of your
brother’s eye.”
- Matthew 7:3-5, The New Testament
“Let he who is without sin cast the first stone.”
- John 8:7, The New Testament
“Judge not, and ye shall not be judged. Condemn not, and ye shall not be condemned. Forgive, and ye shall be
forgiven.”
-Luke 6:37, The New Testament
A similar advice about searching for darkness within us before finding it in others is found in Indian teachings.
4
“Bura Jo Dekhan Main Chala, Bura Na Milya Koye
Jo Munn Khoja Apna, To Mujhse Bura Na Koye”
Translation: “I searched for evil, but didn’t find evil anywhere
When I searched myself, I found the biggest evil within.”
– Kabir’s Dohe
Carl Jung in his works grappled with the dark side within people (also called “shadow” in his work) and
wrote about the need for humans to realize their dark aspects to achieve self-actualization. He called it a moral
problem requiring considerable moral effort.
“Unfortunately, there can be no doubt that man is, on the whole, less good than he imagines himself or wants to be.
Everyone carries a shadow, and the less it is embodied in the individual’s conscious life, the blacker and denser it
is. At all counts, it forms an unconscious snag, thwarting our most well-meant intentions.”
– Carl Jung, Psychology and Religion: West and East
“The shadow is a moral problem that challenges the whole ego-personality, for no one can become conscious of the
shadow without considerable moral effort. To become conscious of it involves recognizing the dark aspects of the
personality as present and real. This act is the essential condition for any kind of self-knowledge.
– Carl Jung, Aion
The need for acknowledging the dark side within us and being alert to our moral weaknesses and moral
limitations is also an important theme in literature, along with the theme of recognizing the moral complexity of
others.
“The world isn't split into good people and Death Eaters. We've all got both light and dark inside us.”
- J.K. Rowling, Harry Potter and The Order of the Phoenix
"Many that live deserve death. And some that die deserve life. Can you give it to them? Then do not be too eager to
deal out death in judgment. For even the very wise cannot see all ends."
-J.R.R. Tolkien, The Lord of the Rings: The Fellowship of the Ring
A similar theme of moral complexity was invoked in the social messaging of civil rights movement leader, Martin
Luther King Jr.
“We must recognize that the evil deed of the enemy-neighbor, the thing that hurts, never quite expresses all that he
is. An element of goodness may be found even in our worst enemy …When we look beneath the surface, beneath the
impulsive deed, we see within our enemy-neighbor a measure of goodness and know that the viciousness and
evilness of his acts are not quite representative of all that he is… there is some good in the worst of us and some evil
in the best of us. When we discover this, we are less prone to hate our enemies”
-Martin Luther King Jr., Strength to Love
Finally, recently, Barack Obama, former US president, spoke against our tendency for moral righteousness and
condemnation, and encouraged seeing moral strengths in others.
“This idea of purity and you’re never compromised... you should get over that quickly. The world is messy, there are
ambiguities. People who do really good stuff, have flaws. People who you are fighting may love their kids and share
certain things with you. There is the view that the way for me to make change is to be as judgmental as possible
about other people, and that’s enough. That’s not bringing about change. If all you are doing is casting stones, you
are probably not going to get that far”.
5
-Barack Obama
Together, these examples convey how our stories, teachings, and writings urge us towards shining a light
on our moral weaknesses and flaws, so that we can grow morally, treat others with grace, and find goodness in them.
Thus, moral humility as conceptualized here, is consistent with ideas and teachings about morality embedded in our
culture.
What Moral Humility Is Not?
Moral Humility and Moral Relativism
One construct moral humility might be confused with is moral relativism. Moral relativism is a metaethical
viewpoint that entails believing that there are no universal or objective moral truths and what is morally appropriate
is dependent on the standards of a culture or group or person (Gowans, 2021). Does being morally humble mean
being a moral relativist? I will test this empirically. However, there is a least a conceptual distinction to be made
between these ideas. Moral relativism is a stance towards the objective nature of morality in general (do objective
moral truths exist), moral humility involves a stance towards the fallibility of one’s own morality (do I have moral
limitations).
Thus, a morally humble person can be so without being a moral relativist. Such a person might think, for
instance, that objectively true moral positions do exist, but they themself might not have accurate insight into it, or
that they can learn from others moral viewpoints to get at the objective truth. Or they might believe that they do
have knowledge of the correct moral perspective but fail to translate into appropriate moral behaviors. These
possibilities show that being morally humble does not necessitate believing there are no objectively correct moral
answers or ways. Thus, I conceptualize moral humility as distinct from moral relativism and adopt this distinction in
this project henceforth. However, I do concede that moral humility can and perhaps does psychologically co-occur
with moral relativism in people. For instance, Smith and Kouchaki (2018) suggest that too much moral humility
might manifest as moral relativism. Thus, these constructs are not psychologically incompatible, and hence I do not
argue that they are entirely distinct.
Moral Humility and Intellectual Humility
Intellectual humility is a type of humility that has recently received a lot of scholarly attention. It is
humility in the domain of knowledge or ideas (Krumrei-Mancuso & Rouse, 2016; Leary et al., 2017). That is, it is
the recognition of one’s intellectual or epistemic limitations (Porter et al., 2022). One might wonder, do we need the
6
moral humility construct to be a separate psychological construct, to be a different type of humility? Or is moral
humility just intellectual humility in the moral domain—i.e., humility about moral ideas and knowledge?
There are good reasons to think that there are important conceptual and empirical differences between these
two constructs. First, moral humility can go beyond just cognitive elements. For instance, when we make
judgements about our own moral imperfections, or when we deem others as morally better than us in some regard,
we are sometimes judging actions, emotions, or characters, as opposed to just knowledge and beliefs. Indeed, virtues
have been conceptualized to have multiple psychological components such as knowledge, behavior, motivation, and
disposition (Fowers et al., 2021).
Second, even when it comes to just knowledge and beliefs, research suggests that moral beliefs have unique
psychological signatures compared to non-moral beliefs — they are more laden with intense emotions, are more
rigid and extreme, and motivate more intolerance, inflexibility, and desire for punishment (Clifford, 2019; Delton et
al., 2020; Goodwin & Darley, 2008; Jung & Clifford, 2024; Kovacheff et al., 2018; Luttrell, et al., 2016; Pretus et
al., 2023; Ryan, 2017; Skitka et al., 2021; Skitka & Mullen, 2002; Van Bavel et al., 2012). In a similar vein,
morality is very central to people’s self-conception and how they interact and organize their social world, making it
an important and distinct aspect of psychological experience (Haidt & Kesebir, 2010; Haidt, 2012; Strohminger &
Nichols 2014; Stanley et al., 2020; Brambilla et al., 2021; Goodwin et al., 2014; Nicolas et al., 2022; Leach et al.,
2015; Halevy et al., 2015). Together, these reasons suggest that moral humility is at least somewhat conceptually
and empirically distinct from intellectual humility, and that these differences should drive us to study it as separate
from intellectual humility. This is the path taken in this project. However, I do concede at the outset that moral
humility has components that can be considered as aspects of intellectual humility. For example, humility about
one’s moral knowledge and beliefs can be considered intellectual humility about one’s moral knowledge. Thus, I do
not argue that they are entirely distinct.
What do we know or don’t know about moral humility so far?
Moral humility is thus far a largely empirically unexplored construct. One study investigated the effect of a
leader’s perceived moral humility on followers in an organizational context (Owens et al., 2019). Within two
different samples (in China and US), team members or employees reported their leader’s moral humility i.e. the
extent to which the team leader acknowledged their own moral limitations, were open to moral learning, and
recognized moral strengths in other team members. The team members’ ethical and prosocial behavior was also
7
measured (other-report or self-report). The study found that when team leaders were rated higher in moral humility
by the followers, the followers were reported to engage in more ethical and prosocial behavior. The authors
suggested that this effect was observed because of enhanced moral self-efficacy on the part of team members
because of team leader’s humility. This study suggested that moral humility can have positive interpersonal
outcomes on ethical and prosocial behavior. Further, the study found these effects over and above the effects of
general humility of leaders, suggesting that domain-specific forms of humility, like moral humility, can enhance the
predictive efficacy of the humility.
However, this study did not test people’s own moral humility and what implications this has on different
types of outcomes. It also used measures of moral humility that were not psychometrically validated. Vallabha et al.,
(2024) conducted the first test that we know of where people’s own moral humility was measured and used to
predict people’s level of affective polarization, support for political compromise, and other attitudes towards
political outgroups. Across three samples (national and student samples) in the US, participants reported their own
levels of moral humility on face-valid measures, as well as a host of political attitudes capturing political
polarization. People higher in moral humility were less inclined to be polarized i.e., they expressed lower negative
attitudes and behavioral intentions towards political outgroups. Because the current political relations between the
partisan groups in the United States (Republican and Democrats) have been characterized by moral disdain,
extremity, and aversion (Finkel et al., 2020, 2024; Garrett & Bankert, 2020; Enders & Lupton, 2021; McGarry et al.,
2023; Puryear et al., 2023), the authors reasoned that moral humility might help attenuate such tendencies. This
provided the first evidence suggesting that people’s own moral humility can have positive outcomes in conflictual
contexts rooted in moral divisions. However, like Owens et al. (2019), Vallabha et al. (2024), also used measures of
moral humility that were not psychometrically validated. Further, this work like others didn’t explore moral
humility’s nomological network, restricting insight into how it is related to other psychological constructs. Finally,
previous work also didn’t examine if and how moral humility could be changed, and whether it plays a causal role in
outcomes of interest. Thus, we have limited information on the nature of moral humility as a psychological construct
and its implications.
The Current Project
In the present project, I address the limitations of previous empirical work on moral humility and extend it
further. Specifically, I (i) develop a measure of moral humility using psychometric factor analytic methods (EFA
8
and CFA), (ii) probe the nomological network of moral humility, such as how it is related to personality, religiosity,
ideology, political extremity, and other relevant constructs, (iii) test it’s predictive and criteria validity in reducing
divisions in a morally relevant context, (iv) test its incremental validity in reducing divisions over other related
constructs like moral relativism and intellectual humility, and (v) develop interventions that change moral humility
and test its impact on political divisions. Studies addressing points i through iv were conducted before the
dissertation committee provided feedback, thus they are reported together below first in the “Chapter II”. After the
committee provided feedback, I extended this previous work in a causal direction and developed and tested moral
humility interventions. This is described after the previously conducted studies are reported and summarized; they
are in the “Chapter III”.
Context of the Study: Polarization in USA
The predictive/criteria validity of moral humility is tested in the context of present-day political
polarization in the USA. Although polarization has many meanings, in this project it is used to describe and measure
the phenomena of animus or antipathy towards political outgroups, which has significantly increased in the US over
the last two decades (Iyengar et al., 2019; Finkel et al., 2020; Mason, 2015; Druckman & Levy, 2022). The rise in
partisan antipathy has raised alarm and spurred investigations on understanding its nature and ways to reduce it. The
primary reason for choosing this as the context in which I study moral humility is because this type of polarization
provides us with a morally relevant context that has the features associated with the dark aspects of our morality
described above.
Specifically, this type of polarization has been described as a “quasi-religious phenomenon” (Finkel et al.,
2024), where people have affective attachments akin to religious attachments to their political ingroup, believe in the
moral righteousness and supremacy of their political side, judge the other side as immoral, and dehumanize political
outgroups (Cassese, 2021; Finkel et al., 2020 Finkel et al., 2024; Lees & Cikara, 2020; Martheus et al., 2021;
McGarry et al., 2023; Puryear et al., 2023; Tappin & McKay, 2019). Consistent with this, research shows that the
more people engaged with politics in moral terms, the greater aversion they expressed toward the political outgroup,
rejected political compromise, and punished politicians who engaged in compromise (Garrett & Bankert, 2020;
Grubbs et al. 2020; Ryan, 2017). Political animus has also been associated with political extremity (Brandt &
Vallabha, 2024), selective and partisan engagement with information (Hobolt et al., 2023; Levy, 2021), support for
9
political violence (Kalmoe & Mason, 2022) and anti-democratic attitudes and behaviors (Graham & Svolik, 2020;
Kingzette et al., 2021; McCoy et al., 2018; Simonovits et al., 2022; McCoy & Somer, 2019).
Thus, political polarization in the US represents a morally fraught conflict that has been associated with
threats to social cohesion and democratic commitments and procedures. Accordingly, it thus provides an apt context
for the study moral humility. Theoretically, the expectation is that moral humility will show positive outcomes in
any moral situation that brings out of the dark aspects of our moral nature. However, here I only test moral
humility’s role in a polarized and moralized political context as a case study of moral humility’s effects.
10
Chapter II: Scale Development, Nomological Associations, Predictive Validity
Seven studies were conducted wherein the main aims were to (i) create a moral humility scale, (ii) probe its
nomological network, and (iii) test the scale’s predictive and criteria, and incremental validity. The studies and
results are described in the order of these three broader questions that I aimed to address. First, all the samples that
went into addressing these three broad aims are described together. Then factor analysis, nomological associations,
and predictive/criteria validity tests are described in three separate sections, each section accompanied by the
description of research design, methodology, measures, and results associated with that section. Overall, the aim of
the studies was to understand the nature of moral humility, its nomological network, and test its mitigatory role in
political polarization.
Dataset and Participants
The studies were conducted on seven adult American samples, including samples from Prolific, Michigan
State University psychology student pool, and YouGov national samples. The details of the different samples are in
Table 1. Prolific is an online service that facilitates the crowdsourcing of research participants (Douglas et al., 2023;
Peer et al., 2022). YouGov samples were collected as part of national surveys conducted by the Polarization
Research Lab (Sample 4) and CCES (Cooperative Election Study, Sample 6).
Participants were paid $0.85 for doing the study (~5 minutes) in Sample 1, $1.68 in Sample 2 (~10
minutes), $1.68 in Sample 3 (~10 minutes), 0.50 research credits in Sample 5 (~20 minutes), and $2.20 in Sample 7
(~15 minutes). In the Prolific samples (Samples 1, 2, 3, 7), participants were recruited evenly from those who self-
identified as Democrats and Republicans in Prolific’ s prescreening to circumvent the problem of samples collected
online being liberally biased.
Sample size justifications for Sample 3-7 are in “Predicting Polarization Outcomes” section later as the
sample size for those samples were determined based on estimates of relationship with political outcomes. Sample 1
was collected primarily for EFA and Sample 2 for CFA and exploring nomological associations. In general, for EFA
and CFA, larger samples are better, and all the samples are > 500. A sample size of 800 in Sample 1 gave between a
very good to excellent sample size for an EFA (Comrey & Lee, 1992). 800 also gives about a 17:1 sample to item
ratio which is above the 15:1 recommended as best practices (Pett et al., 2003). Along similar lines, a sample size of
500 for CFA in Sample 2 also gave a very good sample size for factor analysis (Comrey & Lee, 1992) as well as an
80% power to detect small correlations of r ~ .1, useful for probing nomological associations.
11
Table 1
Sample characteristics
Sample
Number
Sample
Platform
Scale
Used
N
Mage
SDage
1
2
3
4
5
6
7
Full
Prolific
Full
Prolific
Prolific
Full
YouGov Short
Student
Full
YouGov Short
Full
Prolific
823
518
1499
1000
930
1000
805
24.43
26.84
44.94
49.67
19.52
50.97
23.13
13.97
14.5
14.2
17.33
1.5
17.56
13.07
%
Men
46.9
44.5
49.03
46.4
24.01
44
50
% Women
%
White
%
Black
%
Other Ethnic
46.29
48.26
48.76
53.6
72.95
55.2
47.64
73.39
76.83
72.18
64.9
65.99
66
74.32
9.6
6.17
10.14
10.8
7.06
14
7.94
17.01
17
17.68
24.3
26.95
20
17.74
Transparency and Openness
Sample 4 and 6’s study design, hypotheses, and planned analyses were preregistered and can be found at
OSF. The data, code, materials, and SOM for all studies is also at OSF. All studies were approved by the Michigan
State University Institutional Review Board.
Scale Development: Exploratory Factor Analysis
Measures
Moral Humility
Exploratory factor analysis was conducted in Sample 1. 45 items capturing various aspects of moral
humility such as moral fallibility, moral limitations, moral openness, moral learning, moral disdain, and moral
righteousness were constructed. To this end, items from other humility scales (e.g., McLaughlin et al., 2023) were
adapted, items from previous moral humility work (Vallabha et al., 2024) borrowed, and new items generated based
on previous theoretical work (e.g., Smith & Kouchaki, 2018). The full set of items are included in Table 2.
Participants were instructed: “Take a moment and think about your strongly held moral beliefs and values
that help you navigate moral issues and concerns in everyday life and in society. When you think about your moral
views, ideas, and values, please answer the following items about how you see yourself and other people.”1 They
1 About half the participants received some additional instructions telling them about various domains and aspects of
morality people care about (see SOM) before they were saw these instructions. This was done as we were testing if
the two sets of instructions would yield different factor solutions, but after observing that they yielded similar factor
solutions, we proceeded to treat and analyze them as one measure.
12
then indicated their response on the 45 items on a 7-point scale (1 = Strongly Disagree, 2= Moderately Disagree,
3= Slightly Disagree, 4= Neutral, 5= Slightly Agree, 6= Moderately Agree, 7 = Strongly Agree).
Analysis and Results
In the first step of exploratory factor analysis, the number of factors to extract was determined using a scree
plot (Figure 1), parallel analysis, and a priori theoretical considerations of moral humility content. Parallel analysis
suggested six factors, but a scree plot showed that there were three or four most distinct factors or components, with
other factors or components being less distinct. Additionally, there weren’t a priori reasons to expect many (>4)
distinct factors. Thus, based on these theoretical and empirical considerations, three and four factors were decided
upon to be extracted for further consideration.2 Three factors accounted for 48% of the total variance, while the
fourth factor added only 4% additional variance.
In the second step, three and four factors were extracted using minimum residual factor analysis and rotated
with oblimin rotation, thus allowing factors to be correlated. In the third step, the number of items were reduced by
identifying items that clearly and strongly loaded onto a single factor. For that purpose, factor loadings less than 0.3
(indicated by blank spaces in Table 2) were removed. Then, how the items performed in both the three and four
factor solutions was compared and finally only those items that consistently loaded on the same factor and did not
load on multiple factors across the solutions were picked.
This resulted in three coherent factors that were named as Moral Fallibility, Moral Learning/Openness, and
Moral Superiority. The fourth factor wasn’t coherent as it consisted of mostly smaller loadings (<0.3) or cross-
loadings. The items chosen for the three factors using the aforementioned criteria are highlighted in bold in Table 2.
The final chosen factors and the respective items are also separately shown in Table 3.
The 30-items that were retained for the Moral Humility scale had an overall reliability of 0.95 and excellent
subscale reliabilities (Moral Learning/Openness α = 0.95, Moral Fallibility α = 0.94, Moral Superiority α = 0.91).
2 I did investigate the results of five and six factor solutions. No coherent themes seemed to emerge from additional
factors beyond four. Additionally, the fifth and sixth factors accounted for only 4% and 3% variance, which
provided further rationale to proceed with three and four factor solutions for further analysis.
13
Figure 1
Scree plot of eigen values (Sample 1)
Table 2
Standardized Loadings with Minimum Residual Factor Analysis with Oblimin rotation (Sample 1)
3 factors
F1
F2
F3
0.342
0.379
0.392
1
2
3
I would never change what I
believe about moral topics.
I doubt I would change my mind
on moral issues.
I don’t think there is anything
that could happen that would
shift how I think about moral
issues.
4 My mind is settled about moral
0.305
0.412
issues.
5 My beliefs about moral issues
0.732
4 factors
F2
F3
F4
0.324
F1
0.450
0.535
0.502
0.506
0.756
6
7
8
9
may be incorrect. f
I would be willing to revise what
I believe about moral topics.
I realize that my perspective
about moral issues may be
wrong. f
I understand that my moral
beliefs about these statements
may be limited. f
I’d be excited to learn from
others about moral issues.*
0.428
0.470
0.598
0.346
0.692
0.578
0.760
0.464
0.703
0.761
14
Table 2 (cont’d)
10
11
I enjoy hearing diverse
perspectives about moral
issues. *
I am interested to learn how
other people think about moral
issues. *
I am open to exploring moral
topics more in the future. *
13 My moral ideas are much more
12
just and fair than the ideas of
those who disagree with me. †
14 Thoughts and behavior that
conflict with my ideas about
morality are wrong. †
15 The moral values of those who
disagree with me on moral
issues are probably misguided.
†
17
16 Some people hold moral views
that are so horrible, there is no
point listening to them. †
I don’t expect others to adopt
the same ideas of right and
wrong as me, because there
will always be a diversity of
moral viewpoints in the world.
*
I have a better grasp on
important moral topics than
those who disagree with me. †
I respect that there are ways of
thinking about moral issues
that are different from mine. *
18
19
20 When another person
disagrees with me on moral
issues, it is highly likely that
they have a mistaken view. †
I welcome different ways of
thinking about important
moral topics. *
21
22 My views about moral issues
0.483
0.668
23
are just as likely to be wrong as
other views. f
I recognize that my views
about moral issues are based
on limited evidence. f
24 Although I have particular views
about moral issues, I realize that
I don't know everything that I
need to know about it.
It is quite likely that there are
gaps in my understanding
about moral issues. f
25
0.719
0.663
0.662
0.756
0.656
0.678
0.406
0.740
0.742
0.717
0.703
0.583
0.588
0.353
0.362
0.420
0.794
0.765
0.588
0.627
0.743
0.666
0.679
0.706
0.599
0.713
0.412
0.343
0.422
0.678
0.596
15
Table 2 (cont’d)
26 My sources for information
27
about moral issues might not
be the best. f
I am open to new information
in the area of moral issues that
might change my view. *
28 My views about moral issues
today may someday turn out to
be wrong. f
29 When it comes to my views
about moral issues, I may be
overlooking evidence. f
30 My views about moral issues
0.725
0.783
0.664
0.611
0.633
0.698
0.721
0.696
0.401
0.465
0.514
0.373
0.549
0.641
32
31
may change with additional
evidence or information.
I realize that I fall short of my
own moral standards, and that
there is room for improvement.
I am not always able to act in
accordance with my values and
principles.
I consistently live up to my
moral values and principles.
34 My behavior and actions are
consistently in line with my
moral beliefs.
33
35 Moral values and principles
that I disagree with may be
more appropriate in important
moral situations. f
36 Moral values and principles that
I hold are usually appropriate for
most important moral decisions.
0.351
0.553
0.527
0.361
0.302
0.522
-0.49
0.468
-0.487
-0.34
0.332
-0.402
0.372
0.465
-0.342
0.356
0.344
-0.427
0.307
37 My moral values and
0.734
0.640
38
39
principles may be flawed and
incomplete. f
I am open to learning about
different moral values from
people I typically disagree
with. *
I am a more virtuous and
righteous person than most
others. †
I have a stronger sense of what
is right and wrong than most
people. †
I am more committed to moral
values and ethical behavior
than most people. †
42 My moral choices and
41
40
behaviors contribute more to
the greater good than most
people's behavior. †
0.668
0.695
0.715
0.732
0.751
0.700
0.603
0.684
0.713
0.684
16
Table 2 (cont’d)
43
I am not confident that my
moral choices have a positive
impact on the world. f
I don’t always make the right
moral decisions.
I have much to learn from other
people’s moral choices and
behaviors.
44
45
0.339
0.617
0.380
0.571
0.328
0.501
0.575
Note: Moral Fallibility (marked with f signs): Items 5,7,8, 22, 23, 25, 26, 28, 29, 35, 37, 43.
Moral Learning/Openness (marked with * sign): Items 9, 10, 11, 12, 17, 19, 21, 27, 38.
Moral Superiority (marked with † sign): Items 13, 14, 15, 16, 18, 20, 39, 40, 41, 42.
Factor loadings less than 0.3 are indicated by leaving space blank.
Table 3
Final extracted three factors with corresponding items
Moral Fallibility
My beliefs about moral issues may be incorrect.
I realize that my perspective about moral issues may be wrong.
My views about moral issues are just as likely to be wrong as other views. †
I recognize that my views about moral issues are based on limited evidence.
It is quite likely that there are gaps in my understanding about moral issues.
My sources for information about moral issues might not be the best.
When it comes to my views about moral issues, I may be overlooking evidence. †f
My views about moral issues today may someday turn out to be wrong. †f
Moral values and principles that I disagree with may be more appropriate in important moral situations.
My moral values and principles may be flawed and incomplete.
I am not confident that my moral choices have a positive impact on the world.
Moral Openness & Learning
I'd be excited to learn from others about moral issues.
I enjoy hearing diverse perspectives about moral issues. †
I am interested to learn how other people think about moral issues.
I am open to exploring moral topics more in the future.
I respect that there are ways of thinking about moral issues that are different from mine. †f
I welcome different ways of thinking about important moral topics.
I am open to learning about different moral values from people I typically disagree with. †f
I am open to new information in the area of moral issues that might change my view.
I don't expect others to adopt the same ideas of right and wrong as me, because there will always be a
diversity of moral viewpoints in the world.
Moral Superiority
My moral ideas are much more just and fair than the ideas of those who disagree with me. †f
Thoughts and behavior that conflict with my ideas about morality are wrong.
Some people hold moral views that are so horrible, there is no point listening to them.
The moral values of those who disagree with me on moral issues are probably misguided.
I have a better grasp on important moral topics than those who disagree with me.
When another person disagrees with me on moral issues, it is highly likely that they have a mistaken
view. f*
I am a more virtuous and righteous person than most others.
I have a stronger sense of what is right and wrong than most people.
I am more committed to moral values and ethical behavior than most people.
My moral choices and behaviors contribute more to the greater good than most people's behavior. †
1
2
3
4
5
6
7
8
9
10
11
1
2
3
4
5
6
7
8
9
1
2
3
4
5
6
7
8
9
10
17
Table 3 (cont’d)
Note: Items included in Sample 4 are marked with † those included in Sample 6 are marked with f . Items included
in Sample 4 were also used in Experiment 1 and 2 in Section II, with the addition of one more item marked with *.
An item “I understand that my moral beliefs about these statements may be limited” was additionally extracted for
moral fallibility but was removed in later studies (Sample 5) after realizing that it had weird wording, resulting in a
final 30-item measure. This item was thus included in survey in Sample 2, 3, and 7 but not included in any analysis.
Scale Development: Confirmatory Factor Analysis
Measures
Moral Humility
The full 30-item measure extracted using EFA (in Table 3) was included in four subsequent samples
(Samples 2, 3, 5, 7) and Confirmatory Factor Analysis (CFA) was performed on each of them.
Analysis and Results
The three-factor structure was replicated using Confirmatory Factor Analysis (CFA). 3 Additionally, the
three-factor structure was also tested against two and one factor structure. A higher order structure was also
explored.
CFA were done using the lavaan package for R Statistical Software (Rosseel, 2012). Structural equation
modeling based on maximum likelihood estimation was used to estimate and compare the three-factor model with
one-factor and two-factor models. That is, three models were tested: one factor model, two factor model, and three
factor model. The one-factor model had all items loading on a single factor, a two-factor model had positively
worded items (all items from moral learning/openness and moral fallibility subscales) and negatively worded items
(all items from moral superiority subscale) loading on two separate factors, and a three-factor model had the items
load on the factors identified in Sample 1. A hierarchical model was also tested in which all three factors loaded on a
higher order factor.
The fit statistics are noted in Table 4 for different samples. Smaller relative chi square (c2/df, relevant
statistics noted in first three columns) 4, higher CFI and TLI (≥.9), smaller RMSEA and SRMR (<.09), and smaller
AIC and BIC indicate better fit (Bentler 1990; Hu & Bentler, 1999; Schumacker & Lomax, 2010; Hooper et al.,
3 Samples 4 and 6 used a short version of the moral humility scale. Because each factor did not have 3 or more
items, I did not perform a CFA.
4 Some consider significance value of the chi square test as the criteria important to judge model fit wherein non-
significant p-values are considered to indicate better model fit. However, p-value has been found to be heavily
biased by sample size, with bigger samples almost always leading to significant (p <.05) values, hence not providing
good information on model fit. To circumvent this problem, relative chi-square value (c2/df ) that adjusts for sample
size has been suggested as a better criterion for evaluation for model fit where lower is better.
18
2008). In line with this, it was observed that the three-factor model and hierarchical model were a better fit across
samples compared to the two and one factor models. Consistently, chi-square difference test, used to compare the fit
of two models, was also significant and indicated that the three-factor and higher order solution were better fit
compared to the one and two factor solution across all studies. Table 5 has the model comparisons. The results for
the hierarchical model with all three factors loading on a higher order factor showed similar fit indices as the three-
factor model (the rows named “Higher” under each sample in Table 4). 5 Because the hierarchical model fits the
intended conceptualization and use of moral humility measure as a general scale with subscales, this was the model
of preference. A good model fit of this higher order model lends confidence to such a conceptualization and
application. All items in the three-factor and higher order models had loadings of >.3 on their respective factors (see
SOM).
Finally, the correlations between the three factors from CFA are in Table 6. The three factors were all
significantly correlated (p <.001). Moral Openness/Learning was positively correlated with Moral Fallibility (Mr=
.66 across samples) and negatively correlated with Moral Superiority (Mr= -.23). Moral Fallibility was negatively
correlated with Moral Superiority (Mr= -.41).
Summary of Scale Development
The exploratory factor analysis identified a thirty-item scale with three factors or subscales: moral fallibility, moral
learning/openness, and moral superiority. The confirmatory factor analysis (a) replicated the three-factor structure of
moral humility across different samples, (b) showed that the three-factor structure as well as the hierarchical
structure with the three factors loading onto a higher order factor had decent fit, and significantly better fit than one
or two factor structures, and (c) showed that the three factors are significantly correlated with each other, with moral
learning/openness and fallibility correlated most, followed by moral fallibility and moral superiority, and then moral
learning/openness and moral superiority. Although the three-factor and hierarchical models showed decent fit across
samples, there remains room for improvement in the fit indices (e.g., >.95 CFI and TLI), something that can be
explored in future work.
5 The model showed negative variance for the moral fallibility factor in all samples, so the error variance of the
factor was fixed to zero.
19
χ2
1 Factor
4748.37
2 Factor
2861.89
3 Factor
1202.43
Higher
1203.06
df
405
404
402
403
p
c2/df
CFI
TLI RMSEA SRMR
AIC
BIC
Sample 2
11.52
0.585 0.554
0.149
0.765 0.747
0.112
0.924 0.917
0.064
0.924 0.917
0.064
Sample 3
0.151
0.096
0.070
0.070
49256.50
49507.18
47372.02
47626.87
45716.56
45979.77
45715.19
45974.23
CFI
TLI RMSEA SRMR
AIC
BIC
0.634
0.607
0.133
0.133
140721.04 141036.33
0.788
0.772
0.102
0.088
136424.99 136745.54
0.909
0.902
0.067
0.073
133051.21 133382.27
0.909
0.902
0.067
0.073
133053.68 133379.49
Sample 5
CFI
TLI
RMSEA SRMR
AIC
BIC
0.535
0.501
0.121
0.141
77301.17 77583.36
0.741
0.721
0.090
0.098
75179.97 75466.86
0.892
0.883
0.058
0.067
73621.48 73917.79
0.891
0.882
0.059
0.070
73631.01 73922.61
Sample 7
Table 4
Robust fit indices for Samples 2,3,6, and 7
<.001
<.001
<.001
<.001
<.001
<.001
<.001
p
<.001
<.001
<.001
<.001
7.07
2.99
2.99
c2/df
26.19
15.61
7.29
7.28
c2/df
13.6
8.2
4.11
4.14
χ2
df
p
1 Factor 10606.37 405
<.001
2 Factor 6308.32
3 Factor 2930.54
Higher
2935.01
χ2
1 Factor 5208.35
2 Factor 3085.14
3 Factor 1522.66
Higher
1534.19
χ2
1 Factor 6265.17
2 Factor 3714.35
3 Factor 1786.60
Higher
1789.83
404
402
403
df
405
404
402
403
df
405
404
402
403
p
c2/df
CFI
TLI
RMSEA SRMR
AIC
BIC
<.001
<.001
<.001
<.001
15.47
0.608
0.579
0.135
0.138
78414.31
78695.16
9.19
4.44
4.44
0.779
0.762
0.101
0.094
75865.49
76151.02
0.907
0.900
0.066
0.072
73941.74
74236.63
0.907
0.900
0.066
0.072
73942.97
74233.18
Note: Chi square difference test was significant; three factor and higher order models were significantly better than
two and one factor CFA models. See Table 5.
20
Table 5
Model Comparison for Samples 2,3,6, and 7
Higher vs 1-factor
χ2 Diff
df
p
ΔCFI
ΔTLI
ΔRMSEA ΔSRMR
ΔAIC
ΔBIC
Sample 2
3545.3
2 <.001
-0.339
Sample 3
7671.4
2 <.001
-0.275
Sample 5
3674.2
2 <.001
-0.355
Sample 7
4475.3
2 <.001
-0.299
-0.363
-0.295
-0.381
-0.321
0.085
0.067
0.062
0.069
0.081
0.061
0.071
0.062
3541.306
3532.95
7667.354
7656.844
3670.161
3660.755
4471.34
4461.978
Higher vs 2-factor
χ2 Diff
df
p
Δ CFI
Δ TLI
Sample 2
Sample 3
1658.8
3373.3
1 <.001
1 <.001
Sample 5
1551.1
1 <.001
Sample 7
1924.5
1 <.001
-0.158
-0.121
-0.150
-0.129
-0.17
-0.13
-0.161
-0.138
Δ
RMSEA
0.048
0.035
0.032
0.036
Δ SRMR
Δ AIC
Δ BIC
0.026
0.015
0.027
0.017
1656.822
3371.308
1652.644
3366.053
1548.957
1544.254
1922.52
1917.84
Note: The 3-factor model had similar model comparisons with 1 and 2 factor model as the hierarchical models.
Therefore, only the model comparisons with the hierarchical model is shown for space consideration.
Table 6
Correlations between the three Moral Humility factors or subscales (Moral Learning/Openness, Moral Fallibility,
Moral Superiority) in Samples 2,3,6, and 7
Moral Fallibility
Moral Superiority
Moral Fallibility
Moral Superiority
Moral Fallibility
Moral Superiority
Moral Fallibility
Moral Superiority
Moral Fallibility
-0.44
Moral Fallibility
-0.42
Moral Fallibility
-0.33
Moral Fallibility
-0.45
Sample 2
Moral Openness/Learning
0.66
-0.27
Sample 3
Moral Openness/Learning
0.73
-0.27
Sample 5
Moral Openness/Learning
0.57
-0.10
Sample 7
Moral Openness/Learning
0.68
-0.26
21
Nomological Associations
Having constructed a moral humility scale and replicated the factor structure, the next aim was to explore
the scale’s association with other psychological constructs. Specifically, I tested if it is related to other constructs in
theoretically expected and sensible ways. For instance, I expected the moral humility scale to be positively
associated with constructs that capture humility (e.g., modesty, intellectual humility), open-mindedness and
flexibility (e.g., openness to experience, intellectual humility, need for cognition), and positive other-orientedness
(e.g., more agreeableness, less psychopathy, less narcissism). I also expected it to be negatively associated with
measures that captured absolutism and extremism (e.g. political extremity, religious exclusiveness, moral
grandstanding, moral absolutism).
To capture these nomological associations, personality (HEXACO, BFI), dark triad (psychopathy,
Machiavellianism, narcissism), self-esteem, political identity (partisan and ideological identity), political extremity
(partisan and ideological extremity), religiosity, religious exclusivism, moral grandstanding, moral relativism,
attitude-specific moral conviction, intellectual humility, and need for cognition were measured. These measures
were included in different samples and sometimes in more than one sample. Results are presented together in this
subsection from across the samples. Example items are given below for the measures but see SOM for the full list of
items wordings.
Measures
Personality, Dark Triad, and Self-Esteem
Measures of personality were included as the expectation was that moral humility will be positively
associated with the modesty/humility and openness to experience personality traits, given the overlap of different
moral humility facets with these constructs. I also expected moral humility to be negatively associated with Dark
Triad traits as moral humility involves positive other-orientedness, such as lack of moral superiority or moral disdain
towards others. Specifically, narcissism also has associations with ego and pride (Miller et al., 2021; Tracy et al.,
2009), which I especially expected to be negatively associated with moral humility. Self-esteem was included as I
wanted to test that moral humility is not a lack of self-esteem, given that humility is sometimes confused with low
self-regard (Tangney, 2000).
Sample 2 included the short 60-item HEXACO measure (Ashton & Lee, 2009) used to assess the six
personality dimensions of honesty-humility (α = 0.80), emotionality (α = 0.81), extraversion (α = 0.85),
22
agreeableness (α = 0.81), conscientiousness (α = 0.82), and openness to experience (α = 0.84). Sample 5 included
the short 15-item Big Five Inventory (BFI-S; Soto & John, 2017) used to assess the five personality dimensions of
extraversion (α = 0.63), agreeableness (α = 0.61), conscientiousness (α = 0.45), openness to experience (α = 0.61),
and neuroticism (α = 0.67). Additionally, Sample 5 also included the 12-item Dark Triad scale assessing
psychopathy (α = 0.76), Machiavellianism (α = 0.75), narcissism (α = 0.73) (Jonason & Webster, 2010), and 1-item
self-esteem scale (Robins et al., 2001). All measures were reported on a 7-point measure (1 = strongly disagree, 7 =
strongly agree).
Political Identification and Political Extremity
Measures of political identification and extremity were included because these constructs have been
implicated in political animus in previous work (Brandt & Vallabha, 2024; Ganzach & Schul, 2021). I was
especially interested in moral humility’s relationship with political extremity given that I expected negative
associations between moral humility and constructs that captured an orientation towards extremism. This is because
moral humility is a construct that taps into a willingness to accept one’s limitations or a willingness to be modest in
one’s own abilities and knowledge, and a willingness to be open to change—these are orientations that one would
expect to be antithetical to extreme attitudes, positions, and values.
To assess political extremity, partisan and ideological extremity was measured. To assess political identity,
partisan and ideological identity was measured. In all samples, partisan identity was measured using the 7-point
partisanship scale where higher values meant stronger Republican identification (where 1 = Strong Democrat and 7
= Strong Republican). In all samples except YouGov Sample 4, ideological identity was measured using the 7-point
ideology scale where higher values meant stronger conservative identification (1 = Very Liberal and 7 = Very
Conservative). In YouGov Sample 4, it was instead measured using a 5-point scale. In multiple samples,
partisanship and/or ideology scale also had other options like “Not sure”, “Don’t know” and “Haven’t thought much
about it”. In all such cases, participants who chose these options were rescored to the midpoint of the scale.
Partisan extremity was calculated by folding the partisanship scale at midpoint and forming a 4-point
measure where higher value represented stronger partisan identification. Ideological extremity was calculated by
folding the ideology scale at midpoint and forming a 4-point measure where higher value was strong ideological
identification (except in YouGov Sample 4 where it was a 3-point measure).
23
Religiosity and Religious Exclusivism/Inflexibility
Religiosity and religious exclusivism/inflexibility was included as I expected moral humility to be
negatively associated with both. I expected negative associations with religiosity as religious worldviews can be
accompanied by rigidity and religion is a domain that is often linked with absolute notions of morality (Zmigrod et
al., 2019; Hare, 2019; Brandt & Reyna, 2010; Hill et al., 2010). I expected negative associations with religious
exclusivism given that religion often informs religious people’s moral convictions — a person who considers their
religion or religious views to be superior to others’ might likely also consider themselves morally superior and be
low in moral openness and fallibility.
In Sample 2, religiosity was measured using 6 items borrowed from Project Implicit (Schmidt et al., 2023)
and another religiosity scale (Plante & Boccaccini, 1997) which were averaged to create a scale (α = 0.98). Example
items were, “I am a religious person.”, “My religious faith is extremely important to me.” (1 = strongly disagree, 7
= strongly agree).
In Sample 2, religious exclusivism (or religious inflexibility) was measured using 7 face-valid items that
were borrowed from Project Implicit (Schmidt et al., 2023) or self-written. The items were averaged to create a scale
(α = 0.73). The measure intended to capture whether the participants held rigid or exclusivist religious attitudes
(only one/my religion can be true). Example items were, “Everyone should have the same religious views that I
have.”, “There are many different religions, but only one can be true.” (reverse coded). They were measured on a 7-
point scale (1 = strongly disagree, 7 = strongly agree).
Moral Grandstanding
Moral grandstanding was included as it is a construct linked with morality that has been recently
investigated in the context of political polarization. It has been found to be associated with more political animus,
extremity, and conflicts (Grubbs et al., 2019, 2020). Moral grandstanding is understood as the use of moral speech to
gain social status (Tosi & Warmke, 2020). A moral grandstander expresses a moral claim with the intention of being
recognized by others for their moral qualities. As evident, this is a different construct from moral humility as moral
grandstanding involves both status considerations and publicity consideration. Neither is part of the
conceptualization of moral humility. Nevertheless, moral grandstanding may be a symptom of lower moral humility,
especially considering its previous associations with both extremity and disagreeableness. Thus, I expected a
negative association between moral humility and moral grandstanding.
24
Moral grandstanding was included in Samples 2 (α = 0.86) and 5 (α = 0.77). It was measured using 10
items on a 7-point scale (1 = strongly disagree, 7 = strongly agree) using a previously validated scale (Grubbs et al.,
2020). Example items were “I often share my moral/political beliefs in the hope of inspiring people to be more
passionate about their beliefs.”, “When I share my moral/political beliefs, I do so to show people who disagree with
me that I am better than them”.
Moral Relativism
Moral relativism was included because it has also been associated with moral and political tolerance
(Wright & Pölzler, 2022; Conrique, 2020; Collier-Spruel et al., 2019). I conceptualized moral humility and moral
relativism to be distinct. However, these might nevertheless be psychologically congruent for some people. Indeed,
previous theoretical work has suggested that very high levels of moral humility might manifest as moral relativism
(Smith & Kouchaki, 2018). For these reasons, I expected moral humility to be positively associated with moral
relativism.
Moral relativism was included in Samples 2 (α = 0.91), 3 (α = 0.87), 5 (α = 0.85), and 7 (α = 0.90). It was
measured using 6 items on a 7-point scale (1 = strongly disagree, 7 = strongly agree) using a previously validated
scale (Conrique, 2020). Example items were, “There is no absolute standard in morality”, “What is morally good in
one context may be morally bad in another”.
Moral Conviction
Moral conviction was included as moral humility might be mistaken with a lack of moral conviction. Moral
conviction, or the extent to which beliefs are moralized or treated as a matter of good/bad or right/wrong has been
associated with intolerance (Skitka et al., 2021) and hence may be related to moral humility. However, moral
conviction and moral humility are conceptually distinct, such that having moral humility doesn’t necessarily mean
not having moralized beliefs. According to my conceptualization of moral humility, a person can have moralized
beliefs yet be humble about them.
Moral conviction was measured differently across samples 5, 6, and 7. Moral conviction was measured on
different issues in each sample as each sample had a different main task (see the predictive validity section below)
which assessed different issues for the purpose of the task. The tasks are not described in this section, only the
relevant questions and measures are.
25
In Sample 5, participants answered their position on three randomly assigned political issues (of eleven
possible issues). These issues included many topics that are moralized in politics, for example, abortion, guns,
transgender rights (see SOM for full list). After answering their position on the issue, they were asked to report their
moral conviction on each issue. For example, if they indicated their stance on abortion, they were asked of their
moral conviction on abortion with “To what extent is your position on the issue of abortion a reflection of your core
moral beliefs and convictions?” (1= Not at All, 7 = Very Much). Moral conviction on the three issues were averaged
to create a moral conviction measure (α = 0.78).
In Sample 6, participants were asked to choose one of thirteen issues that was most important for them.
These issues were selected on the basis of the results of a national survey run by YouGov previously in 2023 on the
issues Americans found most important (Frankovic et al., 2023), such as, healthcare, immigration, etc. (see SOM for
full list). After selecting their most important issue, participants were asked about their moral conviction on the issue
with “To what extent is your position on [their chosen issue] a reflection of your core moral beliefs and
convictions?” (1= Not at All, 7= Very Much).
In Sample 7, participant’s moral conviction was assessed on five issues that were the basis of the main task
in that sample. These issues were guns, abortion, gender equality, racial equality, and immigration. Participants
were asked, “To what extent is your position on each of the following issues a reflection of your core moral beliefs
and convictions?” (1= Not at All, 7= Very Much). Moral conviction on the five issues was averaged to create a
measure of overall moral conviction (α = 0.79).
Intellectual Humility and Need for Cognition
Intellectual humility was included as it is a type of humility that has been associated with lower levels of
political polarization (Bowes, et al., 2020; Hoyle et al., 2016; Leary et al., 2017) and is also probably one of the
constructs which overlaps the most with moral humility. As outlined in the introduction, I conceptualized
intellectual humility and moral humility to be conceptually distinct to some extent. However, these are overlapping
constructs consisting of dimensions associated with humility (e.g., openness, acknowledgement of fallibility). Thus,
one can imagine a person who is generally humble in many domains, i.e., a person who is humble about their
intellectual attributes and also humble about moral attributes. For these reasons, I expected moral humility to be
positively associated with intellectual humility.
26
Intellectual humility was included in Samples 2 (α = 0.91), 3 (α = 0. 87), 5 (α = 0. 87), and 7 (α = 0.86). It
was measured using 6 items on a 7-point scale (1 = not at all true or characteristic of me, 7 = extremely true or
characteristic of me) using a previously validated scale (Leary et al., 2017). Example items were “I question my
own opinions, positions, and viewpoints because they could be wrong.”, “I like finding out new information that
differs from what I already think is true.”
Need for cognition is another construct related to people’s epistemic tendencies, specifically one’s
enjoyment of activities that require thinking. It was included given that moral humility has aspects of open
mindedness and learning, both of which have been associated with need for cognition along with lower dogmatism
(Cacioppo & Petty, 1982; Furnham & Thorne, 2013; Olson et al., 1984; Liu & Nesbit, 2023). The Need for
Cognition scale was used (α = 0.70; Coelho et al., 2020) in Sample 5, assessing 6 items6 on a 7-point scale (1 =
strongly disagree, 7 = strongly agree). Example items were, “I would prefer complex to simple problems.”, “I like to
have the responsibility of handling a situation that requires a lot of thinking.”
Analysis and Results
Before estimating correlations of the total moral humility scale with other constructs, the items from moral
superiority subscale were reverse scored and combined with items from moral fallibility and moral
openness/learning subscales such that higher values on the moral humility scale indicated overall higher moral
humility. Higher values on the moral fallibility subscale indicated higher moral fallibility, higher values on the
moral superiority subscale indicated higher moral superiority, higher values on the moral learning/openness subscale
indicated higher moral openness/learning.
For any measure that was assessed in the same way in multiple samples (e.g., partisan extremity,
intellectual humility), I present a meta-analytic estimate pooled from data across samples and computed using meta
package in R (Balduzzi, Rücker, and Schwarzer, 2019). For measures that were assessed just once or differently
across sample (moral conviction, personality), estimates from the relevant samples is presented.
Personality, Dark Triad, and Self-Esteem
The correlations between moral humility scale and its subscales with HEXACO, BFI, Dark Triad, and self-
esteem from Samples 2 and 5 are in Table 7. The correlations show that higher moral humility is significantly
6 Two items were contributing to negative reliability for the scale due to its negative correlations with other items, so
they were removed, and final analyses were done using the rest four items.
27
associated with higher agreeableness with mean correlation of Mr= 0.14, and higher openness to experience with
mean correlation of Mr = 0.22, across HEXACO and BFI. These correlations were primarily driven by moral
learning/openness subscale and then moral fallibility subscales, i.e., agreeableness and openness to experience were
most strongly and reliably correlated with the moral learning/openness part of the moral humility, followed by moral
fallibility.
Moral humility scale was not correlated with the honesty-humility subscale of the HEXACO inventory.
However, the modesty facet of the honesty-humility subscale, which captures the humility aspect and was of my
interest was positively associated with moral humility (r = 0.16, p <.001). This was primarily driven by a negative
correlation of modesty with moral superiority subscale (r = - 0.36, p <.001), suggesting that humility/modesty
captured in HEXACO is most closely related to the moral superiority subscale. This makes sense as the items
capturing modesty (or humility) assessed how much people think they are better than others (e.g., “I think that I am
entitled to more respect than the average person is”) which is most similar to the content of the moral superiority
subscale.
Correlations between moral humility scale and the Dark Triad (psychopathy, Machiavellianism, narcissism)
revealed that higher moral humility was significantly negatively associated with all three dark traits, psychopathy,
Machiavellianism, and narcissism, with similar pattern of correlations at subscale level too. However, there was an
unexpected small positive correlation between the moral fallibility subscale and psychopathy. This might indicate
that people with higher scores on moral fallibility are also more willing to report that they have moral failings (such
as the traits included in psychopathy). Finally, the moral humility scale was not significantly associated with self-
esteem.
In summary, the correlations indicate that the moral humility scale taps into the constructs of
modesty/humility, openness, and other-orientedness (higher agreeableness and lower levels of dark traits) as was
expected. Moral humility did not tap into self-esteem, indicating humility is not simply self-deprecation or a lack of
self-regard (Tangney, 2000).
28
Table 7
Correlations between the Moral Humility scale and subscales with HEXACO, Big 5, Dark Triad, and Self Esteem
scales
Modesty
Honesty-Humility
Emotionality
Extraversion
Agreeableness
Conscientiousness
Openness to Experience
Extraversion
Agreeableness
Neuroticism
Conscientiousness
Openness to Experience
Machiavellianism
Psychopathy
Narcissism
Overall
HEXACO (Sample 2)
Moral Humility
0.16***
0.01
0.08
-0.04
0.15***
-0.06
0.27***
Moral Learning/
Openness
0.03
0.01
0.05
0.11*
0.24***
0.11*
0.40***
BFI (Sample 5)
-0.07*
0.13***
0.02
-0.03
0.05
0.34***
0.06
-0.05
0.17***
0.42***
Dark Triad (Sample 5)
-0.11***
-0.12***
-0.13***
-0.15***
-0.25***
-0.06
Self-Esteem (Sample 5)
Subscales
Moral
Fallibility
0.01
-0.09*
0.09
-0.11*
0.08
-0.17***
0.19***
-0.03
<.01
0.05
0.10**
0.07*
0.06
0.09**
0.04
Moral Superiority
-0.36***
-0.13**
-0.03
0.08
-0.05
0.03
-0.07
0.17***
0.04
0.05
0.12***
0.11**
0.17***
0.12***
0.28***
Self Esteem
Note: *p<0.05; **p<0.01, ***p<0.001
-0.04
0.03
-0.04
0.06
Political Identification and Political Extremity
Meta-analytic correlations of moral humility with partisan identity (higher scores indicated stronger
Republican identity), ideological identity (higher scores indicated higher conservatism), partisan extremity, and
ideological extremity (extremity is distance from the middle of the scale) are in Table 8. The mean correlations
across samples show that moral humility is significantly and negatively associated with conservatism (Mr= - 0.25),
Republican identity (Mr= - 0.18), partisan extremity (Mr = -0.12), and ideological extremity (Mr= -0.19). The
subscales were largely consistent with the results for entire scale. In summary, the correlations indicate that the
moral humility scale taps into lower extremist inclinations, as was expected. There was also a positive relationship
between moral humility and liberal ideological leanings (alternatively negative relationship between moral humility
29
and conservatism) which although not predicted a priori, might be moral humility tapping into openness given that
liberal ideology has previously been linked with openness to experience (Sibley et al., 2012)
Table 8
Meta-analytic correlations between the Moral Humility scale and subscales with Ideological Extremity, Partisan
Extremity, Ideological Identity, Partisan Identity, Moral Relativism, Intellectual Humility, and Moral
Grandstanding
Overall
Moral
Humility
-0.19**
-0.12**
-0.25**
-0.18**
0.61**
0.56**
-0.27**
k
6
6
6
6
4
4
2
N
5670
5647
5559
5646
3661
3670
1392
Moral
Learning/
Openness
-0.11**
-0.06**
-0.24**
-0.18**
0.57**
0.65**
-0.04**
Subscales
Moral
Fallibility
Moral
Superiority
-0.18**
-0.10**
-0.23**
-0.17**
0.57**
0.47**
-0.09**
0.16**
0.11**
0.08
-0.07
-0.27**
-0.17**
0.47**
Ideological Extremity
Partisan Extremity
Ideological Identity (higher
conservatism)
Partisan Identity (stronger
Republican)
Moral Relativism
Intellectual Humility
Moral Grandstanding
Note: *p<0.05; **p<0.01, ***p<0.001
Religiosity and Religious Exclusivism/Inflexibility
Correlations of moral humility with religiosity and religious exclusivism from Sample 2 are presented in
Table 9. The correlations show that moral humility is significantly and negatively associated with religiosity and
religious exclusivism. The subscales were consistent with the results for entire scale. The correlations with
religiosity and religious exclusivism indicate that the moral humility scale taps into lower rigidity, superiority, and
absolutism.
Moral Grandstanding
Meta-analytic correlations of moral humility with moral grandstanding are in Table 8. Moral humility was
significantly and negatively associated with moral grandstanding (Mr = -0.27) with the strongest associations with
moral superiority subscale. In summary, the correlations with moral grandstanding indicate that the moral humility
scale taps into lower levels of status-seeking, sanctimoniousness, extremity, and superiority.
Moral Relativism
Meta-analytic correlations of moral humility with moral relativism are in Table 8. Moral humility was
significantly and positively associated with moral relativism (Mr = 0.61). The subscales were largely consistent with
the results for entire scale, with moral openness/learning and moral fallibility subscales showing the strongest
30
correlations. In summary, the correlations with moral relativism indicate that the moral humility scale taps into
lower absolutism. It also suggests that moral relativism and moral humility though conceptually distinct are
empirically the most overlapping constructs.
Moral Conviction
Correlations of moral humility with moral conviction from Samples 5, 6, and 7 are in Table 9 (not pooled
because they were measured differently). The correlations show that moral humility is not significantly associated
with moral conviction. However, there were small significant positive correlations with the moral learning/openness
subscale and the moral superiority scale. In summary, the correlations with moral humility indicate that moral
humility is not simply a lack of moral convictions.
Table 9
Correlations between the Moral Humility scale and subscales with Religiosity, Religious Exclusivism Moral
Conviction, and Need for Cognition
Overall
Moral Humility
Moral Learning/
Openness
Religiosity
Subscales
Moral
Fallibility
Sample 2
-0.23***
-0.10*
-0.21***
Sample 2
-0.47***
Sample 5
Sample 6
Sample 7
-0.03
-0.01
0.06
Sample 5
Note: *p<0.05; **p<0.01, ***p<0.001
0.16***
Religious Exclusivism
-0.30***
Moral Conviction
0.10**
0.09**
0.15**
Need for Cognition
0.29***
-0.36***
-0.04
0.01
0.05
0.11***
Moral Superiority
0.23***
0.46***
0.12**
0.13**
0.04
-0.05
Intellectual Humility and Need for Cognition
Meta-analytic correlations of moral humility with intellectual humility and need for cognition are in Table
8. Moral humility was significantly positively associated with intellectual humility (Mr = 0.56) and need for
cognition (r = 0.16). The subscales were largely consistent with the results for entire scale, with the moral
openness/learning subscale showing the strongest correlations across samples. In summary, the correlations with
intellectual humility and need for cognition indicate that the moral humility scale taps into more openness,
flexibility, acceptance of fallibility, and willingness to learn.
31
Summary of Nomological Associations
The correlations suggest that moral humility was associated with constructs in sensible and largely
expected ways. Taken together, the correlations indicate that a person high in moral humility is inclined towards less
extremism and absolutism (e.g., less political extremity, religious exclusiveness, moral grandstanding, moral
absolutism, religiosity), shows openness and flexibility (e.g., openness to experience, intellectual humility, need for
cognition), shows humility (e.g., modesty, intellectual humility), and positive orientation towards others (e.g., more
agreeableness, lower psychopathy, lower narcissism, etc.). They also suggest that moral humility is not simply a lack
of self-esteem or moral conviction.
The positive correlations with modesty and intellectual humility provided evidence for moral humility’s
convergent validity. The lack of correlations with self-esteem and small correlations with moral conviction provide
evidence of moral humility’s discriminant validity. Together, the correlations provide important empirical evidence
that the moral humility scale is working as it should and provide insight into the nature of the moral humility
construct.
The strength of the various nomological associations varied, with moral humility having the strongest
correlations with moral relativism and intellectual humility (r ~ 0.5-0.6). This is likely due to both having the
greatest content and/or construct overlap with moral humility. This suggests that these constructs are very close to
moral humility in the nomological network, and it is important for future studies to establish the incremental validity
of these constructs when explaining relevant outcomes. This could help clarify whether these constructs are distinct
enough to be meaningful. I examine this further in the following sections.
The correlations with other constructs, such as modesty, agreeableness, openness to experience, dark traits,
political extremity, religiosity were small to medium (~ 0.1-0.3). These smaller correlations, compared to moral
relativism and intellectual humility, is likely because these other constructs share some conceptual/construct
similarities with moral humility, but at a more distal level. It also suggests that moral humility is not reducible to
these personality and personality-type constructs.
Interestingly, there were also small, unexpected associations of the moral fallibility subscale with
psychopathy, r = .09. In a similar vein, although not described in main text, small negative correlations of moral
fallibility with sincerity and fairness facets of honesty-humility subscale were also observed, reflected in a negative
correlation of moral fallibility with honesty-humility (r = -.09) in Table 7. What might these small correlations
32
convey? All three of these constructs (psychopathy, sincerity, and fairness) captured inclination towards things that
might be considered morally questionable. For example, psychopathy had participants indicate how much they
agreed with, “I tend to be insensitive” sincerity had “I wouldn't use flattery to get a raise or promotion at work, even
if I thought it would succeed.” (reverse coded), and fairness had “I’d be tempted to use counterfeit money, if I were
sure I could get away with it.”. A small positive correlation of moral fallibility with these suggest that moral
fallibility might also be capturing some people who show inclination towards morally questionable behaviors and
also acknowledge such inclinations as moral limitations. These correlations also suggest indirectly that moral
humility scale is not capturing social desirability.
Predicting Polarization Outcomes
Having constructed a scale, confirmed its factor structure, and tested if it was associated with constructs in
sensible and expected ways, the next aim was to investigate the central question in the project. That is, examining
whether moral humility could counteract the dark features of morality as they manifest in the context of political
polarization in USA. I examined if moral humility was associated with lower levels of polarization outcomes and if
moral humility was associated with lower levels of polarization over and above the effects of other close constructs.
Thus, across Samples 3-7, I tested moral humility’s predictive or criteria validity and incremental validity. The
negative associations between moral humility and political extremity observed thus far foreshadow that moral
humility might have a counteractive effect on polarization. In the next steps, I investigated this link across additional
indicators of polarization.
In Sample 3, the aim was to test if moral humility predicted lower polarization (e.g., animus towards
political outgroup, perceptions of threat from political outgroup, anger and negative affect towards political
outgroup, social distance from political outgroup) as well as anti-democratic outcomes, especially when
misperceptions of outpartisans are corrected. Previous research has found that people often hold inaccurate
perceptions of outpartisans. These misperceptions can be about the characteristics (e.g., demographic) of
outpartisans (Ahler & Sood, 2018), about how hostile, prejudiced, or negative outpartisans are towards one’s
ingroup (Moore-Berg & Hameiri, 2020; Ruggeri et al., 2021), or about outpartisans’ policy positions, values, or
support for ethically questionable actions (Enders & Armaly, 2019; Pasek et al., 2022; Mernyk et al., 2022; Moore-
Berg & Hameiri, 2024). For example, people hold inaccurate perceptions about the extent to which political
outgroup would engage in violence or anti-democratic actions (Pasek et al., 2022; Mernyk et al., 2022). Such
33
misperceptions have been suggested to drive hostility towards political outgroups (Moore-Berg & Hameiri, 2024).
This is evidenced by work suggesting that correcting such misperceptions reduces negativity towards political
outgroups (Moore-Berg & Hameiri, 2024; Voelkel et al., 2023).
Adapting a misperception correction task used in previous work (Lees & Cikara, 2020; Voelkel et al.,
2023), I investigated if higher moral humility would predict lower polarization and anti-democratic attitudes, and if
this would be greater when people were given a misperception correction treatment (compared to control). The
misperceptions corrected were about hostility of the outgroup towards the ingroup. Apart from polarization
outcomes, I also tested anti-democratic attitudes as outcomes because polarization has been linked with a decline in
support for democratic practices (Finkel et al., 2020). Notably, utilizing a misperception correction task also allowed
to test how moral humility is linked with people’s misperceptions about political outgroups. In sum, this study tested
the relationship of moral humility with polarization, anti-democratic attitudes, and misperceptions. Sample 4 which
was a national YouGov sample had pre-included measures of these three outcomes which allowed conceptual
replication of these three relationships using slightly different measures and design (e.g., misperceptions were about
outgroup’s democratic commitments).
In Sample 4 and 5, the primary aim was to test if higher moral humility predicted more openness towards
opposing political standpoints. Previous work has found that all political sides engage in avoidance of information or
opinions from the other side (Frimer et al., 2017), and that lack of cross-cutting information amplifies polarization
(Hobolt et al., 2023; Levy, 2021). Adapting a selective exposure task used in previous work (Frimer et al., 2017), I
investigated if higher moral humility would predict people being more interested in learning about opposing political
viewpoints from people who disagreed with them. The political issues included in the studies included multiple
topics frequently moralized in politics (e.g., voting for Trump/Biden, abortion, guns, transgender, healthcare, etc). A
secondary aim in Sample 4 was to conceptually replicate the relationship of moral humility with polarization, anti-
democratic attitudes, and misperceptions as aforementioned.
In Sample 6, the aim was to test if higher moral humility predicted more support for political compromise.
Compromise serves as a means to recognize, respect, and accommodate pluralistic values in a democracy. Previous
work has found that polarized and moralized individuals reject political compromise (Clifford, 2019; Delton et al.,
2020; Kodapanakkal et al., 2022; Ryan, 2017; Finkel et al., 2020). Thus, adapting a political compromise task used
34
in previous work (Kodapanakkal et al., 2022), I investigated if higher moral humility facilitates more favorable
attitudes towards political compromise.
In Sample 7, the aim was to test if higher moral humility predicted lesser sharing of partisan-consistent
news online (or myside sharing) on moralized topics like abortion, gun, immigration, race, and gender. Previous
work has found that moralized and politically extreme individuals are more likely to share partisan-consistent news
online (Marie et al., 2023). Research also shows that social media aggravates the negative facets of morality such as
moral outrage (Van Bavel et al., 2024). Thus, adapting a myside sharing task used in previous work (Marie et al.,
2023),I investigated if higher moral humility tempers inclination to share partisan and moralized news on social
media.
Sample Size Justification
In Sample 3, sample size was based on estimates of misperception treatment effect and moral humility’s
effect on outcomes from previous work (Voelkel et al., 2023; Vallabha et al., 2024). A sample of 1500 was chosen
based on power analysis using InteractionPoweR (Finsaas & Barangeras, 2018) which indicated between ~80% to
~90% power to detect a moral humility effect in the range r [0.08, 0.2] after accounting for the treatment and
interaction effects of r = 0.07 (Voelkel et al., 2023).
In Sample 4, sample size justification was based on two factors. First, the YouGov national survey where I
proposed to field my study allowed a sample of 1000. Second, in Sample 3 the correlation between moral humility
and polarization outcomes ranged from 0.16 to 0.27. Further, moral humility was correlated with misperceptions at r
= 0.1 and with support for anti-democratic candidates at r = 0.25. Based on these prior results, G-power (Faul et al.,
2007, 2009) computed that the smallest effect size observed previously of ~ 0.1 can be detected with 90% power
(alpha = .05) with a sample of ~1000. This gave confidence in the ability of this study to detect the effects of interest
and was preregistered.
Sample 5 was a replication of Sample 4, thus the same sample size as Sample 4 (N = 1000) was chosen for
the study. Sensitivity analysis in G-power function indicated that the study had 80% power to detect an effect size of
r = 0.09. A post-hoc analysis of achieved power using simr package in R (Green & MacLeod, 2015) showed that
there was 100% power to detect the effects between moral humility and selective exposure.
Sample 6 had similar sample constraints as Sample 4 as it was also fielded in a YouGov national sample
which only allowed for a sample of 1000. Sensitivity analysis function indicated that the study had 80% power to
35
detect an effect size of r = 0.09. A post-hoc analysis of achieved power using, G-power (Faul et al., 2007, 2009)
showed that we had ~99 to 100 % power to detect the effects between moral humility and political compromise
across the outcomes.
In Sample 7, power analysis using Summary Statistics Based R for multilevel analysis (Murayama et al.,
2022) was conducted based on estimates from previous work (Marie et al., 2023) using the same design. Power
analysis suggested that a sample of ~800 had 90% power to detect a small cross-level interaction effect size of b =
0.07 (t = 2.07).
Measures
Moral Humility
In Samples 3, 5, and 7, moral humility was measured using the full 30-item measure of moral humility
(Sample 3 α = 0.94, Sample 5 α = 0.89, Sample 7 α = 0.94). In Samples 4 and 6, due to the limited number of items
that could be included or proposed in the respective YouGov national surveys, shortened measures were proposed.
The items for the shortened measure were sampled almost equally from all subscales (Sample 4: 8-item measure, α
= 0.75; Sample 5: 6-item measure, α = 0.67). These items were picked using genetic algorithm for creating
shortened scales (Scrucca & Sahdra, 2016). The algorithm iteratively evaluates different combinations of items on
various parameters (like reliability, content coverage) and eventually selects items that maximize these parameters.
See Table 3 for the selected items marked using symbols.
Subscales were also used to explore relationships with outcomes at the subscale level, i.e., with Moral
Learning/Openness (Sample 3 α = 0.93, Sample 4 α = 0.75, Sample 5 α = 0.91, Sample 6 α = 0.71, Sample 7 α =
0.93), Moral Fallibility (Sample 3 α = 0.93, Sample 4 α = 0.75, Sample 5 α = 0.86, Sample 6 α = 0.68, Sample 7 α =
0.93), Moral Superiority (Sample 3 α = 0.88, Sample 4 α = 0.66, Sample 5 α = 0.84, Sample 6 α = 0.6, Sample 7 α =
0.89).
Polarization or Social Cohesion Measures
Several polarization outcomes were measured, drawing from previous work capturing intergroup
perceptions and behaviors in the political context (Iyengar et al., 2019; Kuhne & Kamin, n.d; Vallabha et al., 2024).
These included Social Cohesion Impact Measures (SCIM), a group of measures collated by academics and
practitioners to measure depolarization (Kuhne & Kamin, n.d). Example of measures and items are in Table 10.
36
Table 10
Example of polarization measures used in Sample 3 (including SCIM measures).
Construct
Example Items
Affective Polarization/
Partisan Affect Gap
• How would you rate Democrats?
• How would you rate Republicans?
(0= Very cold or unfavorable feeling, 100=Very warm or favorable feeling)
Social Distance
• How comfortable are you having friends who are [outgroup members]? 0= Not at
all, 10 = Extremely
Humanization
• How often do you think [outgroup members] experience the following emotions?
0 = Never to 10 = Very frequently
(i) Hope.
(ii) Admiration.
Morality
• Would you say that [outgroups members] are generally good people? 0= Not at
all, 10 =Absolutely
Intergroup Empathy
•
•
I find it difficult to see things from [outgroup members] point of view. 0 =
Strongly disagree, 10 = Strongly agree
It is important to understand [outgroup members] by imagining how things look
from their perspective. 0 = Strongly disagree, 10 = Strongly agree
Respect/
Understanding
• Even if I don’t agree with them, I understand people have good reasons for
•
voting for [outgroup] candidates. 0 = Strongly disagree, 10 = Strongly agree
I respect [outgroup members'] opinions even when I do not agree.
0 = Strongly disagree, 10 = Strongly agree
Pluralist Norms
• How important to you is it that [ingroup] elected officials make compromises
with [outgroup] elected officials to solve important problems? 0= Not at all, 10 =
Extremely
• How likely would you be to vote for a [ingroup] candidate who said they would
ban [extreme outgroup] group rallies on the state capitol grounds? 0= Not at all,
10 = Extremely
Perceived Threat
• Would you say [outgroup members] are a serious threat to the United States? 0=
Not at all, 10 =Absolutely
Anger
Identity
• How angry do you get just thinking about [outgroup members]? 0= Not at all, 10
= Extremely
• How much do you agree with the statements “if I met someone who is a [member
of ingroup], I’d feel connected to that person”?
0 = Strongly disagree, 10 = Strongly agree
Note: Inter-item correlation for Humanization was 0.8 (p <.001), Empathy was 0.5 (p <.001),
Respect/Understanding 0.7 (p <.001), and Pluralistic Norms was 0.2 (p <.001).
To give a brief overview, Sample 3 included measures capturing negative orientation towards outgroup like
partisan affect gap (popularly used as a measure of affective polarization; Iyengar et al., 2019), negative affect
37
towards outgroup, anger towards outgroup, social distance from outgroup, perceptions of outgroup threat, and close
identification with ingroup. Measures capturing positive orientation towards outgroup included perceptions of
outgroup morality, empathy towards political outgroup, humanization of political outgroup, respect/understanding
towards political outgroup, and support for pluralistic norms. Sample 4 had the partisan affect gap or affective
polarization measure only, captured using thermometer ratings towards Republicans/Democrats.
Anti-Democratic Attitudes
Anti-democratic attitudes were measured in Samples 3 and 4. Sample 3 used 6 items from previous work
on polarization and democratic outcomes (Voelkel et al., 2023), which were averaged together to create a measure
of anti-democratic attitude (α = 0.91). These items captured support towards political inparty candidates who engage
in anti-democratic practices such as gerrymandering, disputing election legitimacy, and suppressing media. An
example item was, “How likely would you be to vote for the [Inparty] candidate if you learned that they said that
[Inparty] should not accept election results if they do not win?”.
In Sample 4, democratic attitudes were similarly measured by taking an average of 4 items that were
included in the national survey which captured support for anti-democratic norms (α = 0.74). Participants rated how
much they agreed or disagreed with various anti-democratic actions like censorship, subverting court decision or
Congress/Executive, and interfering with polling. An example item was, “Do you agree or disagree with the
following: The government should be able to censor media sources that spend more time attacking [inparty] than
[outparty].” See SOM for full details on items from both samples.
Misperceptions
Misperceptions were measured and corrected in Sample 3, and only measured in Sample 4. In Sample 3, the
misperception correction task was taken from previous work (Lees & Cikara, 2020; Voelkel et al., 2023). Therein,
participants were randomly assigned to either the misperception correction treatment (N = 732) or control (N = 739)
condition. In the misperception correction treatment condition, participants were randomly presented with one of
five scenarios wherein their ingroup was undertaking an action that would possibly disadvantage the outgroup. An
example scenario was “A [Inparty] controlled state legislature is considering a law that would require sitting
governors to disclose their tax returns and all possible financial conflicts of interests. The law would go into effect
immediately, and the current sitting governor is a [Outparty Person].” See SOM for all five scenarios. Participants
were asked to indicate the extent to which they thought a partisan outgroup member would dislike the action, oppose
38
the action, and find the action politically unacceptable. These, when subtracted from outgroup members’ actual
dislike, opposition, and unacceptability, indexed participant’s level of misperceptions7.
The misperception correction took place next wherein the participants in the treatment condition were
informed of the real responses from partisan outgroup members on how much they actually disliked/opposed/found
unacceptable the actions. The real responses presented were based on a previous nationally representative survey
(Lees & Cikara, 2020). The control condition did not take part in the misperception task and directly moved to
answering the outcome measures.
In Sample 4, misperceptions about support for anti-democratic actions were only measured. Misperceptions
about the outparty were calculated in two steps. First the percentage of Democratic and Republican participants
agreeing to four anti-democratic actions described above (censorship, subverting court decision or
Congress/Executive, and interfering with polling) was estimated. For example, “Do you agree or disagree with the
following: The government should be able to censor media sources that spend more time attacking [inparty] than
[outparty].” See SOM for full details on all items. This indexed actual support towards anti-democratic actions by
Democrats and Republican respectively. In the second step, these actual agreement values for the outgroup were
subtracted from each participant’s response to four questions assessing participants’ perceptions of outgroup
members’ approval of anti-democratic actions noted above. For example, for media censorship, they were asked,
“What percent of [outparty] voters do you think agree with the following: The government should be able to censor
media sources that spend more time attacking [outparty] than [inparty].” The values obtained from subtraction of
actual values estimated in first step from perceived values in second step for each of the four democratic action was
averaged to create measure of misperceptions (α = 0.85). These gave estimates of the magnitude of each
participant’s misperception of outgroups.
Selective Exposure
Selective exposure, the primary outcome in Sample 4 and 5, was measured by assessing participants’
interest in opposing political standpoints. For this, I adapted a selective exposure task from previous work (Study 2-
3; Frimer et al., 2017). In both Samples 4 and 5, for one (Sample 4) or three (Sample 5) political issues answered by
7 Misperceptions could only be calculated for the treatment group who actually reported their (mis)perceptions of
how much they thought outgroup members would oppose, dislike, and find unacceptable certain actions. Following
the materials from Voelkel and colleagues (2023), the control group did not complete these initial judgments.
39
participants previously in the study, participants were asked how interested they would be in hearing from someone
with the opposing viewpoint on the issue.
The issues were chosen slightly differently in both samples. In Sample 4, there were ten issues that a
participant could answer plus one issue that everyone had to answer. Of the ten possible issues, each participant was
randomly assigned to answer only five issues. From these five issues and one mandatory issue8 answered by
everyone, only one issue was then randomly selected for selective exposure task. In Sample 5, of the eleven total
issues in Sample 1, each participant answered three randomly selected issues and then completed selective exposure
task for the three issues. The issues included environment, healthcare, transgender, abortion, defunding police, guns,
trade, taxes, unions, marijuana, presidential vote. These issues were already part of the national survey wherein I
fielded moral humility and selective exposure measures, so I leveraged participants’ response to these issues for the
selective exposure task. See SOM for full wording of how issues were assessed.
For example, if a participant answered the transgender issue, they were first asked,
“Some believe that transgender athletes should be allowed to compete on teams that match the gender they identify
with. Others believe that transgender athletes should be required to compete on teams that match the sex they were
assigned at birth. Still others fall somewhere between these two positions. Where do you stand on this issue?” They
then indicated their response on a 7-point scale (1= Allow transgender athletes to compete on teams matching their
gender identity, 4 = Middle of the road; see the pros and cons of both sides, 7 =Require transgender athletes to
compete on teams matching their sex assigned at birth).
If someone answered 1, 2, or 3 on the scale, they then were asked later, “You indicated that you leaned
towards allowing transgender athletes to compete on teams matching their gender identity. How interested are you in
hearing from someone who supports requiring transgender athletes to compete on teams matching their sex assigned
at birth?”. If they instead answered 5, 6, or 7, they read, “You indicated that you leaned towards requiring
transgender athletes to compete on teams matching their sex assigned at birth. How interested are you in hearing
from someone who supports allowing transgender athletes to compete on teams matching their gender identity?”. If
they answered 4, another issue was randomly selected for selective exposure task in Sample 4 wherein participant’s
8 The mandatory question was about their vote choice in previous election, which was asked of everyone at the
beginning of the study
40
response was not a 4. In Sample 5, if participant answered 4 on any issue, they did not get the selective exposure
task for that issue.
Participants then indicated their response on a -100 to 100 scale [−100 (very uninterested), 0 (neutral), and
100 (very interested)]. Higher values indicated more interest in learning from someone who disagreed with them
politically, or more interest in cross-cutting exposure.
Political Compromise
The measure for political compromise, the main outcome in Sample 6, was adapted from Kodapanakkal et
al. (2022). Participants were first asked to choose one of thirteen issues that was most important for them. The issues
included in this list were selected on the basis of results of a national survey run by YouGov in 2023 on the issues
Americans found most important (Frankovic et al., 2023), such as, abortion, healthcare, immigration, guns. (see
SOM for full list). After selecting their most important issue, participants completed a political compromise task
from Kodapanakkal et al. (2022) where they were asked about two candidates with different approaches to the issue
they had selected as most important,
“We would like your opinion on two candidates with different approaches to [Most Important Issue]. The
candidates might be competing for their party’s nomination to run for Congress. Both candidates agree with your
position on [Most Important Issue], but they differ on how they plan to negotiate with their political opponents.
Candidate A is uncompromising and will vote against any proposal that does not support your position.
Candidate B will dislike proposals that do not support your position, but will be willing to negotiate and
make concessions in this area if it leads to a gain in other areas that are important to you.”
Participants indicated their likelihood of supporting each candidate separately, “How likely are you to
support Candidate [A/B]?” using a 7-point Likert scale (1 = not at all, 7 = very likely). Support for Candidate A (the
uncompromising candidate), support for Candidate B (the compromising candidate), and the relative support of
compromising candidate over uncompromising candidate (Candidate B - Candidate B) formed the three main
outcome measures for the study.
Myside Sharing
The materials and measures for myside sharing or partisan sharing, the main outcome in Sample 7, was
taken from Marie et al. (2023; Experiment 1 &10). Participants were randomly assigned to one of two news sets
which contained 12 real news items each. Thus, in total there were 24 news items but to prevent participant fatigue,
participants were assigned to only one set of news stories.
41
The news items were on divisive issues: gun control, abortion, gender, race, and immigration. Each set
contained 12 news, two news items per these five issues, one which was congruent for liberals (e.g. pro-abortion)
and one congruent for conservatives (e.g. anti-abortion). This resulted in 10 news items; additionally, there were 2
neutral news items in each set. All news items were real and had been picked from mainstream news media
websites. Each news item had a headline, a short introductory snippet, and a picture. All news items are in SOM.
After reading the 12 news items, participants were asked about their inclination to share those news stories
on social media, “How likely would you be to share this news article on social media?” (1=Extremely Unlikely, 2=
Unlikely, 3=Somewhat Unlikely, 4 = Somewhat Likely, 5 = Likely, 6 = Extremely Likely).
Participants also indicated their position on the five issues on which the news items were based. For race,
they were asked “What is your position on the issue of racial equality? (0= I don't care at all, 100 = Extremely in
favor), for gender “What is your position on the issue of gender equality?” (0= I don't care at all, 100 = Extremely
in favor), for abortion “What is your position on the issue of abortion?” (0= Extremely pro-life, 100 = Extremely
pro-choice), for guns “What is your position on the issue of guns?” (0= Extremely pro-gun rights, 100 = Extremely
pro-gun control), and for immigration “What is your position on the issue of immigration?” (0=Extremely in favor,
100= Extremely opposed). Participants’ position on the issues was used to determine which news items (liberal or
conservative for each issue) was politically congruent and incongruent to participants.
The study was opened to only those participants on Prolific who indicated on their prescreening survey that
they used at least one of the following social media: Facebook, Twitter, Instagram, Tik Tok, and Reddit. This choice
was made based on past work on online news sharing that used similar eligibility criterion (Mosleh et al., 2020).
This was to ensure that our study was realistic to participants doing the study as the study pertains to online news
sharing.
Controls
Partisan identity, ideological identity, partisan extremity, ideological extremity, and demographics (age,
race, gender) were included as the standard set of control variables in all analyses conducted in Samples 3-7.
Gender (male = 0.5, female = -0.5) and ethnicity (0.5 = white, -0.5 = non-white) were contrast coded.
Samples that were not collected as part of national surveys (wherein I was limited in the number of items I
could include, i.e., Sample 4 and 6) included moral relativism and intellectual humility as control variables as well
to establish moral humility’s incremental validity over these closely related psychological constructs. These two
42
were chosen as these two constructs are most closely conceptually and/or empirically related to moral humility.
They were thus included in Samples 3, 5, and 7. Sample 6 included moral conviction as well as it had been
preregistered as control for exploratory purposes.
Analysis and Results
Sample 3 & 4: Misperceptions, Polarization, Anti-Democratic Attitudes
Sample 3 Results. In Sample 3, the social cohesion outcomes and democratic outcomes were regressed on
moral humility, the condition variable (contrast coded: Control= -5, Misperception Correction Treatment = .5), and
the interaction term between them for the first set of models. In the second set, the same models were re-estimated
including all the covariates. I expected main effects for moral humility and the condition variable, as well as a
significant interaction term between them. Moral humility was mean-centered before analyses to improve
interpretability for the interaction analyses.
Unlike previous work (Lees & Cikara, 2020; Voelkel et al., 2023), the misperception correction treatment
did not work as expected and didn’t have a significant effect on any of the outcome measures (all p’s > .05 p range
[0.16, 0.96]). There also wasn’t a significant interaction of the treatment with moral humility (all p’s > .05 p range
[0.16, 0.86]). This suggests that misperception corrections did not reduce animosity towards outgroup, and this non-
significant effect was similar across levels of moral humility. There were however main effects of moral humility on
the polarization and democratic outcomes across conditions. Additionally, moral humility was also associated with
lower magnitude of misperceptions for the experimental group (the control group did not complete these measures).
These associations are shown in Figure 2 (upper part).
Higher moral humility was associated with lower partisan affect gap (or affective polarization), lower
desire to socially distance from the political outgroup, lower negative affect and anger towards the political
outgroup, lower perceptions of threat from the political outgroup, lower misperceptions of the outgroup members,
and lower support towards anti-democratic candidates. Higher moral humility was also associated with more
empathy, humanization, and respect/understanding of political outgroup, more support for pluralistic norms, and
higher perceptions of outgroup morality. These relationships were robust to the inclusion of control variables like
intellectual humility and moral relativism (except for humanization).
43
Figure 2
Sample 3 and 4 Relationship between Moral Humility and Political Outcomes
Note: The blue lines with circles indicate models where moral humility was the only predictor; the orange lines with
triangles indicate the models where all controls were also included. Outcomes were recoded to range from 0-1. The
controls in Sample 3 included the standard controls plus intellectual humility and moral relativism, controls in
Sample 4 only included the standard controls (see Control section above).
Sample 4 Results: Conceptual Replication of Sample 3. Three outcomes from Sample 3, misperception,
partisan affect gap (or affective polarization), and anti-democratic attitudes were conceptually replicated in a
national sample in Sample 4. Linear regression was used to estimate two models each, with and without controls, for
the three outcomes. The results are in the lower part of Figure 2. Moral humility had negative relationships with all
three outcomes (with and without controls).
Sample 3 & 4 Subscale Result. I also explored if the results observed in Sample 3 and 4 held or varied
across the moral humility subscales. To this end, the models in Samples 3 and 4 were re-estimated using the moral
humility subscales. The results are in Figures 3 and 4. Results observed for the whole scale largely held for all the
subscales too. Some outcomes were less robust to inclusion of controls when using a particular subscale (such as
pluralism when predicted by moral learning/openness, or misperceptions when predicted by moral superiority).
However, the general direction of results across the outcomes for each subscale was consistent with that of the
44
whole scale. Moral superiority subscale is consistent with whole scale when it predicts outcomes in the opposite
direction than the full scale i.e., more negative outcomes and less positive outcomes. The other two subscales should
show results in same direction as the whole scale to be consistent. There was one exception. In Sample 5, higher
moral fallibility predicted more support for anti-democratic norms.
Figure 3
Sample 3 Relationship between Moral Humility Subscales and Political Outcomes.
Note: The green lines with triangles indicate where moral learning/openness was predictor; the orange lines with
circles indicate the model where moral fallibility was the predictor; the blue lines with diamonds indicates the model
where moral superiority was the predictor. Outcomes were recoded to range from 0-1. The controls in Sample 3
included the standard controls plus intellectual humility and moral relativism.
45
Figure 4
Sample 4 Relationship between Moral Humility Subscales and Political Outcomes
Note: The green lines with triangles indicate where moral learning/openness was predictor; the orange lines with
circles indicate the model where moral fallibility was the predictor; the blue lines with diamonds indicate the model
where moral superiority was the predictor. Outcomes were recoded to range from 0-1. The controls in Sample 4 only
included the standard controls (see Control section above).
Sample 4 and 5: Selective Exposure
In both Samples 4 and 5, selective exposure was regressed on moral humility, both without and with
controls. The analytic strategy was a little different in both samples given that Sample 5 was repeated measures
design, unlike Sample 4. That is, in Sample 5, each participant completed selective exposure for three issues instead
of just one.
In Sample 4, linear regression was used wherein selective exposure for one issue was regressed on the
moral humility measure. Both the models, with and without controls, also included a dummy coded control for the
issue that the participant had been randomly assigned to complete as part of the selective exposure task (described in
methods). In Sample 5, as each participant completed selective exposure for three issues, multilevel regression
analyses was used instead. Selective exposure was regressed on moral humility and issues were nested in persons.
Random intercepts for both the issues and the persons were included.
46
Both samples found similar results for selective exposure (Figure 5). Higher moral humility was associated
with more interest in cross-cutting exposure. This effect emerged over and above the control variables. Like before,
models with each moral humility subscale were also estimated. The results are in Figure 6. Results observed for the
whole scale largely held for all the subscales too.
Figure 5
Sample 4 and 5 Relationships between Moral Humility and Selective Exposure
Note: The blue line with triangle indicates the model where moral humility was the only predictor; the orange lines
with circles indicate the model where all covariates were also included. Outcomes were recoded to range from 0-1.
Sample 4 was a national survey and hence has fewer controls. The coefficients for dummy codes for the issues in
Sample 4 are hidden in the figure to keep the figure neat but were estimated in the model.
47
Figure 6
Sample 4 & 5 Relationship between Moral Humility Subscales and Selective Exposure
Note: The blue line with triangle indicates the model where moral humility was the only predictor; the orange lines
with circles indicate the model where all covariates were also included. Outcomes were recoded to range from 0-1.
Sample 4 was a national survey, and hence fewer controls were included. The coefficients for dummy codes for the
issues in Sample 4 are hidden in the figure to keep the figure neat but were estimated in the model.
Sample 6: Political Compromise
The three political compromise outcomes (support for Candidate A, support for Candidate B, the relative
support towards Candidate B over A) were each regressed on moral humility without and then with control
variables.
Moral humility was associated with more support for Candidate B (the more compromising candidate),
lesser support for Candidate A (the lesser compromising candidate), and more support of Candidate B over A, both
with and without controls (Figure 7). The results for the subscales are in Figure 8. They show that the results
observed for the whole scale largely held for all the subscales too. However, moral superiority more reliably
predicted support for uncompromising candidate, whereas moral learning and fallibility more reliably predicted
support for the compromising candidate.
48
Figure 7
Sample 6 Relationship between Moral Humility and Political Compromise
Note: The blue line with triangle indicates the model where moral humility was the only predictor; the orange lines
with circles indicate the model where all covariates were also included. Outcomes were recoded to range from 0-1.
The coefficients for dummy codes for the issues in Sample 6 are hidden in the figure to keep the figure neat but were
estimated in the model.
Sample 7: Myside Sharing
First, the political congruency of the ten news items (two for each of the five issues) that the participants
saw as part of the task was determined for each participant. To this end, participants’ position on the five issues were
reverse coded where needed (i.e., on immigration question) such that lower values on all issue indicated more
conservative position and higher values indicated more liberal position. If the participant had a response <50 on an
issue, then the conservative news item on the issue was considered as politically congruent and the liberal news item
as politically incongruent. If the participant had a response ≥50 on an issue, then liberal news items were considered
congruent and conservative news items incongruent. Two contrast codes indexing news item congruence were
created for analyses wherein incongruent news stories were treated as reference. The two contrast codes were
congruent (congruent = 0.5, other = -0.5) and neutral (neutral = 0.5, other = -0.5).
Multilevel regression models were used for the final analysis with random intercepts for participants and
news item. Moral humility was grand-mean centered. In the first set of models, willingness to share the news item
was first regressed on congruent dummy code, neutral dummy code, and congruent dummy’s interactions with
moral humility. In the second set of models, control variables were included.
49
Figure 8
Sample 6 Relationship between Moral Humility Subscales and Political Compromise
Note: The blue line with triangle indicates the model where moral humility was the only predictor; the orange lines
with circles indicate the model where all covariates were also included. Outcomes were recoded to range from 0-1.
The coefficients for dummy codes for the issues in Sample 6 are hidden in the figure to keep the figure neat but were
estimated in the model.
Results showed that people were indeed more likely to share congruent news items more than incongruent
news items (b = 0.71, SE = 0.03, t = 23.54, p <.001). The same was observed for neutral new items (b = 0.72, SE =
0.17, t = 4.06, p <.001). However, there wasn’t a significant interaction of moral humility with congruent dummy,
contrary to what was predicted (b = -0.03, SE = 0.02, t = -1.33, p = 0.18). Moral humility did not have a main effect
either; it wasn’t associated with levels of news sharing (b = 0.03, SE = 0.04, t = 0.8, p = 0.40). Results for the
subscale were largely consistent with the results of the whole scale. There wasn’t a significant interaction of moral
learning/openness with congruent dummy (b = 0.01, SE = 0.02, t = 0.8, p = 0.41). Moral fallibility (b = -0.04, SE =
0.02, t = -1.80, p = 0.06) and moral superiority (b = 0.04, SE = 0.02, t = 1.80, p = 0.07) had marginal p values and
small effects in the direction of predicted effects (below the threshold of what our study was powered to detect at
90%). The results for the scale and subscale held with inclusion of control variables. Overall, the results suggested
50
that moral humility and its various facets didn’t significantly temper the myside sharing effect. If there is indeed an
effect in the predicted direction in case of moral fallibility and superiority, the effect might be small.
Incremental Validity Robustness Check
Moral humility significantly predicted polarization and related outcomes over and above important controls
like moral relativism and intellectual humility. This provided support for its incremental validity. However, recently
some research has advised against using multiple regression solely to establish evidence of incremental due to the
likelihood of Type I error from measurement unreliability (Westfall & Yarkoni, 2016). The authors instead
recommend methods like structural equation modeling (SEM) which accounts for measurement unreliability. Given
that establishing moral humility’s incremental validity over closely related constructs like moral relativism and
intellectual humility was one of the central aims in this project, I checked for the robustness of the incremental
validity results in Sample 3 and 5 (these had the two controls in analyses) using this recommended SEM approach.
Instead of computing the mean score for moral humility, moral relativism, and intellectual humility and then
entering these mean scores as simultaneous predictors in a multiple regression, I specified the measurement models
for these three constructs. Thus, moral humility (hierarchical model with three factors), moral relativism (one factor
model), and intellectual humility (one factor model) were entered as latent factors as part of the structural regression
model. Accordingly, the measurement error became an explicit part of the full model. I tested the models with the
control variables in Sample 3 using this approach (see SOM). Essentially, I replicated moral humility’s incremental
validity over the control variables, and notably over and above moral relativism and intellectual humility.
Interestingly, most significant effects of intellectual humility that were observed before no longer held when using
structural equation modeling. Further, the effect sizes for moral humility also became larger.
Predictive Validity Effect Size
I computed standardized regression coefficients for all models across studies (see SOM). A summary of the
moral humility effect size from all models with and without control are in Table 11. For comparison, I also present
effect size from the models with controls for intellectual humility — one of the constructs closest to moral humility
which has also been investigated in the polarization context before, and political (partisan and ideological) extremity
— the construct that has been theorized to be central to polarization. Compared to both intellectual humility and
political extremity, moral humility showed a stronger effect. Notably, these moral humility effects were even
51
stronger (median β = 0.46) when SEM was used for analyses which accounts for measurement unreliability (see
incremental validity robustness section above).
The summary statistics (Table 11) show that the average effect of moral humility
(β ~ 0.20) was small according to Cohen’s standards. According to different standards, an effect of 0.2 is medium
(Funder & Ozer, 2019) while an effect of 0.1 is small. According to these guidelines, the effects for moral humility
were on an average, medium-sized, and those of intellectual and political extremity were small-sized. In sum, the
effect size of moral humility being comparable (or larger) to political extremity suggests that moral humility is
perhaps of substantive importance in understanding polarization.
Table 11
Predictive Validity Effect Size Summary for Moral Humility, Intellectual Humility, and Political Extremity
MH
(Without
Control)
MH
(With
Control)
IH
(With
Control)
Mean
SD
Median
Min
Max
0.23
0.09
0.23
0.07
0.42
0.20
0.08
0.19
0.07
0.34
0.09
0.06
0.11
0.02
0.19
Party
Extremity
(With
Control)
0.12
0.09
0.09
0.01
0.28
Ideological Extremity
(With
Control)
0.13
0.07
0.14
0.02
0.23
Summary of Polarization Results
The broader aim of the analyses conducted in this section was to test if in a moralized and conflictual
context, moral humility could attenuate its negative aspects, such as having low opinion of the outgroup, antagonism
and derogation towards outgroup, rigidity in one’s own views, rejection of compromise and contact, and adoption of
morally questionable means.
In line with these expectations, in the context of political polarization in the US, I found moral humility to
negatively predict a range of polarization and other associated outcomes across studies. Moral humility was
associated with more positive opinion towards political outgroup such as higher perceptions of outgroup morality,
lower perceptions of outgroup threat, and lower misperception of the outgroup’s hostility towards their ingroup.
Moral humility was also associated with lesser antagonistic feelings towards political outgroups such lower negative
affect and lower anger, as well as lower gap between negative/positive feelings towards political ingroups and
outgroups. It was further associated with a range of other-oriented outcomes, such as more empathy, more respect,
and more understanding towards political outgroup and their perspectives. Similarly, it was associated with less
52
rigidity and more openness, such as expressing more willingness to learn from disagreeing others and their views,
more willingness to engage in political compromise, and not very strong attachments to the political ingroup
identity. Finally, it was associated with lower support for morally questionable means such as support for anti-
democratic and anti-pluralistic actions. Thus, overall people higher in moral humility demonized the political
outgroup less, had more positive, respectful and open orientation towards the political outgroup and towards
opposing political viewpoints, and were more committed to norms and practices accommodating diverse values,
perspectives and interests.
These studies taken together provided evidence for moral humility’s criteria and predictive validity in a
moralized context. Further, most of these effects were observed over and above the effects of closely related
psychological constructs such as intellectual humility, moral relativism, and political extremity. This thus provided
evidence for moral humility’s incremental validity and suggested that it has distinct value in explaining moralized
conflicts. Further, the effect of moral humility on polarization outcomes, although modest, was comparable (and
even stronger) to the effects of political extremity, an explanatory variable considered very central in polarization
literature. This suggests that moral humility is perhaps of substantive importance.
However, there were a few findings that did not fully align with expectations. First, moral humility did not
predict lower myside sharing on social media, and the effects of moral humility facets which showed effects in
prediction-consistent direction were small and marginal. There could be a few reasons for that. Political
conversations are a small proportion of total conversation on social media and a very small number of social media
users (~9%) produce the most political content (Pew Research Center, 2019, 2021). This suggests that most people
do not engage in political sharing on social media. This is supported by our own data which found the median
sharing to be 2 (indicating “Unlikely” on a scale of 6, with mean of M = 2.5 with SD = 1.5). Thus, maybe moral
humility might not have much variation to explain, or other factors might be more important in explaining political
sharing on social media.
Second, in one of the samples (Sample 4), moral fallibility subscale was positively associated with more
support for anti-democratic actions. This was not expected a priori, but aligns with the results found in previous
section, vis-à-vis moral fallibility’s small positive associations with psychopathy, which had suggested that moral
fallibility might be capturing some people who accept engaging in or supporting morally questionably acts and also
accept their immoral inclinations as moral limitations. In a similar vein, the results for anti-democratic attitudes
53
suggest that people who support anti-democratic means might also be willing to accept that they might be morally
flawed sometimes. However, there is a caveat here — this positive relationship was only found in one sample and
not the other (Sample 3) suggesting it would need to be replicated and investigated further before strong conclusions
can be drawn.
Otherwise, the results for subscales largely held in direction consistent with the overall scale. Some effects
were weaker for a particular subscale or were sometimes less robust to controls, suggesting that different facets of
moral humility might be sometimes playing a stronger role in regard to some outcomes. For example, support for the
less compromising candidate (Sample 6), support for anti-democratic candidate (Sample 3) and lower support for
pluralistic norms (Sample 3) was more reliably and robustly predicted by moral superiority. Or higher willingness to
cross-cutting exposure (Sample 4, 5) was more reliably and robustly predicted by moral fallibility and moral
openness/learning. Thus, the different facets of moral humility play unique role in explaining outcomes while being
largely consistent with results of the whole scale, supporting the view of moral humility construct as one construct
with different and unique yet correlated factors. Taken together the results indicate that moral humility predicts
outcomes in sensible and expected ways across the scale and the subscales, and has unique explanatory power over
other constructs of substantive importance.
54
Chapter III: Intervention Development and Causal Assessment
The studies conducted so far used cross-sectional designs. Further, moral humility was measured like a
trait, i.e., as a stable dispositional aspect of a person. In the next step of my research program on moral humility, and
as part of the new studies in the dissertation involving the committee, I conducted two experiments to take the study
of moral humility in a causal direction, as well as investigate if moral humility can be studied as a state. That is, I
tested if and how moral humility can be changed. The studies attempted to change moral humility using newly
designed interventions that manipulated aspects of moral humility like fallibility, openness, learning, and other-
orientedness in order to move people’s moral humility. Further, the studies also tested if these changes have
downstream effects and are associated with reduced levels of polarization outcomes. Together, these studies thus
attempted to find ways that moral humility can be changed and establish the causal effect of moral humility on
political outcomes. These studies were the first experimental studies that I know of that attempted to do this.
First, I developed five moral humility interventions, details of which are described first below. I then pilot
tested these and subsequently ran two experiments, testing these interventions individually (Experiment 1) and
together (Experiment 2). The Pilot, Experiment 1, and Experiment 2 are described below in separate sections after
describing the five interventions.
Transparency and Openness
Both the main experiments’ study materials, design, hypotheses, and planned analyses were preregistered
and can be found at OSF. All studies were approved by the Michigan State University Institutional Review Board.
Moral Humility Interventions
Five interventions or treatments were developed to increase moral humility (see SOM for the
interventions). These interventions each had a vignette that the participants would read, which talked about some
moral ideas that I believed might move moral humility. The vignettes were presented along with congruent pictures
and interactive questions based on the vignette. These interactive questions were not intended to be the outcome
variables in main experiments, but were a feature of the treatments themselves, included to make the task of reading
these vignettes more engaging. There was also a control vignette with pictures and interactive questions on a topic
(artificial intelligence) unrelated to morality, to be used in the control condition in the main experiments.
The five moral humility vignettes aimed to induce moral humility by manipulating or making salient its
various attributes — moral fallibility, moral openness/learning, and moral superiority. Each vignette invoked one or
55
more of these attributes. It might be argued that a better design would have different vignettes targeting different
moral humility attributes separately than employing a mix of them in the vignettes. This systematic approach might
be useful in future work. At this nascent stage where no moral humility manipulations exist in the literature, the aim
was to adopt more of a proof-of-concept approach to see if moral humility can be moved in principle and what kind
of ideas broadly work well as moral humility manipulations. To that end, a more general approach was adopted to
designing manipulations, using any ideas that might capture and move moral humility more generally. These ideas
were thus a mix of moral humility attributes that I believe might move moral humility.
In a similar vein, a secondary justification for adopting a general approach of using a mix of moral humility
attributes in the intervention vignettes is that previous work testing multiple interventions to reduce polarization
(Voelkel et al., 2023) found that multifactorial interventions (i.e., treatments that employed multiple strategies) were
more effective. Thus, at a proof-of-concept stage where the aim was to move moral humility, employing strategies
that maximize effectiveness rather than isolate the precise causal factors might be preferable. Notably, such an
approach has a long history in psychological intervention literature (Wilson et al., 2010; Rozin, 2006) wherein
identifying strategies that can shift an important outcome is given preponderance over identifying or isolating exact
theoretical mechanism or causal factors that can move the outcome of interest (example of such recent works
include Broockman & Kalla, 2016; Voelkel et al., 2023; for a review see Paluck et al., 2021) . The former is
captured in research that is called “problem-oriented” whereas the latter in research that is called “process-oriented”
wherein studying the phenomenon versus studying the process is respectively emphasized — the former is
recognized as a suitable approach especially during initial investigations of a problem or phenomenon where
researchers are still exploring “is” questions (e.g., what is the phenomenon?) (Wilson et al., 2010). A recent example
of problem-oriented research is Broockman and Kalla (2016) where researchers aimed to reduce transgender
prejudice. The problem here was prejudice reduction. To that end, the study had canvassers going door-to-door
talking to people about transgender laws, including showing them a video of opposing views on transgender issues,
asking them to talk about a time when they were judged for being different, asking them to engage in perspective
taking with transgender experiences, and asking for report on whether the whole exercise changed their minds. Thus,
the study employed a variety of strategies to the end of prejudice reduction instead of focusing on isolating and
identifying individual causal factors. The idea behind the experimental studies testing moral humility interventions
was also somewhat similar.
56
It is important to note that the interventions were not designed with the aim of producing large or lasting
effects (such as longitudinal treatments), although I do present combinations of the five moral humility treatments in
Experiment 2 to test if doing so would produce bigger effects than when the treatments are presented individually. In
any case, they were light touch interventions delivered in a single online exposure with the intention of temporarily
boosting a state of humility toward one’s own and others’ morality.
The arguments or ideas presented as part of the vignettes were based on moral psychological and
philosophical work (Baumeister, 1999; Zimbardo, 2004; MacAskill et al., 2020; Williams, 2015; Tersman, 2022;
Cole, 2023). All five moral humility vignettes as well as the one control vignette were largely matched for length
and style of presentation. The participants were asked to read these vignettes as part of a study that aims to
understand people’s thoughts about various philosophical ideas. See SOM for the exact wording and presentation of
the vignettes.
The first vignette highlighted our psychological tendency towards bias and overconfidence. It pointed out
how our views about right and wrong are often biased due to things like our personality, background, and social
relationships, how we are usually blind to our biases and fail to correct them, and on top of it are overconfident in
our views. It suggests that a more accurate understanding of what is morally apt may be easier if people lowered
their confidence in their own moral standpoints and had more openness to learning from people with different moral
standpoints. This vignette thus highlighted the facets of moral fallibility, moral learning and openness, and moral
superiority.
The second vignette highlighted the difficulty of being ethical, such as figuring out the right moral view or
action, living up to our values, developing a good moral character, weighing competing moral ideas and values
accurately, understanding the complexities of different moral situations, and accounting the needs and perspective of
all moral stakeholders. It is suggested to the reader that such complexities mean that our moral judgments might be
prone to error, and that making the right moral choice or being morally upstanding can be difficult. It proposed that
being moral is thus a learning process. This vignette thus highlighted the facets of moral fallibility, and moral
learning and openness.
The third vignette highlighted ordinary people’s capacity of evil. It made salient the fact that it is not
especially evil people who commit moral wrongdoings, such as those in Nazi Germany, Maoist China, and the
Soviet Union, but rather ordinary people like us who deceive themselves of doing a noble deed. It suggests that we
57
should therefore be mindful of our ability to commit horrible misdeeds. This vignette thus highlights the facets of
moral fallibility, moral openness, and moral superiority.
The fourth vignette highlighted humankinds’ morally imperfect past and suggested that like every
generation before us, we are also likely blind to our moral limitations and unknowingly participate in terrible moral
transgressions. It suggested that one way to do a better job of identifying our moral mistakes and correcting them
would be to set up a social environment supporting exchange of ideas so that different people could combine their
knowledge and abilities to identify and correct our moral blind spots. This vignette thus highlighted the facets of
moral fallibility, moral learning and openness, and moral superiority.
The fifth vignette highlighted the existence of moral disagreements amongst equally rational, morally
sensitive, well-informed, and well-intentioned people. It suggests that such disagreement might indicate that other
people might have useful moral perspectives that might help us reach accurate understanding in moral matters. It
also highlighted that people who we morally disagree with often have other moral strengths. This vignette thus
highlighted the facets of moral fallibility, moral learning and openness, and moral superiority.
The control vignette explored whether AI (artificial intelligence) can think and understand as humans. It
presents John Searle’s Chinese Room thought experiment to the reader. This vignette was designed with the
intention of not evoking any content overlapping with moral humility and its facets, while still being engaging.
Possible Theoretical Mechanism. Although identifying and isolating the precise theoretical or
psychological mechanisms undergirding the interventions was not a priority, I speculated that some of the following
factors and mechanism will perhaps drive the interventions’ effects (if any). First, the interventions might increase
the salience of certain aspects undergirding moral humility. For instance, reading about the difficulty of being
ethical, or about our tendencies towards bias or overconfidence, or the atrocities committed by our ancestors might
activate the attribute of moral fallibility thereby making our moral limitations salient.
The interventions might also be educational. The interventions inform and make people aware of our moral
flaws and limitations such as our biases and overconfidence, or of the ordinariness of people who committed well
known historical atrocities. Being educated about these might be a morally humbling experience. These might
therefore move people to recognize the necessity of moral humility.
Along that line, the interventions might implicitly suggest various aspects of moral humility as solutions to
the problems raised in the interventions, such as increasing moral learning/openness or reducing moral superiority as
58
a solution to our biases and overconfidence. Similarly, suggesting recognition of one’s moral fallibility as a solution
to our ability to commit immoral deeds. The interventions therefore implicitly aim to increase the value of moral
humility to the reader and presents it as an important virtue by framing it as a solution to some problems (moral or
otherwise) that people might relate to. The idea is similar to when a political party first makes a problem salient to
the voters and then presents voting for their candidate or party as a solution to fixing that problem. Or an
advertisement from a company makes the consumer salient of an issue or need and then presents their product as a
solution to addressing that issue or need. Such techniques have been called by different names in the persuasion
literature, like problem–reaction–solution (PRS) framing (Drinkwater et al., 2018), or demand or need creation
strategy (Priem et al., 2018). The interventions thus are in a way intended to create a “demand” for moral humility.
The interventions might also evoke certain emotions that might aid in boosting moral humility — for
instance highlighting the moral wrongdoing of others like them or their ancestors might evoke shame,
embarrassment, or fear which might induce humility about morality. However, it is important to note that past work
on humility interventions suggests that exposure to negative self-information is not conducive to inducing humility
by itself unless accompanied by self-affirmation, given that self-affirmation helps put people in a less defensive state
of mind and facilitates acknowledgement of one’s limitations (Ruberton et al., 2016). Along these lines, in order to
not make the reader feel attacked, despondent, or nihilistic when reading about moral flaws and limitations, the
interventions incorporated solutions when presenting problems.
All the mechanisms noted here are often used in psychological interventions (e.g., for salience
manipulation see Burke et al., 2010, Van Tongeren et al., 2016; for educational manipulation see Paluck et al., 2021;
for problem-solution manipulation Tannenbaum et al., 2015; for emotion manipulations see Ruberton et al., 2016). I
conjectured that some or all of these mechanisms might be at play in the moral humility interventions.
59
Pilot Study
Both experiments were preceded by a small Pilot study wherein I tested the five moral humility
interventions that I developed and are described above. The aim was to assess if the five interventions to be used in
the main experiments were well received by the participants, with the intention of making changes in the
interventions accordingly before the main experiments.
Dataset and Participants
The Pilot study was conducted on Prolific. Prolific is an online service that facilitates the crowdsourcing of
research participants (Douglas et al., 2023; Peer et al., 2022). The participants on Prolific are more diverse than the
average college sample than other crowdsourcing platforms like MTurk (Palan & Schitter, 2018) and provide high
quality data (Peer et al., 2022). The Pilot was opened to 50 participants aged over 18 and evenly recruited from those
who self-identified as Democrats and Republicans in Prolific’s prescreening to make sure the interventions are well-
received by participants of diverse ideological leaning. Participants were paid $4 for doing the study (~20 minutes).
The Pilot did not have treatment or control conditions. All participants read all five moral humility treatments and
completed measures that assessed their impressions of the treatments. Demographics was not collected in the Pilot.
Materials and Procedure
Moral Humility Interventions
All participants read and interacted with all five moral humility interventions described in the previous
section and then answered some questions (described next). The control vignette was not included in the pilot-
testing to avoid making the Pilot too long and was directly used in the main experiments.
Outcome Variables
After reading and interacting with the five treatments in the Pilot, the participants answered outcome
measures assessing their evaluation of what they just read. The outcome measures are summarized in Table 12 (table
also contains a summary of measures included in the main experiments).
60
Table 12
Post-treatment outcomes and pre-treatment covariates included in different studies
Pilot
Experiment 1
Experiment 2
Pilot
Experiment 1
Experiment 2
Outcomes
(Post-Treatment)
Moral Humility
Intellectual Humility
Moral Relativism
Political Sectarianism
Selective Exposure
Political Compromise
Epistemic Emotions
Message Quality
Thoughts
Covariates
(Pre-Treatment)
Moral Humility
Political Sectarianism
Selective Exposure
Message Quality: The perceived quality of the vignettes was assessed along two indicators: (i) the ease of
comprehensibility, i.e., if participants found the vignettes easy to read, and (ii) interestingness, i.e., if participants
found the vignettes engaging to read. Participants answered these after reading each vignette. These were intended
to help provide valuable information to evaluate the weaknesses of the vignettes so that changes could be made if
needed. On a 7-point scale from 1 (extremely disagree) to 7 (extremely agree), participants answered the following 2
items after each vignette — “The text was easy to read.” (to assess ease of comprehension), and "I found the content
of the text interesting” (to assess the interestingness).
Emotions. Participants also reported the strength of various epistemic emotions they experienced during
the reading of the vignettes on Epistemically-Related Emotion Scales (Pekrun et al., 2017). The scale is meant to
capture affective states that occur during cognitive activities involving acquisition or generation of knowledge, such
as surprise, curiosity, enjoyment, confusion, anxiety, frustration, and boredom. Assessing these epistemic emotions
in the Pilot provided information on whether the treatments were well-received by participants, which were again
intended to be used to make changes to the treatments before the main experiments if needed. The fourteen items
61
capturing the seven core emotions (two items each) were surprised, amazed, curious, interested, excited, happy,
confused, muddled, anxious, nervous, frustrated, irritated, bored, and monotonous, and were answered on a 7-point
scale (1=Not at All, 4=Moderate, 7=Very Strong). Unlike message quality, these were not assessed after each
vignette, but at the end of reading all vignettes to assess participants’ overall experience with the treatments.
Thoughts. After each vignette, participants were also provided with an open-ended box where they were
asked to briefly respond to “What did you think of the idea you just read? What came to your mind when you read
it? Did you like or dislike anything?” These open-ended responses were used to get a more detailed picture of what
the experience of the participants was when reading the vignettes, and if they negatively responded to some ideas.
Analysis and Results
The aim of the Pilot study was to examine if the five treatments or interventions that would be used in
Experiment 1 and 2 were received well by the participants.
Ease of comprehension
Descriptive statistics for the ease of comprehension for each of five treatments was estimated. Results are
in Table 13 and presented in Figure 9. Results indicated that all treatments were easy to understand, with mean
perceived easiness being above midpoint for all treatments. Further, most ratings clustered above the midpoint,
indicating that most participants generally found the vignettes easy to read.
Table 13
Summary statistics for ease of comprehension across the five treatments
SD
1.19
1.09
1.22
1.24
1.02
Min
Max
2
3
1
2
3
7
7
7
7
7
Mean
Median
biased worldviews
difficulty ethics
evil capacity
imperfect past
moral disagreements
6
5.96
6.12
6.02
6.1
6
6
7
7
6
62
Figure 9
Distribution of the perceived ease ratings across the five treatments
Interestingness
Descriptive statistics for the perceived interestingness of each of five treatments was estimated. Results are
in Table 14 and presented in Figure 10. Results indicated that all five treatments were perceived to be largely
interesting, with mean perceived interestingness above midpoint for all treatments. Further, most ratings clustered
above the midpoint, indicating most participants generally found the vignettes interesting to read.
Table 14
Summary statistics for interestingness across the five treatments
Mean
Median
biased worldviews
difficulty ethics
evil capacity
imperfect past
moral disagreements
5.49
5.7
5.92
5.76
5.78
6
6
6
6
6
63
SD
1.3
1.2
1.02
1.21
1.03
Min
Max
1
3
3
2
4
7
7
7
7
7
Figure 10
Distribution of the perceived interestingness ratings across the five treatments
Emotions
Descriptive statistics for the fourteen epistemic emotions (two items each for seven core emotions) was
estimated. Unlike other indicators, these were not reported for each treatment but asked at the end of all five
treatments to get an overall sense of the emotions evoked by the treatments. Results are in Table 15 and presented in
Figure 11. The results indicate that interest and curiosity were the most strongly experienced emotions, with mean
scores above midpoint for both. This suggests that overall reading the treatments was interesting and curiosity-
inducing for participants. Mean level of emotions that would indicate a negative experience with the treatments such
as muddled, monotonous, irritated, confused, frustrated, and bored were all below midpoint.
64
SD
1.91
1.45
1.79
1.14
1.56
1.64
1.67
1.72
1.37
1.59
1.54
1.52
1.55
1.83
Min
Max
1
1
1
1
1
1
1
1
2
1
1
1
1
1
7
6
7
5
7
7
7
7
7
6
7
7
7
7
Table 15
Summary statistics for epistemic emotions
Mean
Median
amazed
anxious
bored
confused
curious
excited
frustrated
happy
interested
irritated
monotonous
muddled
nervous
surprised
Figure 11
3.12
2.08
2.49
1.84
4.75
3.22
2.25
3.75
5.39
2.08
2.69
2.2
1.96
3.04
Distribution of the strength of epistemic emotions
3
1
2
1
5
3
1
4
6
1
3
2
1
3
65
Thoughts
The open-ended thoughts that participants reported at the end of reading each treatment were also analyzed,
wherein their thoughts were categorized as either positive, negative, mixed, or neutral. Results of the subjective
coding suggested that no treatment was uniquely negatively perceived by the participants to a concerning extent.
There were ~5-7 negative responses for each treatment out of a total of 51-52 responses. Results are in Table 16.
Table 16
Summary of qualitative coding of open-ended thoughts provided at end of each treatment
Positive/
Agreement
Negative/
Disagreement
Mixed
Neutral/
Non-answers
difficulty ethics
biased worldviews
imperfect past
evil capacity
moral disagreements
11
4
13
12
7
Note: Positive/Agreement or Negative/Disagreement is conceptualized broadly. Former includes responses where
participants are echoing thoughts in the text and latter includes responses where participants are expressing a counter
thought to the text presented.
33
38
27
30
35
6
5
6
7
6
1
5
5
3
3
Pilot Summary
The results of the pilot indicated that the text in all treatments was easy to understand, perceived as
engaging, and did not evoke negative emotions or thoughts. Taken together, the indicators suggested that all five
treatments were perceived positively and received well by the participants, and thus these experimental materials
were well-suited for use in the main experiments without modification.
Experiment 1
Experiment 1 tested the five moral humility interventions or treatments individually. The experimental
setup thus had five experimental conditions tested against a control condition. The experiment used a pre-post
design to increase precision (Clifford et al., 2021) wherein outcomes were measured before the treatment as well as
after the treatment. The idea behind these designs is that when the pre-treatment outcome measure is correlated with
the post-treatment outcome measure, it is possible to increase the precision of the estimates and the power of the
study. This study aimed to (i) provide a proof-of-concept for moral humility being amenable to change, ii) to
examine which treatments work best to increase moral humility, and (iii) assess the causal relationship between
moral humility and polarization. All preregistered study details can be found at OSF.
66
Dataset and Participants
Experiment 1 was conducted on Prolific. The study was opened to 2700 US participants aged over 18. The
sample details are in Table 17. Participants were paid $2.60 for doing the study (~13 minutes) which is $12/hr.
Experiment 1 had five experimental conditions and one control conditions. A sample of 2700 people (~ 450 in each
condition) provided good statistical power (~80-85%) in a pre-post experimental design to detect a small effect size
between each of the five treatment conditions with control condition (d ~ 0.15) where the pre and post treatment
outcomes are correlated at .65 (for reference, minimum split-half reliability of moral humility in previous work was
~0.8). If and when the pre- and post-treatment outcomes’ correlations are higher, the achieved statistical power
would be higher.9 Conversely, if the correlations are lower, the achieved statistical power would be lower.
Table 17
Sample characteristics in the two experiments
Sample
Platform
N
Mage
SDage
%
Men
%
Wome
n
%
White
%
Black
%
Other
Ethnic
Prolific
2789
38.42
13.09
49.08
48.26
58.87
25.89
15.24
Prolific
1564
39.04
13.32
48.59
48.21
70.01
9.84
20.15
Sample
Number
Exp 1
Exp 2
Data
Collection
Month
December
2024
February
2025
Materials and Procedure
Independent Variable
Participants were randomly assigned to one of six conditions: five moral humility interventions or
treatments or a control (about artificial intelligence). Interventions are described in SOM.
Dependent Variables
A summary of the dependent variables included in the experiment is in Table 12.
Moral Humility. The main outcome measure was moral humility. This was measured on a 9-item moral
humility scale (α = 0.82), a shorter moral humility scale that included a subset of the 30-item scale constructed and
used in the previous studies. 3-items each were sampled from each of the moral humility subscales based on genetic
algorithm used to construct shorter scales (Schroeder et al., 2016). The items used are indicated in Table 3.
9 This power calculation is different from what was preregistered, as there was a mistake made in the preregistered
power calculation. The preregistered power analysis underestimated the power of the sample.
67
The impact of the interventions was also explored at the subscale level, i.e., with moral learning/openness
(α = 0.85), moral fallibility (α = 0.83), and moral superiority (α = 0.76). This 9-item shorter scale was used instead
of the full scale to keep the study a manageable length.
Political Outcomes. Three political outcomes were measured. Political sectarianism (Finkel et al., 2024)
was assessed as it is a measure meant to capture the moralized nature of partisan disdain in the US context. The 9-
item measure of political sectarianism (Finkel et al., 2024) was used (α = 0.97), where participants indicated their
agreement (1 = strongly disagree, 7 = strongly agree) on the scale’s following items which were averaged to create
a measure: “I am different from the typical [Republican/Democrat].”, “I feel distant from the typical
[Republican/Democrat]”, “No matter how hard I try, I can’t see the world the way the typical
[Republican/Democrat] does.”, “I hate the typical [Republican/Democrat].”, “My feelings toward the typical
[Republican/Democrat] are negative”, “The typical [Republican/Democrat] has lots of negative traits.”, “The typical
[Republican/Democrat] is immoral.”, “The typical [Republican/Democrat] is evil.”, “The typical
[Republican/Democrat] lacks integrity.” The scale included three subscales capturing othering (first three items; α =
0.93), aversion (middle three items; α = 0.91), and moralization (last three items; α = 0.94). Accordingly, I explored
the effect of the interventions on the subscale separately as well, in addition to testing the intervention’s effect on the
whole scale.
Selective exposure was assessed in the same way as in Samples 4 and 5 in Chapter II, wherein participants
indicated how interested they were in hearing from someone who held the opposing view on an important issue (-
100 = very uninterested, 0 = neutral, 100 = very interested). However, for the sake of keeping the study short,
participants completed the selective exposure task for only one issue that they selected was an important issue for
them. Political compromise was measured the same way as in Sample 6 in Chapter II, wherein participants indicated
how likely they were to vote for a compromising candidate and an uncompromising candidate (1 = not at all, 7 =
very likely) who differed in their approach to negotiating on an issue important to the participant. The main
difference in both selective exposure and political compromise tasks compared to the prior studies was in the
assignment of issues for each task, i.e., on which issues the participants completed the task and how it is chosen for
each participant. In Experiment 1, participants indicated the two important issues for them from a list of issues.
These issues were the eleven issues used in Sample 4 and 5. One of these issues was used for the selective exposure
68
task and one for the political compromise task. The importance of the issue and participant’s position on the issue
for selective exposure were both assessed before participants underwent the moral humility intervention.
Moral Relativism and Intellectual Humility. Moral relativism and intellectual humility, two constructs
that have had strong correlations with moral humility in previous studies were also measured to see if and the extent
to which moral humility interventions have an impact on them. Moral relativism was measured using a subset of 10-
item scale (Collier-Spruel et al., 2019). Specifically, six highest loading items were chosen while balancing the
breadth of content covered (α = 0.71). Example items were: “There is a moral standard that all actions should be
held to, even if cultures disagree.”, “There are moral rules that apply to everyone regardless of personal beliefs.”.
See SOM for full list of items. Again, fewer items were used to keep the study a manageable length for participants.
Intellectual humility was measured using the same 6-item (Leary et al., 2017) scale used in all the previous studies
(α = 0.89).
Emotions. As in the Pilot, participants reported the intensity of various epistemic emotions they
experienced during the reading of the interventions on the Epistemically-Related Emotion Scales (Pekrun et al.,
2017). Only the six most informative and simple ones, i.e., surprised, curious, confused, anxious, frustrated, and
bored were included in Experiment 1 to keep the study a manageable length. Participants answered these on a 7-
point scale (1=Not at All, 4=Moderate, 7=Very Strong). Again, the aim was to see if the interventions impacted
these emotions differently, which would be used to provide insight into any possible differences between the
effectiveness of these interventions.
Pre-Treatment Covariates
The experiment used a pre-post experimental design; a pre-post experimental design involves measuring
the dependent or outcome variable both before and after the treatment. This type of design has been shown to
increase precision of estimates as well as provide more power to detect an effect (Clifford et al., 2021), given that
the pre- and post- measures are highly correlated. The design essentially allows the examination of how participants'
attitudes change over the course of a study and whether the pattern of this change differs among those assigned to
different conditions in an experiment.
Accordingly, outcome variables were measured before the treatment as well, which were then treated as
covariates in the analysis (Clifford et al., 2021). Specifically, in Experiment 1, moral humility (Moral Humility α =
0.78; Moral Learning/Openness α = 0.81; Moral Fallibility α = 0.78; Moral Superiority α = 0.71) and political
69
sectarianism (Political Sectarianism α = 0.97; Othering α = 0.91; Aversion α = 0.90; Moralization α = 0.93) were
measured before the treatment (using the same post-treatment measures described above). Pre-treatment measures of
selective exposure and political compromise were not included as I conjectured that pre-treatment political
sectarianism might be highly correlated with all three political outcomes (political sectarianism, selective exposure,
political compromise) and serve as a useful pre-treatment covariate for all. Similarly, pre-treatment moral humility
was considered a pre-treatment measure for intellectual humility and moral relativism also, apart from post-
treatment moral humility. These decisions were made to keep the study a manageable length. If all outcomes were
measured before the treatment, that would have made the study very long. A summary of pre-treatment variables
included in the experiment is provided in Table 12.
Exploratory Variables
Open-Ended Report. Participants also reported their general thoughts in an open-ended box at the end of
the treatments or control. They read: “You read about (moral) ideas that philosophers and psychologists have
thought about, written about, and studied. Please take a few moments to think about the things you read. Please tell
us very briefly: What did you think about the ideas you just read about? Was there any information that challenged
or expanded your thinking (about morality)?”. This was collected for two purposes. Primarily, it was collected under
the guise of the main outcome variable we are interested in as researchers. The whole task or experiment was set up
such that participants were told that we researchers are interested in what they think of various philosophical views.
After reading the various vignettes, this open-ended question would serve as the place where researchers collect the
participant’s views. Secondarily, these responses were intended to be used as a potential exploratory variable which
would be qualitatively analyzed to understand how the participants perceived the vignettes. This was however not a
variable of interest and was more of a filler question as part of the cover story and was intended to be analyzed only
if time permits.
Analysis and Results
The aim of the study was to examine if the five moral humility interventions or treatments each increase
moral humility and decrease the polarization outcomes compared to the control condition, and whether the effect of
the treatments on the polarization outcomes is mediated by moral humility. The effects of the five treatments on
intellectual humility, moral relativism, and epistemic emotions was also examined. All preregistered analyses and
predictions can be found at OSF.
70
Moral Humility
To test the interventions’ effect on moral humility, linear regression was used wherein post-treatment moral
humility was regressed on five dummy-coded condition variables and pre-treatment levels of moral humility. Each
dummy code compared each of the five treatments to the control. The prediction was that people assigned to the five
moral humility intervention conditions will show higher moral humility compared to those in the control condition.
The correlation between pre- and post- moral humility was r = 0.82, p <.001.
Consistent with the predictions, moral humility was significantly higher in each of the five experimental or
treatment conditions compared to the control condition (Table 18). Results are shown in Figure 12. The average
effect size or Cohen’s d across the five treatments was d = 0.23 [0.14, 0.33].
Figure 12
Estimated difference in moral humility between the treatments and control (accounting for pre-treatment levels of
moral humility)
Note: X-axis coefficients (b) are on the original 1-7 scale.
71
Table 18
Estimates (and standard errors) of the difference in moral humility between each of the five treatments versus
control. These are estimates for Figure 12
Biased Worldviews
Difficulty Ethics
Evil Capacity
Imperfect Past
Moral Disagreements
Moral Humility
(Pre-Treatment)
Moral Humility
b
0.18**
(0.04)
0.12**
(0.04)
0.08*
(0.04)
0.14**
(0.04)
0.12**
(0.04)
0.89**
(0.01)
Effect Size
Cohen’s d
0.33
0.22
0.14
0.21
0.22
-
Observations
Note: *p<0.05; **p<0.01. Regression coefficients (b) are on the original 1-7 scale. The effect size or Cohen’s d was
computed by dividing the marginal means by residual standard deviation in an ANOVA.
2,755
The effect of each of the five treatments on the three moral humility subscales was also examined. The
results for the subscales showed that different aspects of moral humility (moral learning/openness, moral fallibility,
moral superiority) contributed to the treatment effect of the five interventions to different extent. For example,
biased worldview and imperfect past interventions most strongly impacted moral fallibility, moral disagreement
intervention most strongly impacted moral learning/openness, and the difficulty ethics and evil capacity
interventions most strongly impacted moral superiority. Results are in Table 19.
Moderation analysis was not preregistered but explored. Specifically, I examined if the effect of the five
interventions on moral humility was moderated by party identity or education to assess if the interventions work
similarly across people of different political leanings and educational levels. To test this, a party identity dummy
variable (Democrat/Democrat leaning = 0, Republican/Republican leaning = 1) and education dummy variable (less
than bachelor’s education = 0, bachelor’s education and higher = 1) was created and an interaction variable between
these and the experimental condition dummy variable was added in the regression. There weren’t any significant
interactions, suggesting that the interventions worked similarly across political leanings and education levels.
72
Finally, I also examined if the five interventions significantly differed from each other using post-hoc
pairwise comparisons in ANOVA. Across ten pairwise comparison, there were no significant differences between
the interventions.
Table 19
Estimates (and standard errors) of the difference in moral humility subscales between each of the five treatments
versus control
Biased Worldviews
Difficulty Ethics
Evil Capacity
Imperfect Past
Moral Disagreements
Moral Learning
(Pre-Treatment)
Moral Fallibility
(Pre-Treatment)
Moral Superiority
(Pre-Treatment)
Observations
Moral Learning
b
0.14**
(0.05)
0.08†
(0.05)
0.05
(0.05)
0.08†
(0.05)
0.12**
(0.05)
0.84**
(0.01)
Moral Fallibility
b
0.26**
(0.06)
0.12†
(0.06)
0.08
(0.06)
0.20**
(0.06)
0.10†
(0.06)
0.77**
(0.01)
2,693
2,694
Moral Superiority
b
-0.09†
(0.06)
-0.15**
(0.06)
-0.10†
(0.06)
-0.07
(0.06)
-0.09
(0.06)
0.76**
(0.01)
2,694
Note: †p<0.1, *p<0.05; **p<0.01. Regression coefficients (b) are on the original 1-7 scale. Marginally significant
results are also highlighted as analysis at subscale level can reduce power and precision.
Taken together, the results suggested that all five moral humility interventions significantly increased moral
humility compared to the control, the five interventions were not significantly different from each other. The
interventions worked similarly in increasing moral humility for Democrats and Republicans and people of different
education levels, with a variation in how strong the effects of the interventions were on the outcome.
Political Outcomes
To test the interventions’ effect on polarization, three political outcomes — political sectarianism, selective
exposure, political compromise — were examined, and a similar test was conducted as moral humility, but on the
political outcomes. That is, linear regression was used to test if the three political variables significantly differed
between each of five treatments and control. For each political outcome analysis, I controlled for pre-treatment
levels of political sectarianism. The prediction was that compared to the control condition, people in the moral
73
humility intervention conditions will express (i) lower political sectarianism, (ii) more interest in cross-cutting
exposure or exposure to opposing political viewpoint, and (iii) more willingness towards political compromise.
Notably, a logic error made in the study’s Qualtrics program led to half the sample getting the wrong the
political outgroup for the political sectarianism measure. This compromised the power for the analyses of the three
political outcomes as approximately half the sample had to be excluded for the preregistered analyses for these
political outcomes. Regardless of this mistake, the experiment was able to still provide good, high-powered (80%
power to detect d = 0.15) information for at least one of these outcomes, i.e., political sectarianism. This was
because the pre-treatment political sectarianism measure was highly correlated with the post-treatment political
sectarianism measure (r = 0.90, p <.001), which helped enhance power. However, the tests for the other two political
outcomes, i.e., selective exposure and political compromise, didn’t have good power (~ 40 - 55% power to detect d
= 0.2) because of the loss of sample and low-correlation with pre-treatment political sectarianism (r ~ 0.15 - 0.30).
Given these, I am more confident in the political sectarianism results than those for selective exposure and political
compromise.
Political Sectarianism. Political sectarianism was significantly higher for three of the five treatments
compared to the control. These treatments were biased worldviews, evil capacity, and moral disagreements (Table
20). Results are shown in Figure 13. The average effect size or Cohen’s d across the five treatments was d = 0.21
[0.11, 0.30].
The effect of each of the five interventions on the three political sectarianism subscales (othering, aversion,
moralization) was also examined. The results for the subscales show that different aspects of political sectarianism
(othering, aversion, moralization) contribute to the treatment effect of the five interventions to different extent
(Table 21). All five interventions had the biggest impact on the othering aspect of political sectarianism, followed
largely by aversion, and then moralization.
Like with moral humility outcome, I conducted moderation analysis (not preregistered, was exploratory) to
examine if the effect of the five interventions on political sectarianism was moderated by party identity or education.
Largely, there weren’t any significant interactions of both with the five interventions, suggesting that the
interventions worked similarly in reducing political sectarianism across political leanings and education levels.
There was one marginally significant interaction — between imperfect past intervention and party identity (b = 0.21,
74
p = 0.059) — suggesting that it might be that imperfect past intervention is more effective for reducing political
sectarianism for those who identify as Democrats than Republicans.
I also examined if the five interventions significantly differed from each other in reducing political
sectarianism using post-hoc pairwise comparisons in ANOVA. Across ten pairwise comparison, I found no evidence
of significant differences between the interventions.
Figure 13
Estimated difference in political sectarianism between the treatments and control (accounting for pre-treatment
levels political sectarianism)
Note: X-axis coefficients (b) are on the original 1-7 scale.
75
Table 20
Estimates (and standard errors) of the difference in political sectarianism between each of the five treatments versus
control. These are estimates for Figure 13
Biased Worldviews
Difficulty Ethics
Evil Capacity
Imperfect Past
Moral Disagreements
Political Sectarianism
b
-0.18**
(0.06)
-0.07
(0.06)
-0.12*
(0.06)
-0.09
(0.06)
-0.16**
(0.06)
0.96**
(0.06)
Effect Size
Cohen’s d
0.30
0.11
0.21
0.14
0.27
-
Political Sectarianism
(Pre-Treatment)
Observations
Note: *p<0.05; **p<0.01. Regression coefficients (b) are on the original 1-7 scale.
1369
Table 21
Estimates (and standard errors) of the difference in political sectarianism subscales between each of the five
treatments versus control
Political Sectarianism
Othering b
Political Sectarianism
Aversion b
Biased Worldviews
Difficulty Ethics
Evil Capacity
Imperfect Past
Moral Disagreements
Political Sectarianism
Othering (Pre-Treatment)
Political Sectarianism
Aversion (Pre-Treatment)
Political Sectarianism
Moralization (Pre-Treatment)
-0.22**
(0.06)
-0.12
(0.06)
-0.13†
(0.06)
-0.16*
(0.06)
-0.19*
(0.06)
0.93**
(0.01)
-0.19**
(0.07)
-0.08
(0.07)
-0.13†
(0.07)
-0.05
(0.07)
-0.16*
(0.07)
0.93**
(0.011)
Political
Sectarianism
Moralization b
-0.13†
(0.07)
-0.01
(0.07)
-0.11
(0.07)
-0.05
(0.07)
-0.12
(0.07)
0.93**
(0.012)
Observations
†p<0.1, *p<0.05; **p<0.01. Regression coefficients (b) are on the original 1-7 scale. Marginally significant results are
also highlighted as analysis at subscale level can reduce power and precision.
1,369
76
Political Sectarianism Mediation. The results of the mediation analysis are in Table 22, wherein I tested
the extent to which moral humility explained the effects of the interventions on political sectarianism. The ACME
(Average Causal Mediation Effect) tells us how much of the effect of the interventions on political sectarianism is
mediated by moral humility. Thus, it is the indirect effect of the interventions on political sectarianism through
moral humility. If ACME is significant (i.e., confidence interval doesn’t contain zero), it suggests that moral
humility as the mediator plays a significant role in explaining the relationship between the interventions (moral
humility treatments) and political sectarianism (outcome).
The ADE (Average Direct Effect) is part of effect of the interventions on political sectarianism that is not
mediated by moral humility. Thus, it is the direct effect of the interventions on political sectarianism after
accounting for moral humility. If ADE is significant, it means the interventions still influence political sectarianism
even after controlling for moral humility. The Total Effect is the overall effect of the interventions on political
sectarianism (ACME + ADE). The Proportion Mediated tells us how much of the Total Effect is explained moral
humility.
The expectation was that the ACME’s will be negative and significant, indicating that the interventions
increase moral humility and an increase in moral humility leads to decrease in political sectarianism. Result showed
that moral humility significantly mediated (ACME) the effect of the four interventions on political sectarianism.
These were: biased worldviews, imperfect past, evil capacity, and moral disagreements. The interventions also had
significant direct and total effects on political sectarianism for two of these interventions, i.e., biased worldviews
and moral disagreements.
Table 22
Mediation results for the effect of the interventions on political sectarianism via moral humility
Difficulty
Ethics
Biased
Worldviews
-0.01
-0.05
-0.06
0.19
-0.04**
-0.12*
-0.16**
0.27
Imperfect
Past
-0.03**
-0.02
-0.05
0.66
Evil
Capacity
-0.04**
-0.08
-0.12*
0.33
Moral
Disagreements
-0.03**
-0.12*
-0.15**
0.20
ACME
ADE
Total Effect
Prop Mediated
Note: *p<0.05; **p<0.01.
I also conducted sensitivity analysis to address the sequential ignorability assumption (Imai et al., 2010;
Tingley et al., 2014). To that end, the correlation between the residuals of the mediator and outcome regressions was
77
chosen as the sensitivity parameter. This is because the relationship between moral humility and political outcomes
was not causal, and thus there existed the possibility of the existence of unobserved pre-treatment confounders
affecting both the mediator and the outcome, making the correlation between the residuals not zero. The sensitivity
analysis varied the value of this correlation between -0.9 and +0.9 by 0.1 increments and examined how the
estimated indirect effect changes. The results of this sensitivity analysis showed that across the significant mediation
models, a correlation of r ~ 0.1- 0.2 due to unmeasured confounders could nullify the mediation effect.
Taken together, the results suggested that only three of the five moral humility interventions — biased
worldviews, evil capacity, and moral disagreements — significantly reduced political sectarianism compared to the
control. The effects of all three was mediated by moral humility. However, the five interventions were not
significantly different from each other. Two of these might have been not significantly different from control due to
weaker treatment effect (e.g., difficulty ethics) or heterogenous effects across groups weakening the overall
treatment effect (e.g., imperfect past); in fact, for one of these (i.e., imperfect past), there was a significant indirect
effect through moral humility. However, sensitivity analysis suggests that the mediation models in general are very
sensitive to the existence of confounders. Overall, the results suggested that the interventions worked similarly in
reducing political sectarianism, had similar effects for Democrats and Republicans and people of different education
levels, with a variation in how strong the effects of the five interventions were on the outcome.
Selective Exposure. The willingness towards cross-cutting exposure or engaging with opposing political
viewpoint on the participant’s self-selected important issue was significantly higher for three of the five treatments
compared to the control. These treatments were difficulty ethics, evil capacity, and imperfect past (Table 23).
Results are shown in Figure 14. The average effect size or Cohen’s d across the five treatments was d = 0.2 [0.13,
0.25].
78
Figure 14
Estimated difference in selective exposure between the treatments and control (accounting for pre-treatment levels
of political sectarianism)
Note: X-axis coefficients (b) are on the original -100 - +100 scale. Higher and positive values indicate more interest
in exposure to opposing viewpoint.
Table 23
Estimates (and standard errors) of the difference in selective exposure between each of the five treatments versus
control. These are estimates for Figure 14
Selective Exposure
b
Effect Size
Cohen’s d
Biased Worldviews
Difficulty Ethics
Evil Capacity
Imperfect Past
Moral Disagreements
Political Sectarianism
(Pre-Treatment)
Observations
0.13
0.23
0.23
0.25
0.16
-
1,058
7.84
(6.46)
13.90*
(6.40)
13.84*
(6.42)
15.04*
(6.51)
9.78
(6.45)
-13.06**
(1.142)
79
Table 23 (cont’d)
Note: *p<0.05; **p<0.01. Regression coefficients (b) are on the original -100 - +100 scale.
Like before, I conducted moderation analysis (not preregistered, was exploratory) to examine if the effect
of the five interventions on selective exposure was moderated by party identity or education. Largely, there weren’t
any significant interactions of both with the five interventions, suggesting that the interventions worked similarly in
reducing political sectarianism across political leanings and education levels. There was one significant interaction
— between imperfect past intervention and education level (b = 27.47, p = 0.043) — suggesting that it might be that
imperfect past intervention is more effective for increasing interest in cross cutting exposure for those with
bachelor’s education or higher.
I also examined if the five interventions significantly differed from each other in increasing cross-cutting
exposure using post-hoc pairwise comparisons in ANOVA. Across ten pairwise comparison, I found no evidence of
significant differences between the interventions.
Selective Exposure Mediation. The results of the mediation analysis are in Table 24, wherein I tested the
extent to which moral humility explained the effects of the interventions on interest in cross-cutting exposure. The
expectation was that the ACME’s will be positive and significant, indicating that the interventions increase moral
humility and an increase in moral humility leads to increase in interest in cross-cutting exposure.
Results showed that moral humility mediated (ACME) the effect of the four interventions on willingness to
engage with opposing political viewpoint either significantly or marginally significantly. These were: difficulty
ethics, biased worldviews, imperfect past, and evil capacity. There were two (difficulty ethics, evil capacity) and
three (difficulty ethics, evil capacity, moral disagreements) marginally significant or significant direct and total
effects respectively.
80
Table 24
Mediation results for the effect of the interventions on selective exposure via moral humility
Difficulty
Ethics
Biased
Worldviews
Imperfect
Past
Evil
Capacity
Moral
Disagreements
ACME
ADE
Total Effect
2.05†
12.23*
14.41*
4.65**
0.41
5.05
4.36*
6.45
10.82
2.42†
9.89†
12.31*
1.80
8.08
9.88†
Prop Mediated
Note: †p<0.1, *p<0.05; **p<0.01. Marginally significant results are also highlighted as analysis had lower power due
to exclusion of sample that got wrong political items.
0.40
0.20
0.15
0.18
0.91
I also conducted sensitivity analysis, following the same procedure as used before for political sectarianism
mediation models. Like before, the results of the sensitivity analysis showed that across the mediation models, a
correlation of r ~ 0.1- 0.2 due to unmeasured confounders could nullify the mediation effect.
Taken together, the results suggested that only three of the five moral humility interventions — difficulty
ethics, evil capacity, and imperfect past — significantly increase cross-cutting exposure compared to the control.
The effects of all three was mediated by moral humility. However, the five interventions were not significantly
different from each other. Two of these might have been not significantly different from control due to weaker
treatment effects; in fact, for one of these (i.e., biased worldviews), there was a significant indirect effect through
moral humility. However, sensitivity analysis suggests that the mediation models in general are sensitive to the
existence of confounders. Overall, the results suggested that the interventions worked similarly in increasing cross-
cutting exposure, had similar effects for Democrats and Republicans and people of different education levels, with a
variation in how strong the effects of the five interventions were on the outcome. However, all of these results
should be interpreted tentatively until they are replicated, given the lower power (~ 60%) for the selective exposure
analyses.
Political Compromise. The five treatments did not significantly impact support towards uncompromising
candidate (candidate A), the compromising candidate (candidate B), or compromising over uncompromising
candidate (candidate B-A) (Table 25) compared to the control. The average effect size or Cohen’s d across the five
treatments for support towards candidate A was d = 0.08 [0.01, 0.16], towards candidate B was d = -0.08 [-0.18,
0.05], and towards candidate B v A was d = 0.09 [-0.18, 0.02].
81
Table 25
Estimates (and standard errors) of the difference in support for candidate A (uncompromising candidate), candidate
B (compromising candidate), and candidate B vs A (uncompromising over compromising candidate), between each
of the five treatments versus control
Candidate A
b
Candidate A
d
Candidate B
b
Candidate B
d
Candidate B-
A
b
Candidate B-
A
d
Biased Worldviews
Difficulty Ethics
Evil Capacity
Imperfect Past
Moral Disagreements
Political Sect
0.25
(0.18)
0.16
(0.18)
0.30
(0.18)
0.02
(0.18)
0.12
(0.18)
0.18**
(0.03)
0.13
0.08
0.16
0.01
0.06
-
-0.12
(0.14)
-0.07
(0.14)
-0.27†
(0.14)
0.07
(0.14)
-0.21
(0.14)
-0.14**
(0.02)
-0.08
-0.05
-0.18
0.05
-0.15
-
-0.36
(0.28)
-0.24
(0.28)
-0.53†
(0.28)
0.05
(0.28)
-0.31
(0.28)
-0.32**
(0.05)
-0.12
-0.08
-0.18
0.02
-0.11
-
Observations
1,355
1,367
1,353
Note: †p<0.1, *p<0.05; **p<0.01. Regression coefficients (b) are on the original 1-7 scale. Marginally significant
results are also highlighted as analysis had lower power due to exclusion of sample that got wrong political items.
As before, I conducted moderation analysis (not preregistered, was exploratory) to examine if the effects of
the five interventions on the three political compromise measures were moderated by party identity or education.
There weren’t any significant interactions of party identity or education with the five interventions for all three
political compromise measures, suggesting that the interventions worked similarly for political compromise across
diverse political leanings and education levels.
I also examined if the five interventions significantly differed from each other in increasing political
compromise across the three measures using post-hoc pairwise comparisons in ANOVA. Across ten pairwise
comparisons, I found no evidence of significant differences between the interventions.
Political Compromise Mediation. The results of the mediation analysis are in Table 26, wherein the extent
to which moral humility explained the effects of the interventions on support for uncompromising (candidate A),
82
compromising (candidate B), and compromising over uncompromising (candidate B-A) candidate was examined.
The expectation was that the ACME for support for the uncompromising (candidate A) will be negative and
significant, indicating that the interventions increase moral humility and an increase in moral humility leads to
decrease in support for uncompromising (candidate A). Next, for support for the compromising (candidate B) and
support for the compromising over uncompromising candidate (candidate B-A), the expectation was that ACME’s
will be positive and significant, indicating that the interventions will increase moral humility and an increase in
moral humility leads to an increase in support for compromising (candidate B) and support for compromising over
uncompromising candidate (candidate B-A).
Results showed that the ACME’s tended to be significantly or marginally significant in the direction that
was expected for almost all five interventions (Table 26). The interventions increased moral humility and
consequently — decreased support for uncompromising candidate, increased support for the compromising
candidate, and increased support for the compromising over uncompromising candidate. However, the direct and
total effects (whenever significant or marginally significant) were in the opposite direction to indirect effects. This
pattern of direct and indirect results being in the opposing directions could be because it is possible that while the
interventions do increase willingness towards political compromise via increased moral humility, they also decrease
willingness towards political compromise via another mechanism. In other words, there might be competing
mediators that counteract the positive effects of moral humility, resulting in overall negative effect.
Further, I also conducted sensitivity analysis, following the same procedure as used before for political
sectarianism and selective exposure mediation models. Like before, the results of the sensitivity analysis showed
that across the mediation models, a correlation of r ~ 0.1- 0.2 due to unmeasured confounders could nullify the
mediation effect.
In any case, all of these results should be interpreted tentatively, given the non-significant effects of the
interventions on political compromise outcomes (Table 25) and the pretty low power (~40%) for these analyses.
83
Table 26
Mediation results for the effect of the interventions on support for candidate A (uncompromising candidate),
candidate B (compromising candidate), candidate B vs A (uncompromising over compromising candidate) via
moral humility
Difficulty
Ethics
Biased
Worldviews
Imperfect
Past
Evil
Capacity
Moral
Disagreements
ACME
ADE
Total Effect
Prop Mediated
ACME
ADE
Total Effect
Prop Mediated
ACME
ADE
Total Effect
Candidate A (Uncompromising)
-0.11**
0.37*
0.27
-0.39
-0.08*
0.17
0.09
-0.82
Candidate B (Compromising)
0.09*
-0.24†
-0.16
-0.57
0.10**
-0.13
-0.03
-3.05
-0.05
0.2
0.16
-0.31
0.04*
-0.13
-0.09
-0.49
Candidate B-A (Compromising v Uncompromising)
0.09*
-0.34
-0.25
0.2**
-0.62*
-0.42
0.18**
-0.30
-0.12
-0.05†
0.37*
0.31†
-0.18
0.09**
-0.38**
-0.29*
-0.3
0.15**
-0.72**
-0.57*
-0.12**
0.25
0.13
-0.87
0.11**
-0.34**
-0.23
-0.50
0.23**
-0.58*
-0.35
Prop Mediated
Note: †p<0.1, *p<0.05; **p<0.01. Marginally significant results are also highlighted as analysis had lower power due
to exclusion of sample that got wrong political items.
-0.47
-1.51
-0.37
-0.69
-0.27
Taken together, the results suggested that the interventions worked similarly, i.e., didn’t reliably impact
political compromise across all political compromise outcomes, and had similar effects across political leanings and
education levels. There was some suggestive evidence that the interventions might indeed be increasing willingness
towards political compromise via increased moral humility, but that there are also other mechanisms at play which
might be reducing or counteracting the positive effects of moral humility on political compromise. These results
would need to be replicated in future studies with good power before any of these conclusions can be made
confidently or warrant any further investigation or speculation.
Intellectual Humility and Moral Relativism
The analysis for moral relativism and intellectual humility followed the same analytic strategy as that for
moral humility. The correlation between pre-treatment moral humility and post-treatment intellectual humility was r
= 0.56, p <.001, and with post-treatment moral relativism was r = 0.47, p <.001.
84
All five interventions did not significantly in(de)crease intellectual humility or moral relativism (Table 27).
The average effect size or Cohen’s d across the five treatments for intellectual humility was d = -0.01 [-0.05, 0.01],
and for moral relativism was d = -0.02 [-0.05, 0.07]. Again, these effects were not moderated by party identity or
educational levels.
Taken together, these results (compared to moral humility results which increased across all five
treatments) provide evidence for moral humility’s discriminant validity with both these psychological constructs.
Table 27
Estimates (and standard errors) of the difference in intellectual humility and moral relativism between each of the
five treatments versus control
Intellectual
Humility
b
0.01
(0.06)
-0.05
(0.06)
-0.05
(0.06)
0.01
(0.06)
0.01
(0.06)
0.63**
(0.02)
Intellectual
Humility
d
0.006
-0.04
-0.04
0.01
0.01
-
Moral
Relativism
b
-0.05
(0.06)
0.02
(0.06)
-0.02
(0.06)
-0.06
(0.06)
0.07
(0.06)
0.53**
(0.02)
2,692
2,691
Moral
Relativism
d
-0.05
0.03
-0.02
-0.07
0.08
-
Biased Worldviews
Difficulty Ethics
Evil Capacity
Imperfect Past
Moral Disagreements
Moral Humility
(pre-treatment)
Observations
Note: *p<0.05; **p<0.01
Epistemic Emotions
Six epistemic emotions — surprise, curiosity, confusion, anxiety, frustration, and boredom — were
measured in all five experimental and control conditions. This was to assess if the interventions differed in the
emotions they evoked, which would help understand any differences in the impact of these interventions on moral
humility or political outcomes. All six conditions were compared pairwise using ANOVA (adjusting for Tukey’s
multiple comparisons). The results are in Table 28.
Results indicated that the moral disagreement treatment made people significantly less surprised and
curious compared to the biased worldview and imperfect past treatments. There were no differences in confusion
and boredom across all pairwise comparisons. Compared to the control, evil capacity and imperfect past vignettes
evoked significantly more frustration. Evil capacity treatment evoked significantly more anxiety and frustration than
85
the biased worldviews and moral disagreement treatments. Evil capacity also evoked more frustration compared to
the difficulty ethics treatment.
Overall, there was largely a lack of very consistent patterns in these comparisons, barring one. Evil capacity
treatment appeared to have the most impact on the negative emotions, especially anxiety and frustration. This might
help explain why that treatment was the one that had the smallest impact on moral humility, although as we saw
before, the impact was not significantly different than the effect of other treatments. Thus, at this point, it was
unclear what the higher levels of frustration and anxiety in the evil capacity treatment might mean substantively for
its effect on outcomes.
Table 28
Pairwise comparisons of epistemic emotions between the six (five treatments and one control) conditions
Contrast
Artificial Intelligence -
Biased Worldviews
Artificial Intelligence -
Difficulty Ethics
Artificial Intelligence -
Evil Capacity
Artificial Intelligence -
Imperfect Past
Artificial Intelligence -
Moral Disagreements
Biased Worldviews -
Difficulty Ethics
Biased Worldviews -
Evil Capacity
Biased Worldviews -
Imperfect Past
Biased Worldviews -
Moral Disagreements
Difficulty Ethics -
Evil Capacity
Difficulty Ethics -
Imperfect Past
Difficulty Ethics -
Moral Disagreements
Evil Capacity -
Imperfect Past
Evil Capacity -
Moral Disagreements
Imperfect Past -
Moral Disagreements
Note: *p<0.05; **p<0.01
Surprised
-0.05
(0.12)
0.11
(0.12)
0.07
(0.12)
-0.02
(0.12)
0.32
(0.12)
0.16
(0.12)
0.12
(0.12)
0.03
(0.12)
0.38
(0.12) *
-0.04
(0.12)
-0.13
(0.12)
0.22
(0.12)
-0.10
(0.12)
0.25
(0.12)
0.35
(0.12) *
Curious
0.01
(0.11)
0.13
(0.11)
0.25
(0.11)
0.01
(0.11)
0.42
(0.12) **
0.12
(0.11)
0.24
(0.11)
-0.00
(0.11)
0.41
(0.11) **
0.12
(0.11)
-0.13
(0.11)
0.29
(0.12)
-0.24
(0.11)
0.17
(0.11)
0.41
(0.11) **
Confused
-0.03
(0.09)
-0.10
(0.09)
-0.08
(0.09)
-0.07
(0.09)
0.15
(0.09)
-0.06
(0.09)
-0.05
(0.09)
-0.04
(0.09)
0.18
(0.09)
0.01
(0.09)
0.03
(0.09)
0.25
(0.09)
0.02
(0.09)
0.24
(0.09)
0.22
(0.09)
Anxious
0.10
(0.10)
-0.04
(0.10)
-0.28
(0.10)
-0.13
(0.10)
0.11
(0.10)
-0.13
(0.10)
-0.38
(0.10) **
-0.22
(0.10)
0.01
(0.10)
-0.24
(0.10)
-0.09
(0.10)
0.14
(0.10)
0.15
(0.10)
0.39
(0.10) **
0.24
(0.10)
Frustrated
-0.14
(0.10)
-0.11
(0.10)
-0.53
(0.10) **
-0.30
(0.10) *
-0.07
(0.10)
0.03
(0.10)
-0.39
(0.09) **
-0.16
(0.10)
0.07
(0.10)
-0.42
(0.10) **
-0.19
(0.10)
0.04
(0.10)
0.24
(0.09)
0.47
(0.10) **
0.23
(0.10)
Bored
-0.13
(0.09)
-0.15
(0.09)
0.01
(0.09)
0.01
(0.09)
-0.10
(0.10)
-0.02
(0.09)
0.14
(0.09)
0.14
(0.09)
0.03
(0.10)
0.16
(0.09)
0.16
(0.09)
0.05
(0.10)
0.01
(0.09)
-0.11
(0.10)
-0.11
(0.10)
86
Experiment 1 Summary
This study tested five moral humility treatments against a control. Specifically, their impact on moral
humility, three political outcomes, epistemic emotions, intellectual humility, and moral relativism was examined.
All five treatments —biased worldviews, difficulty ethics, evil capacity, imperfect past, and moral disagreements —
significantly increased moral humility. Further, all five interventions were not significantly different from each
other. These results together thus suggested that all five treatments can be used as treatments designed to move
moral humility (though the strength of the treatments vary across the treatments). Thus, these results served as a
kind of a successful manipulation check. Further, they also provided evidence of the amenability of moral humility.
However, it is worth noting that the effect sizes were small-to-moderate, suggesting that it would be useful in future
work to find ways to amplify the effect of the treatments further.
The interventions’ downstream impact on polarization was also assessed across three political outcomes —
political sectarianism (capturing partisan animosity), selective exposure (capturing people’s tendency towards cross-
cutting exposure with the opposing political side), and political compromise (capturing people’s willingness to
compromise on important political issues). The treatments significantly reduced political sectarianism and increased
openness to cross-cutting exposure, but didn’t have an impact on political compromise. Specifically, three
treatments (biased worldviews, evil capacity, and moral disagreements) significantly reduced political sectarianism,
while three (difficulty ethics, evil capacity, and imperfect past) significantly increased interest in exposing oneself to
opposing political viewpoints. Notably, across both political outcomes, all five treatments were not significantly
different from each other. Mediation analyses largely supported a mechanistic/causal interpretation wherein the
treatments increased moral humility, which in turn had a downward impact on political sectarianism and selective
exposure in expected ways. However, sensitivity analyses suggested the mediation effect is susceptible to
unmeasured confounders.
Two psychological constructs, intellectual humility and moral relativism share the most conceptual
similarity with moral humility and have been found to be empirically strongly correlated with moral humility in my
previous work. Thus, in this study, the extent to which these treatments, designed to increase moral humility, also
affect these two other constructs was examined. The results would give insight into how close these constructs are to
moral humility. If they both moved in the same direction and to the same extent as moral humility in response to the
interventions, this would suggest that these constructs are very close to each other in the nomological network.
87
Results indicated that intellectual humility and moral relativism did not move in the same direction or to the same
extent as moral humility across these treatments. The results for these two constructs were largely non-significant.
Taken together, these provided evidence for moral humility’s distinct nature from these two constructs. These
constructs perhaps then may be close to each other but maintain their distinctiveness.
Finally, the epistemic emotions experienced by participants were also examined in order to provide insight
into and explain any important differences between the five treatments. Broadly, there weren’t any pattern of
meaningful differences in the emotions (surprise, curiosity, confusion, anxiety, frustration, boredom) evoked by the
five treatments. There was some evidence suggesting that the evil capacity treatment evoked more frustration and
anxiety compared to other treatments. However, given that the treatments did not perform significantly differently
from other treatments when it came to the outcomes, taken in the larger context, evil capacity’s negative valence did
not change the big picture interpretations and inferences made from the study at this point.
Thus, the overall picture suggested that the five treatments were successful in increasing moral humility
and decreasing political sectarianism. It also increased interest in cross-cutting exposure (and its impact on political
compromise was inconclusive). These latter two results would although need to be tested in high-powered samples
before any strong conclusions can be drawn. Additionally, moral humility appeared to be distinct from intellectual
humility and moral relativism. There are some caveats. To reiterate, only the analyses for moral humility, political
sectarianism, moral relativism, intellectual humility, and epistemic emotions were highly powered. Other analyses
had lower power and thus their results deserve caution in interpretation. Further, while mediation analyses supported
a causal pathway, their robustness is weak due to unmeasured confounders. Finally, the effect sizes for moral
humility and political outcomes such as political sectarianism and selective exposure were small-to-moderate. This
was expected as the interventions were designed as light-touch interventions. However, their practical value might
depend on finding ways to enhance the impact of these interventions.
88
Experiment 2
Experiment 1 found that the five treatments worked similarly across outcomes but had small-to medium
sized effects. Experiment 2 built up on Experiment 1. The main aims of Experiment 2 were twofold: (i) to examine
if increasing the dose of moral humility by combining treatments would make the effects on the outcomes stronger,
(ii) to replicate the results of Experiment 1, especially given the lower power of some of the analyses in Experiment
1. Experiment 2 had two experimental conditions, a low-dose and a high-dose condition, and a control condition.
The low-dose condition was like Experiment 1 wherein participants randomly received only one the five treatments
individually. Given that the treatments worked largely similarly in Experiment 1, the five interventions were treated
as multiple stimuli instantiating a common construct, allowing to see if the results can be generalized across the five
treatments or stimuli. Further, this condition allowed the replication of Experiment 1 as the treatments were
presented individually. The high-dose condition involved participants receiving a random selection of two of the
five treatments in a random order; this served to strengthen the boost in moral humility received by the participants.
These two experimental conditions were tested against a control condition. The control was same as Experiment 1
wherein participants read about artificial intelligence. The two experimental conditions were also compared to each
other to see if the high-dose condition was significantly stronger the low-dose condition. Finally, mediation of the
two treatment conditions on the political outcomes was also examined. All preregistered study details can be found
at OSF. Like Experiment 1, Experiment 2 also used a pre-post experimental design to increase precision and power
(Clifford et al., 2021) such that outcomes were measured before the treatment as well as after the treatment.
Dataset and Participants
Experiment 2 was conducted on Prolific. The study was opened to 1500 US participants aged over 18. The
sample details are in Table 17. Participants were paid $2.40 for doing the study (~12 minutes). Experiment 2 had
two experimental conditions (low and high-dose) and one control condition. A sample of 1500 people (~ 500 in
each condition) provided good statistical power (>90%) in a pre-post experimental design to detect a small effect
size between each of the experimental and control conditions (d ~ 0.15) where the pre and post treatment outcomes
are correlated at .7 (for reference, pre- and post-treatment outcomes were correlated at ~ 0.8 - 0.9 in Experiment 1).
If and when the pre- and post-treatment outcomes’ correlations are higher, the achieved statistical power would be
higher. Conversely, if the correlations are lower, the achieved statistical power would be lower.
89
This sample in Experiment 2 also enabled the detection of a difference of approximately d ~ 0.15 between
the two experimental groups (high vs. low dose) with strong power (>90%). This means that if the low-dose
condition (compared to the control) has an effect size of d ~ 0.20, there was >90% power to detect a difference
between the low- and high-dose conditions when the high-dose effect is 75% larger (i.e., d ~ 0.35). If the high-dose
effect is 50% larger than the low-dose effect (i.e., d ~ 0.30), the power to detect this difference drops to about 59%.
However, for effects exceeding a 75% increase, power approaches 99%. Additionally, if pre- and post-treatment
outcomes are more strongly correlated than 0.7, power improves further; for instance, at r ~ 0.80, a 50% larger
effect in the high-dose condition could be detected with approximately 75% power.
Materials and Procedure
Independent Variable
Participants were randomly assigned to one of three conditions: the low-dose or high-dose intervention
condition, or a control (about artificial intelligence). In the low-dose condition, participants received one of the five
treatments from Experiment 1 randomly. In the high-dose condition, participants received two randomly selected
treatments of the five treatments from Experiment 1 in a randomized order.
Dependent Variables
A summary of dependent variables included in the experiment is provided in Table 12. Experiment 2
included fewer outcome measures than Experiment 1 to compensate for the increase in the length of the study due to
double the number of interventions (in high-dose condition) and more pre-treatment measures (see pre-treatment
covariates section below). Specifically, it didn’t include measures of intellectual humility, moral relativism, and
epistemic emotions as these weren’t the main outcomes of interest and were analyzed with good power before.
Further, number of political outcomes measured was also reduced to two — with the longest political measure,
political compromise removed in Experiment 2.
Moral Humility. Moral humility was measured just like Experiment 1, using the 9-item moral humility
scale (α = 0.86). The impact of the interventions was also explored at the subscale level, i.e., with moral
learning/openness (α = 0.87), moral fallibility (α = 0.85), and moral superiority (α = 0.77).
Political Outcomes. Two political outcomes were measured. Political sectarianism (Finkel et al., 2024)
was assessed just like Experiment 1, using the 9-item measure of political sectarianism (α = 0.96). The impact of the
90
interventions was also explored at the subscale level, i.e., othering (α = 0.90), aversion (α = 0.90), and moralization
(α = 0.94).
Selective exposure was again assessed in the same way as in Experiment 1 wherein participants answered
how interested they were in hearing from someone who held the opposing view on an issue that was selected as
important by the participant. Accordingly, in Experiment 2, participants were asked to choose only one issue that
was important to them which was then was used in the selective exposure task. Like Experiment 1, the importance
of the issue and participant’s position on the issue were both assessed before participants underwent the moral
humility intervention.
Pre-Treatment Covariates
Experiment 2 also used a pre-post experimental design wherein outcome variables were measured before
the treatment as well—which were then treated as covariates in the analysis. Like Experiment 1, the variables
included were: moral humility (Moral Humility α = 0.83; Moral Learning/Openness α = 0.83; Moral Fallibility α =
0.78; Moral Superiority α = 0.70) and political sectarianism (Political Sectarianism α = 0.96; Othering α = 0.90;
Aversion α = 0.90; Moralization α = 0.94). Additionally, I added a pre-treatment measure of selective exposure
(unlike Experiment 1) after observing a small correlation between pre-treatment political sectarianism and post-
treatment selective exposure in Experiment 1, which had led to lower power for the selective exposure analyses. I
did this because selective exposure was an important outcome, so it was important to enhance the power of its
analyses. All three were measured in the same way as their respective post-treatment measures. A summary of pre-
treatment measures included in Experiment 2 is provided in Table 12.
Exploratory Variables
Open-Ended Report. Like Experiment 1, participants reported their general thoughts in an open-ended
box at the end of the treatments or control. This was again not a variable of interest and was more of a filler question
as part of the cover story of the study and was intended to be analyzed only if time permits.
Analysis and Results
The aim of the study was to examine if the low-dose and high-dose moral humility treatments increase
moral humility and decrease the polarization outcomes compared to the control condition, and whether the effect of
the high-dose treatment is stronger than the low-dose treatment. Additionally, whether the effect of both treatments
91
on the polarization outcomes is mediated by moral humility was also examined. All preregistered predictions and
analyses can be found at OSF.
Moral Humility
To test the treatments’ effects on moral humility, linear regression was used wherein post-treatment moral
humility was regressed on two dummy-coded condition variables and pre-treatment levels of moral humility. Each
dummy code compared the low-dose and high-dose treatments to the control (Model 1, Table 29). The prediction
was that people assigned to the low- and high-dose treatment conditions will show higher moral humility compared
to those in control condition. The correlation between pre- and post- moral humility was r = 0.85, p <.001,
suggesting a high power to detect small effects. Moral humility was significantly higher in the high-dose condition
compared to the control condition; however, the low-dose condition was not significantly higher in moral humility
compared to control (Model 1, Table 29).
To test if the two experimental groups (low- v high-dose moral humility treatment) significantly differed
from each, the reference group in the regression was changed to be the low-dose condition, allowing the comparison
of the two experimental groups. Again, pre-treatment level of moral humility was included as a covariate in the
analysis. The prediction was that people assigned to the high-dose treatment condition will show higher moral
humility compared to those in low-dose condition. The high-dose condition did have a significantly stronger impact
on moral humility compared to the low-dose condition by almost 40%. (Model 2, Table 29).
Table 29
Estimates (and standard errors) of the difference in moral humility between each of two treatments (low and high
dose) and control (Model 1), and between the two treatments (low vs high dose) (Model 2)
Model 1 b
(Reference =
Control)
0.12**
(0.03)
0.05
(0.03)
High Dose
Low Dose
Control
Moral Humility
(Pre-Treatment)
0.92**
(0.02)
Moral Humility
Effect Size
Cohen’s d
0.24
0.10
-
Model 2 b
(Reference = Low
Dose)
0.07*
(0.03)
-0.05
(0.03)
0.92**
(0.02)
Effect Size
Cohen’s d
0.14
-0.10
-
Observations
Note: *p<0.05; **p<0.01. Regression coefficients (b) are on the original 1-7 scale.
1,540
92
The effect of the low and high dose treatments compared to the control on the three moral humility
subscales was also examined. The results for the subscales showed that different aspects of moral humility (moral
learning/openness, moral fallibility, moral superiority) contribute to the treatment effect to different extent. The
strongest effects were on moral fallibility, followed by moral learning/openness, with negligible effects on moral
superiority. Results are in Table 30.
Table 30
Estimates (and standard errors) of the difference in moral humility subscales between each of two treatments (low
and high dose) and control
Moral Learning b
Moral Fallibility b
Moral Superiority b
0.14**
(0.04)
0.06
(0.04)
0.88**
(0.01)
High Dose
Low Dose
Moral Learning
(Pre-Treatment)
Moral Fallibility
(Pre-Treatment)
Moral Superiority
(Pre-Treatment)
0.19**
(0.05)
0.10
(0.05)
0.82**
(0.02)
1,497
-0.04
(0.05)
-0.01
(0.05)
0.82**
(0.02)
Note: *p<0.05; **p<0.01. Regression coefficients (b) are on the original 1-7 scale.
Moderation analysis was not preregistered but explored. Specifically, I examined if the effect of the low
and high-dose treatments on moral humility was moderated by party identity or education to assess if the treatments
work similarly across people of different political leanings and educational levels. To test this, a party identity
dummy variable (Democrat/Democrat leaning = 0, Republican/Republican leaning = 1) and education dummy
variable (less than bachelor’s education = 0, bachelor’s education and higher = 1) was created and an interaction
variable between these and the experimental condition dummy variable was added in the regression. There weren’t
any significant interactions, suggesting that the treatments worked similarly across political leanings and education
levels.
Taken together, the results suggested the high dose moral humility treatment increased moral humility
compared to the control, the low-dose treatment did not significantly increase moral humility compared to the
control, and the high-dose treatment was significantly stronger than the low-dose treatment. Additionally, both
treatments worked similarly for Democrats and Republicans and people of different education levels.
93
Political Outcomes
To test the interventions’ effect on polarization, two political outcomes, political sectarianism and selective
exposure were examined. A similar test was conducted as moral humility, but on the political outcomes. That is,
linear regression was used to test if the two political variables significantly differed between both treatment
conditions and control, as well as if the high-dose condition had a bigger impact on the outcomes than the low-dose
condition. For political sectarianism analysis, pre-treatment level of political sectarianism was included as a
covariate; for selective exposure analysis, pre-treatment levels of selective exposure was included as a covariate.
The correlation between pre- and post- political sectarianism was r = 0.94, p <.001, and the correlation between pre-
and post- selective exposure was r = 0.93, p <.001, suggesting a high power to detect small effects. The prediction
was that compared to the control condition, people in the low and high-dose treatments will express lower political
sectarianism and more interest in cross-cutting exposure, and that the effects will be stronger in the high-dose
condition versus the low-dose condition.
Political Sectarianism. Both the low-dose and high-dose interventions did not have a significant impact on
political sectarianism (Model 1, Table 31). The high-dose condition was not significantly different from the low-
dose condition (Model 2, Table 31).
Table 31
Estimates (and standard errors) of the difference in political sectarianism between each of two treatments (low and
high dose) and control (Model 1), and between the two treatments (low vs high dose) (Model 2)
Political Sectarianism
Model 1 b
(Reference = Control)
Effect Size
Cohen’s d
High Dose
Low Dose
Control
Political Sectarianism
(Pre-Treatment)
-0.06
(0.04)
-0.02
(0.04)
0.99**
(0.01)
-0.09
-0.03
-
Model 2 b
(Reference = Low
Dose)
-0.04
(0.04)
Effect Size
Cohen’s d
-0.07
0.02
(0.04)
0.99**
(0.01)
0.03
-
Observations
Note: *p<0.05; **p<0.01. Regression coefficients (b) are on the original 1-7 scale.
1,496
The impact on the subscales was also not significant. Looking at just magnitude of effects, the results
showed that different aspects of political sectarianism (othering, aversion, moralization) contribute to the small
94
effect of the low-dose and high-dose treatments to different extent (Table 32). The treatments impacted the othering
aspect of political sectarianism most, followed by moralization, and then aversion.
Like moral humility before, I conducted moderation analysis (not preregistered, was exploratory) to
examine if the effect of the low and high-dose treatments was moderated by party identity or education. There
weren’t any significant interactions of both with the two treatments, suggesting that the treatments worked (or didn’t
work) similarly in reducing political sectarianism across political leanings and education levels.
Table 32
Estimates (and standard errors) of the difference in political sectarianism subscales between each of two treatments
(low and high dose) and control
Political Sectarianism
Othering b
Political Sectarianism
Aversion b
Political Sectarianism
Moralization b
-0.06
(0.05)
-0.03
(0.05)
0.95**
(0.01)
High Dose
Low Dose
Political Sectarianism
Othering
(Pre-Treatment)
Political Sectarianism
Aversion
(Pre-Treatment)
Political Sectarianism
Moralization
(Pre-Treatment)
Observations
-0.04
(0.04)
0.01
(0.04)
0.98**
(0.011)
1,496
-0.06
(0.05)
-0.03
(0.05)
0.96**
(0.010)
Note: *p<0.05; **p<0.01. Regression coefficients (b) are on the original 1-7 scale.
Political Sectarianism Mediation. The results of the mediation analysis are in Table 33 (left side of the
table), wherein I tested the extent to which moral humility explained the effects of the two treatments on political
sectarianism. The ACME (Average Causal Mediation Effect) tells us how much of the effect of the interventions on
political sectarianism is mediated by moral humility. Thus, it is the indirect effect of the interventions on political
sectarianism through moral humility. If ACME is significant (i.e., confidence interval doesn’t contain zero), it
suggests that moral humility as the mediator plays a significant role in explaining the relationship between the
treatments and political sectarianism.
The expectation was that the ACME’s will be negative and significant, indicating that the treatments
increase moral humility and an increase in moral humility leads to a decrease in political sectarianism. Result
95
showed that moral humility significantly mediated (ACME) the effect of the high dose interventions on political
sectarianism (marginally significant mediation for low dose treatment at p = 0.06). The direct and total effects of the
interventions on political sectarianism were not significant.
Table 33
Mediation results for the effect of the low and high dose treatments on political sectarianism and selective exposure
via moral humility
ACME
ADE
Total Effect
Prop Mediated
Note: †p<0.1, *p<0.05; **p<0.01
Political Sectarianism
Selective Exposure
Low Dose
High Dose
Low Dose
High Dose
-0.01
<.01
-0.01
0.79
-0.02**
-0.03
-0.05
0.35
0.22
1.25
1.47
0.15
0.74*
1.90
2.64
0.28
I also conducted sensitivity analysis to address the sequential ignorability assumption (Imai et al., 2010;
Tingley et al., 2014). The results of this sensitivity analysis showed that across both mediation models, a correlation
of r ~ 0.1 due to unmeasured confounders could nullify the mediation effect.
Taken together, the results suggested that both the low-dose and high-dose treatments did not significantly
reduce political sectarianism, both treatments were not significantly different from each other, and the effects were
not moderated by party identity or education levels. Mediation models found some evidence consistent with the
hypothesized causal prediction in the high-dose condition, i.e., treatment increases moral humility which
consequently decreases political sectarianism. However, sensitivity analysis suggested that the mediation models in
general were very sensitive to the existence of confounders.
Why might the mediation effect be significant when the direct and total effects are not? Some authors
suggest (Rucker et al., 2011) that the power of the mediation analysis is stronger than the analysis for direct and total
effects, making it possible to detect very small effects. Thus, it is possible that there was an effect of the treatments,
but given its small size, the tests for direct and total effects (Tables 26 and 27) were nonsignificant; the stronger
power of the indirect effects allowed the mediation pathway to reach significance despite the overall weak impact of
the treatments. However, give the general patterns of small effects and non-significance, it is hard to tell whether
these results are meaningful.
96
Selective Exposure. Both the low-dose and high-dose interventions did not have a significant impact on
willingness to engaging with opposing political viewpoint on participant’s self-selected most important issue (Model
1, Table 34). The high-dose condition was not significantly different that the low-dose.
Table 34
Estimates (and standard errors) of the difference in selective exposure between each of two treatments (low and
high dose) and control
High Dose
Low Dose
Control
Selective Exposure
Model 1 b
(Reference = Control)
Effect Size
Cohen’s d
2.74
(1.78)
1.71
(1.78)
0.11
0.07
Model 2 b
(Reference = Low
Dose)
1.03
(1.78)
Effect Size
Cohen’s d
0.04
Selective Exposure
(Pre-Treatment)
Observations
Note: *p<0.05; **p<0.01. Regression coefficients (b) are on the original -100 - +100 scale.
0.92**
(0.01)
1,228
-
-1.71
(1.77)
0.92**
(0.01)
-0.07
-
Like before, I conducted moderation analysis (not preregistered, was exploratory) to examine if the effect
of the low- and high-dose treatments was moderated by party identity or education. There weren’t any significant
interactions of both with the two treatments, suggesting that the treatments worked (or didn’t work) similarly in
increasing interest in opposing viewpoint across political leanings and education levels.
Selective Exposure Mediation. The results of the mediation analysis are in Table 33 (right side of table),
wherein the extent to which moral humility explained the effects of the interventions on selective exposure was
examined. The expectation was that the ACME’s will be positive and significant, indicating that the treatments
increase moral humility and an increase in moral humility leads to increase in interest in opposing political
viewpoint. Result showed that moral humility significantly mediated (ACME) the effect of the high-dose, but not
the low-dose intervention on selective exposure. The direct and total effects of the treatments on selective exposure
were not significant. The results of the sensitivity analysis showed that across both mediation models, a correlation
of r ~ 0.1 due to unmeasured confounders could nullify the mediation effect.
Taken together, the results suggested that both the low-dose and high-dose treatments did not significantly
decrease selective exposure, both the low-dose and high-dose treatments were not significantly different, and the
97
effects were not moderated by party identity or education levels. Mediation models found some evidence consistent
with the hypothesized causal prediction, i.e., treatments increased moral humility which consequently increased
cross-cutting exposure in the high-dose condition. However, sensitivity analysis suggests that the mediation models
in general were very sensitive to the existence of confounders. Like political sectarianism, it is possible that there
was an effect of the treatments but given its pretty small size, the tests for direct and total effects (Tables 28 and 27)
were nonsignificant; the stronger power of the indirect effects allowed the mediation pathway to reach significance
despite the overall weak impact of the treatments. However, give the general patterns of small effects and non-
significance, again it is hard to tell whether these results are meaningful.
Replication of Experiment 1: Comparing each of the five treatments comprising the low-dose condition to control
So far, when comparing the low-dose condition to the control, all five treatments were analyzed together as
one, as the idea was to see if the results found in Experiment 1 generalize over the different moral humility “stimuli”
or “topics” (Clifford & Rainey, 2024). This decision was taken in the light of the five treatments behaving similarly
in Experiment 1.
However, since participants randomly received one of five moral humility treatments individually in this
low-dose condition, it allowed the comparison of each of these five treatments with the control (and with each
other). Thus, we could assess if the five treatments were behaving similarly in Experiment 2 as well. Since this
analysis was similar to the analysis in Experiment 1, it allowed the replication of Study 1. Because the pre-and post-
treatment measures (for all 3 outcomes: moral humility, political sectarianism, and selective exposure) were highly
correlated (r ~ 0.9), there was good power (>80% to detect d = 0.15) even with just approximately 100 participants
per the five treatments (biased worldview, difficulty ethics, evil capacity, imperfect past, and moral disagreement)
and ~500 participants in control. Thus, each of the five treatments was compared to the control for all three
outcomes, moral humility, political sectarianism, and selective exposure. Results are in Table 35.
For moral humility, compared to the control, two treatments worked significantly to increase moral
humility and had comparable effect sizes like Experiment 1. These were the biased worldviews and moral
disagreement treatments. Thus, these two worked consistently across Experiment 1 and 2. The other three treatments
had either negligible and non-significant effect (difficulty ethics, past treatments), or opposite of the expected effect
(evil capacity, although the effect was not significant). Thus, these latter three didn’t replicate across the two
experiments.
98
Table 35
Estimates (and standard errors) of the difference in moral humility, political sectarianism, and selective exposure
between each of five individual treatments comprising the low dose condition and control
Moral
Humility
b
0.18**
(0.05)
0.02
(0.05)
-0.06
(0.05)
0.001
(0.05)
0.11*
(0.05)
0.94**
(0.02)
Moral
Humility
Cohen’s d
0.36
0.04
-0.12
0.001
0.23
-
Political
Sectarianism
b
-0.08
(0.06)
-0.001
(0.06)
0.02
(0.06)
0.15**
(0.06)
-0.15*
(0.06)
Political
Sectarianism
Cohen’s d
-0.15
-0.003
0.03
0.26
-0.27
Selective
Exposure
b
2.42
(2.96)
1.98
(3.01)
3.94
(3.17)
-6.14*
(3.15)
5.65*
(2.97)
Selective
Exposure
Cohen’s d
0.10
0.08
0.16
-0.24
0.22
0.99**
(0.01)
-
0.92**
(0.01)
-
Biased
Worldviews
Difficulty
Ethics
Evil
Capacity
Imperfect
Past
Moral
Disagreements
Moral Humility
(Pre-Treatment)
Political
Sectarianism
(Pre-Treatment)
Selective
Exposure
(Pre-Treatment)
Observations
Note: *p<0.05; **p<0.01
1,025
997
823
For political sectarianism, compared to the control, again the same two treatments replicated in terms of
significance or comparable effect sizes in decreasing political sectarianism. They were the moral disagreement and
biased worldview treatments. The other three treatments had either negligible and non-significant effect (difficulty
ethics) or opposite of the expected effect (imperfect past and evil capacity, former was stronger and significant).
Thus, these latter three didn’t replicate across the two experiments. So far across the outcomes, looking at direction
of coefficients, two treatments, evil capacity and imperfect past seemed to be working in opposite of expected
direction. Evil capacity seemed to work in opposite of intended direction for moral humility and political
sectarianism, and imperfect past in opposite direction for political sectarianism.
For selective exposure, compared to the control, one treatment worked in opposite of expected direction to
significantly decrease interest in exposure to opposing political viewpoint. This was the imperfect past treatment
which also worked in opposite of the expected direction for political sectarianism. The other four broadly replicated
in terms of significance or comparable effect sizes in increasing interest in exposure to opposing political viewpoint.
99
However, some were stronger and significant than others (i.e., moral disagreement), although the four were not
significantly different from each other.
In sum, across Experiment 1 and 2 taken together, the overall picture from the three outcomes suggested
that some treatments worked more consistently, strongly, and as hypothesized (e.g., biased worldviews, moral
disagreements), some worked inconsistently and at times produced effects in the opposite direction than
hypothesized (e.g., evil capacity and imperfect past), and some worked in the direction as expected but had very
weak effects (e.g., difficulty ethics).
Exploratory Analyses
Exploratory Analysis 1: Zooming in on the treatments in the high-dose condition. Some of the five
treatments in the low-dose condition, i.e., imperfect past and evil capacity, worked in the opposite of the expected
direction on the outcomes when comparing each of them individually to the control. Given this, I next did
exploratory analyses (not preregistered) examining if the presence of these two treatments in the high-dose
condition, wherein they were randomly presented with one other treatment had counteractive effects. That is,
whether the presence of these two (imperfect past and evil capacity) treatments led to a diminished effect or an
effect in the opposite direction to the expected effect on the three outcomes. I compared each of the ten possible
treatment combinations to the control condition (five treatments yield ten possible combinations). Since these were
exploratory analyses with reduced power, I focused on the sign and magnitude of the effects and not the
significance.
Across the ten possible combinations of the five treatments, I did find evidence suggestive of the
counteractive effects of these two treatments (Table 36). For example, the combination of these two treatments
produced the smallest positive effect on moral humility, or whenever there was increased political sectarianism
(positive sign) or decreased willingness towards exposure to opposing viewpoints (negative sign), the treatment
combination included the presence of imperfect past treatment.
Thus, when considered alongside the previous analysis of the five treatments individually in the low-dose
condition wherein evil capacity and imperfect past were behaving oddly, the findings suggest that the imperfect past
and evil capacity treatments tend to be backfiring at times and possibly be having counteractive effects. Thus, it is
possible that their presence maybe obscuring or interfering with the effects of other treatments when analyzed
together in both the low-dose and high-dose conditions.
100
Table 36
Estimates (and standard errors) of the difference in moral humility, political sectarianism, and selective exposure
between each of the ten possible combinations of the five treatments in the high dose condition compared to control
Contrast
(the treatment combination vs control)
Moral Humility
b
Political
Sectarianism b
difficulty ethics & imperfect past
difficulty ethics & biased worldview
difficulty ethics & evil capacity
difficulty ethics & moral disagreements
imperfect past & biased worldview
imperfect past & evil capacity
imperfect past & moral disagreements
biased worldview & evil capacity
biased worldview & moral disagreements
evil capacity & moral disagreements
0.13
(0.07)
0.07
(0.07)
0.16
(0.07)
0.10
(0.07)
0.20
(0.07)
0.03
(0.07)
0.06
(0.07)
0.15
(0.07)
0.14
(0.07)
0.14
(0.07)
0.07
(0.08)
-0.13
(0.09)
-0.14
(0.09)
-0.06
(0.09)
0.08
(0.09)
-0.11
(0.09)
0.03
(0.09)
-0.06
(0.09)
-0.13
(0.09)
-0.10
(0.09)
Selective
Exposure
b
8.44
(3.87)
9.73
(3.83)
0.44
(3.91)
4.73
(4.26)
-3.69
(3.79)
-5.02
(4.21)
4.65
(3.87)
3.96
(4.00)
3.14
(3.87)
0.39
(3.83)
Exploratory Analysis 2: Reanalyzing the main models without the two counteractive treatments.
Next, I reanalyzed the main models for Experiment 2, excluding these two treatments from both the low-dose and
high-dose conditions. Specifically, I examined whether the low-dose and high-dose conditions increase moral
humility, decrease political sectarianism, and increase interest in cross-cutting exposure compared to the control
condition, and whether high-dose conditions have stronger effects than low-dose conditions on the three outcomes.
Results are in Table 37.
Both low- and high-dose conditions significantly increased moral humility compared to the control. The
high-dose condition significantly decreased political sectarianism and significantly increased selective exposure
compared to the control. The low-dose condition marginally decreased political sectarianism (p = 0.05) and
increased selective exposure (p = 0.07) compared to the control. Comparing the low- and high-dose condition to
each other, the three outcomes were not significantly different across the two conditions—meaning the high-dose
treatment didn’t have a meaningfully stronger effect on outcomes than the low-dose treatment.
101
Table 37
Estimates (and standard errors) of the difference in moral humility, political sectarianism, and selective exposure
between each of two treatments (low and high dose) and control, excluding two treatments (evil capacity and
imperfect past)
Moral
Humility
Effect Size
Cohen’s d
Political
Sectarianism
Effect Size
Cohen’s d
Selective
Exposure
Effect Size
Cohen’s d
High Dose
Low Dose
Moral Humility
(Pre-Treatment)
Political Sectarianism
(Pre-Treatment)
Selective Exposure
(Pre-Treatment)
0.10*
(0.04)
0.10**
(0.03)
0.95**
(0.02)
0.21
0.22
-
-0.11*
(0.05)
-0.08†
(0.04)
-0.20
-0.15
6.01**
(2.48)
3.36†
(1.88)
0.99**
(0.01)
-
0.25
0.15
-
0.93**
(0.01)
791
Observations
Note: †p<0.1, *p<0.05; **p<0.01
938
941
Experiment 2 Summary
This study tested a low-dose and high-dose moral humility intervention against a control and each other.
Specifically, their impact on moral humility, political sectarianism, and selective exposure was examined. In the
low-dose intervention, participants received one of five treatments randomly, and in the high-dose intervention they
received two of five treatments randomly. Only the high-dose intervention significantly increased moral humility.
Neither intervention significantly reduced political sectarianism or increased willingness to engage with opposing
political viewpoints. Further, there were no significant difference between the two interventions on these outcomes
(except for moral humility where the high-dose had stronger impact than the low-dose). Mediation analyses
indicated that the interventions influenced political sectarianism and selective exposure indirectly through moral
humility in the high-dose condition. Taken together, given that these indirect effects were small, the direct and total
effects were nonsignificant, and in general most models were non-significant, the overall impact of the interventions
was uncertain.
A closer examination of the five individual treatments that comprised the low-dose and high-dose
conditions – difficulty ethics, biased worldviews, imperfect past, evil capacity, and moral disagreements – revealed
102
that there was a notable variation in their effect. It appeared that the five treatments were not working similarly,
unlike Experiment 1. The most consistent and overall strongest and significant treatments across outcomes and
experiments were the biased worldviews and moral disagreements treatments. The imperfect past and evil capacity
treatments were sometimes working in the opposite direction than expected in Experiment 2 and thus were
inconsistent across the two experiments. The difficulty ethics treatment seemed to have very negligible effects at
times, although overall behaved in the predicted direction across outcomes and experiments.
Taken together, this means that combining the five treatments or treating them as largely similar or
interchangeable isn’t the most effective approach at this stage as their individual impacts vary significantly. Doing
this for the main analysis in this study might have obscured effects when they existed and/or distorted the effects due
to the inter-stimuli/treatment variation. To address this concern to some extent, I did exploratory analysis, re-
estimating the main models with the two unreliable treatments excluded. This revealed that the low-dose and high-
dose conditions did significantly (sometimes marginal) increase moral humility, decrease political sectarianism, and
increase cross cutting exposure. However, the high-dose condition was still not significantly stronger than the low-
dose condition; they both had very similar effect sizes. Thus, combining treatments didn’t have a substantial
compounding effect in increasing the strength of the interventions that I had expected. In a similar vein, overall, the
effect sizes for moral humility and political outcomes were again in the small-to-moderate range like Experiment 1.
In summary, these findings suggested that while the treatments had some impact, their effectiveness was
modest and varied depending on the specific treatments used. Of the five treatments used in both experiments, why
might have some treatments worked more consistently (like biased worldviews or moral disagreements) than others
which sometimes even backfired (like evil capacity and imperfect past)? One possibility is that some treatments like
the latter two are more sensitive to context. Notably, the political context changed between the two experiments—
the change in government between the two experiments created a more politically charged atmosphere. At a time
when political identities were highly charged and the political climate intensely moralized, interventions that
highlighted the capacity of people to do evil like erstwhile Nazis, or compared the moral failing of present
generations to those of their ancestors might have triggered or threatened participants. This could have occurred if
these elements in the treatments were perceived as an attack on their ingroup’s identity or worldview or as validating
moral disdain towards outgroups or vindicating their own ingroup’s worldview. It is also possible that since these
two treatments had the possibility of evoking negative emotions of guilt and shame given that they both talked about
103
wrongdoings of past groups and ancestors, they are more susceptible to backfiring if these emotions trigger moral
identity threat. Future work would thus need to investigate these possibilities and refine these treatments to
minimize unintended backfire effects.
104
Chapter IV: General Discussion, Future Directions, and Conclusion
In this project, across ten studies, I investigated the psychological construct of moral humility, filling
existing gaps in its empirical study. I conceptualized and developed a psychometrically validated measure, examined
its nomological associations, and analyzed its antidotal role in a moralized and conflictual context, i.e., political
polarization, using correlational and experimental studies. Notably, to this latter end, I developed and validated
interventions of moral humility and assessed their causal impact on polarization outcomes. This also enabled the
examination of a more fundamental question—whether moral humility an individual characteristic can be changed
or is relatively stable.
I constructed a thirty-item moral humility scale comprised of three factors — moral fallibility, moral
openness/learning, and moral superiority. An examination into its nomological associations, especially to constructs
such as personality, religiosity, ideology, political extremity, moral grandstanding, moral relativism, and intellectual
humility, showed that moral humility was associated with these constructs in meaningful ways. A person high in
moral humility showed less extremism and absolutism (e.g., less political extremity, religious exclusiveness, moral
grandstanding, moral absolutism, religiosity), more openness and flexibility (e.g., openness to experience,
intellectual humility, need for cognition), more humility (e.g., modesty, intellectual humility), and a more positive
orientation towards others (e.g., more agreeableness, lower psychopathy, lower narcissism, etc.).
Having constructed a scale of moral humility and assessed that it was working sensibly with other
constructs, I then assessed its ability to attenuate the dark features of morality in a moralized and high-conflict
context. In the context of political polarization in the US, moral humility predicted lower antagonism and derogation
towards outgroup, lower rigidity in one’s own views, lower rejection of compromise and contact, and lower
adoption of morally questionable means. Consistently, across over fifteen outcomes and multiple studies in diverse
samples, moral humility was associated with lower levels of polarization. These relationships provided strong
evidence of moral humility being a counteractive force in high-conflict, intergroup contexts that bring out the dark
side of our morality. Further, the studies found that these effects of moral humility emerged over and above effects
of other important, conceptually and empirically related constructs such as intellectual humility and moral
relativism. However, these results were correlational in nature and could not definitely establish the causal role of
moral humility.
105
Five moral humility interventions were then designed and tested in two experiments. These were aimed to
find ways to increase people’s moral humility. They thus served to provide insight into the amenability of moral
humility (i.e., if it is a fixed or changeable aspect of people), which would help shed light on whether it is a
construct that can be fruitfully targeted through interventions. They also helped examine the causal role played by
moral humility in a moralized, conflictual contexts like political polarization. Across the two experiments, two
interventions, biased worldviews and moral disagreements, emerged as the most consistent treatments that increased
moral humility, and subsequently decreased political antagonism towards outgroup and increased openness towards
opposing sides’ viewpoints. A third treatment, difficulty ethics also worked in the predicted direction across
outcomes but sometimes had negligible effects. Finally, two interventions, evil capacity and imperfect past,
produced inconsistent effects across experiments. These variations suggest that the approach taken in the second
experiment, where interventions were treated as interchangeable or combined, should be avoided until their
individual consistency is better established. Nonetheless, the three more reliable interventions provided promising
evidence that moral humility can be increased (although with modest effect sizes), which can then have ameliorative
effects in morally inflamed contexts such as polarization. These findings also highlight the kind of ideas (e.g., biased
worldviews, moral disagreements, difficulty ethics) that show promise for boosting moral humility (either
individually and/or in combination) and counteracting the rigid and antagonistic tendencies that arise in moralized
conflicts. Further, by showing the amenability of moral humility, they supported the value in targeted interventions
to increase moral humility. The modest effectiveness of these interventions however does raise the need for further
refinement of these intervention strategies to enhance their robustness and impact.
Theoretical and Practical Implications
Moral humility as a distinct and meaningful psychological construct
Morality is a fundamental axis of human experience, shaping how our individual selves and social lives are
understood and navigated. Notably, it has a dark side that fuels hate, intolerance, rigidity, cruelty, violence, and
conflict. Against this backdrop of morality’s centrality to our lives and its dark nature, the main thesis of this project
was to propose and show that a domain-specific form of humility—moral humility, or humility in the domain of
morality—is a conceptually and empirically meaningful psychological construct for understanding our morally
infused lives. In particular, this project showed that moral humility can serve as an antidote to the dark features of
our morality, for instance, as they manifest in a polarized, moralized, and conflictual intergroup (political) context.
106
Moral humility as a domain-specific form of humility has not received much conceptual or empirical examination.
This work puts moral humility on the map, as a distinct, theoretical, and meaningful psychological construct, worth
the investigation, especially when it comes to understanding morally relevant contexts and psychological tendencies.
Relatedly, using correlational or/and experimental evidence, this project demonstrated moral humility’s
uniqueness from important constructs such as general humility (in HEXACO), intellectual humility, moral
conviction, and moral relativism. This is important, as it suggests that a distinct form of humility—moral humility—
plays a unique role in navigating morally charged situations, offering a psychological mechanism that is separate
from general or intellectual humility, or moral conviction or relativism. This distinction opens the door for further
research into how moral humility influences key outcomes, especially those that are morally relevant. In particular,
when it comes to countering the dark aspects of our morality, there are few works so far in the moral psychological
literature, directly or indirectly providing solutions that put morality at its core (for exceptions see Kraaijeveld &
Jamrozik, 2022; Kodapanakkal et al., 2022; Wright & Pölzler, 2022; Bastian et al., 2015). This work fills this gap
and provides an additional promising solution to the repertoire of possible antidotes.
Notably, amongst the solutions suggested in existing literature, two are de-moralization or moral relativism
(Kraaijeveld & Jamrozik, 2022; Kodapanakkal et al., 2022; Wright & Pölzler, 2022; Bastian et al., 2015). However,
moral conviction or moralization has also been associated with positive outcomes such as voting or collective action
(Skitka & Bauman, 2008; Van Zomeren at al., 2012). It is thus unclear if reducing moral conviction would
compromise its positive effects, and hence whether doing so would be advisable. Moral relativism has been
considered philosophically and morally objectionable by many scholars due to its implications of “anything goes”
wherein harmful practices such as sexism can be justified as being right according to the norms of a sexist society
(Gowans, 2021). Given these competing considerations for both these proposed antidotes, moral humility offers the
potential of an underexplored psychological antidote that does not rely on demoralization or inducing moral
relativism to counteract our dark moral tendencies. It would instead encourage approaching moral divisions with an
acknowledgement of one’s own moral limitations, recognition of others’ moral strengths, and an openness to moral
learning. Of course, at this stage, there is a lack of definitive comprehensive data to make strong claims about moral
humility’s unequivocal positive effects without also investigating its secondary effects. Thus, any strong claims
await future tests. Nevertheless, the evidence in this project far points towards a promising solution to the dark
aspects of our morality.
107
Moral basis of political conflicts
When it comes to political polarization and political conflicts at large, this work offers additional insight
and support into its moral basis. Politics is not merely about practical concerns but also has moral and value
considerations at its core (Paul et al., 2010; Colombo, 2019). These considerations are often given less attention
when designing interventions as compared to political identity (e.g., Voelkel et al., 2024), despite evidence of its
close connections to people’s political attachments (Lupton et al., 2020). Focusing on political identity is
understandable; however, because political identities are largely stable overtime (more stable than personality traits;
Brandt & Morgan, 2022; Vaisey & Kiley, 2021), identifying routes to depolarization that do not invoke political
identities is fruitful. By showing moral humility’s relevance to political polarization, the work in this project a)
reinforces the moral nature of politics and political conflicts, and b) offers a morality-based approach to addressing
political conflicts. Relatedly, it shows using experimental studies that moral humility is amenable to change,
offering a way to addressing conflicts that is not focused on shifting stable psychological attributes like political
identity. Moreover, in the correlational studies on political polarization, moral humility’s effect size was on average
greater than that of intellectual humility and political extremity, two psychological constructs that have been
investigated in the context of polarization, and at least one that is considered central (i.e., political extremity). This
suggests that moral humility is likely of substantive importance. Taken together, this research on moral humility and
political polarization highlights moral humility as a promising and underexplored avenue for addressing political
polarization—one that acknowledges the moral underpinnings of political divisions.
Strengths, Limitations, and Future Directions
The work in this project had the following strengths. First, it developed the first psychometrically validated
measure of moral humility, allowing future studies to measure this construct and build upon it. Relatedly, in addition
to the first validated moral humility scale, it also offers a shorter scale of moral humility for use in large national and
expensive surveys (used in some studies here). Second, by providing the first comprehensive examination into moral
humility’s nomological network, it provides a solid groundwork on the nature and correlates of the construct which
future work can build on. Third, it is amongst the first to show the relevance of moral humility in a high-conflict,
polarizing, and moralized domain, i.e., politics. This highlights the importance of understanding this construct in the
context of political outcomes and contexts. Notably, the relationship between moral humility with polarization
emerged across multiple outcomes, ranging from hostility towards outgroup, perceptions of outgroup morality and
108
threat, willingness to politically compromise, support for anti-democratic methods, and interest in cross-cutting
exposure. Fourth, it is amongst the first to develop and validate interventions to increase moral humility, showing
that moral humility is responsive to targeted interventions and the kinds of ideas that might help in boosting moral
humility. This provides rationale and motivation for further developing interventions that increase moral humility
even more strongly and in lasting ways in future work by amping up the moral humility treatments (such as by
repeating intervention over time or embedding them in public platforms or conversations). Fifth, this work was
conducted across different types of samples (YouGov nationally representative samples, student samples, nationally
diverse online Prolific samples), with relationships replicating across the diverse samples. The use of both diverse
samples and outcome lends confidence to the reliability and generalizability of the findings. Sixth, this work
provided important evidence (through incremental and discriminant validity tests) to distinguish moral humility
from related constructs like intellectual humility and moral relativism, suggesting an alternative psychological
mechanism through which important and relevant outcomes can be targeted. Seventh, moral humility’s effect size in
the correlational studies being on an average comparable or greater than intellectual humility and political
extremity—two psychological constructs that have been investigated and considered important in the context of
polarization—showed that moral humility is actually of substantive importance. Finally, this work provided both
correlational and causal evidence towards supporting moral humility’s role in mitigating the dark features of our
morality, particularly in the context of political divisions. Overall, this work is amongst the first detailed empirical
investigations into moral humility, its nature, associations, and its impact.
Despite these strengths, there were limitations. First, the relevance of moral humility in countering the dark
features of morality was examined in just one morally inflamed conflict i.e., political polarization. Thus, we have
limited insight into its relevance in other moralized contexts. Future work should investigate the role of moral
humility in other contexts characterized by these dark moral features such as religious intolerance or historical
conflicts. Further, to fully understand the nature and consequences of moral humility, future studies also need to
examine how moral humility is linked with the not-so-dark aspects of morality. For example, how is moral humility
associated with taking action against moral wrongs or fighting against injustice? The current studies showed that
moral humility can counter the dark aspects of our morality such as animosity and rigidity, but can moral humility
also promote the constructive aspects of our morality? Or does moral humility dampen moral action, such as social
activism aimed at moral progress and social change? The answers to these questions, and consequently a fuller
109
understanding of moral humility, await further investigation. Notably, some previous work has found that observing
moral humility in others promotes prosocial and ethical action (Owens et al., 2019), suggesting that moral humility
may promote the good features of morality as well.
Second, the measures used across studies were largely self-report. While self-report measures have notable
strengths (Corneille & Gawronski, 2024), incorporating behavioral assessments can provide a richer understanding
of a psychological phenomenon. Future research on moral humility would benefit from such measures. For instance,
researchers could examine how moral humility influences resource allocation in economic games (e.g., ultimatum or
dictator games) for those who participants perceive to be on the opposing side in a moral situation and those on the
same. Another approach could involve observing people’s punishment or forgiveness behaviors in morally
ambiguous situations where the incentive for virtue signaling or moral grandstanding is high (e.g., Jordan & Kteily,
2022). Alternatively, willingness to sign up for or attend events like talks or discussions that challenges people’s
own moral ideas or worldviews could offer valuable behavioral insights into moral humility. Apart from using
behavioral measures to enhance the understanding of moral humility, self-report measures that capture state-
variations in moral humility across different tasks or moral contexts can enhance the understanding of moral
humility by providing insight into its situational dynamics and help triangulate on it using an alternate method
(Fleeson & Jayawickreme, 2015; Van Tongeren et al., 2023). The measure developed and used in this project
assessed moral humility as a trait or global characteristic of the person. By using state measures of moral humility in
future studies, researchers could better capture the variability in moral humility, i.e.., how moral humility fluctuates
within and across individual when they are placed in various moral situations.
Third, across the correlational and experimental studies, moral humility was observed to have small-to-
medium effects on the outcomes assessed (Funder & Ozer, 2019). This means that while moral humility appears to
be a meaningful psychological construct with real-world implications, its impact may be modest in magnitude.
Given this, future research can explore ways to enhance the effectiveness of moral humility interventions.
Interventions could be made more immersive or sustained over longer periods to see if stronger and more lasting
effects can be achieved. For example, repeated exposure to the moral humility-promoting narratives identified in the
experiments, or its integration into social conversations with peers or in messages by authority figures, such as
leaders or media influencers, could help reinforce its impact. Additionally, examining whether moral humility
interacts with other psychological traits (e.g., personality, social identity, character virtues, cognitive and emotional
110
styles) or contextual factors (e.g., elite signaling, social norms) could also help identify conditions under which its
influence is amplified.
Fourth, while the experimental studies did find that moral humility can be increased through certain
interventions and reduce political divisions, not all interventions that were tested as part of the studies worked
reliably. Two interventions, one emphasizing the capacity for evil in ordinary humans, and one highlighting the
moral imperfections and blindspots of our ancestors, produced inconsistent effects across the studies. In the second
study, they worked counter to expectations on moral humility or political outcomes. Why might this have happened?
One possibility is that these interventions are particularly sensitive to contextual factors. The first and second
experimental studies which tested these interventions were conducted at different times, with a major political shift
occurring between them—Donald Trump’s inauguration and subsequent executive orders substantially changed the
political landscape. It is possible that the heightened political tensions at the time of the second study made
participants react more defensively to these interventions. This might have especially happened with these two
interventions because perhaps they have some elements or ideas that have more susceptibility to backfire effects if
perceived as an attack on their ideological group or moral worldview. Future work would thus need to repeatedly
test these as well as other interventions across different contexts and time points to establish their robustness.
Further, before repeated testing across contexts, these interventions might need to be refined more through more
detailed discussions and/or testing with diverse population subgroups, so that any threatening aspects may be
identified, removed and/or reframed.
Fifth, while this work showed that moral humility can be increased through targeted interventions, the
durability of these effects remains an open question given that the experimental studies in this project did not
follow-up participants after the experiments. Do the observed increases in moral humility persist over time, or do
they fade as the moral humility narratives fade from salience? Longitudinal studies that track moral humility across
extended periods would help determine the stability of these changes in response to interventions.
Sixth, this project focused primarily on moral humility in the political-cultural context of the US. Future
research should explore moral humility in different cultural contexts within and across different societies. Given that
the conceptions of morality and/or humility may vary across cultures, moral humility may manifest differently in
different societies and cultures with different dominant moral frameworks or understanding of humility. For
example, consider two religious cultures, Christianity and Hinduism. In Christianity, moral fallibility of humans is
111
central to its religious and ethical teachings—moral humility in someone influenced by the Christian framework thus
may dominantly take the form of recognizing the inherent moral limitations of oneself and others. Alternatively,
Hinduism embraces moral particularism as an important part of its moral framework — thus moral humility in
someone influenced by the Hindu framework may dominantly take the form of recognizing that moral priorities and
strengths may be different across individuals embedded in different contexts. These differences in the source and
nature of moral humility might mean that interventions and measurements would need to be adapted to account for
these cross-cultural variations. It would also mean refining the theory to account for variations in the manifestations
of moral humility.
Finally, insight into the ontogeny of moral humility is lacking and would be useful in understanding the
antecedents of moral humility. Future work can investigate how virtues like moral humility develop. Are they
primarily shaped during childhood, alongside the development of other core aspects of our morality (e.g., empathy,
fairness, or prosocial behavior), or do they emerge more prominently in adulthood, when we become better at
reflection and recognizing complexity? Are they influenced by certain religious, cultural, or political values, or by
specific life experiences such as exposure to diverse people and perspectives? Insights into these questions await
future research and would enrich our understanding of how the virtue of moral humility can be cultivated.
Digging deeper into unexpected findings
Throughout this project, there were some unexpected findings that highlighted gaps in the theoretical and
practical understanding of moral humility and would benefit from future enquiry. First, while probing moral
humility’s associations with other psychological constructs and political outcomes, I discovered an unexpected small
positive association of moral fallibility with psychopathy and support for anti-democratic means, and small negative
associations with sincerity and fairness (facets of honesty-humility). All these constructs broadly captured an
inclination towards morally questionable things. Taken together, these associations indicate that acknowledging
one’s moral limitations or fallibility is not unequivocally a sign that one is motivated towards moral improvement,
prosociality, or generally towards thinking or acting morally. Instead, these associations suggest that recognizing
one’s moral imperfections could, for some individuals, be compatible with an immoral, amoral, or morally complex
orientation. For example, imagine an ambitious person (perhaps a politician, businessman, or scientist) who wants to
make a tangible impact on society. This person might think that sometimes doing things that are immoral (e.g.,
exploiting workers, flouting ethical guidelines) are a necessary cost that one must pay to achieve great things. Such a
112
person might be willing to acknowledge that they indeed do morally questionable things which makes them morally
imperfect, but they are not inclined to be any different as they are motivated by other ends. As another example,
imagine a person who recognizes that eating meat is morally problematic and doing so makes one morally
compromised as a human being—yet they are not inclined to act differently because being a moral person or being
moral in the domain of food choice is not very important to their identity or self-concept. Or consider another
person who views moral fallibility as inevitable and, based on that perspective, makes little effort to change or grow
morally. They may acknowledge moral shortcomings as an unchangeable aspect of human nature and adopt a
passive or resigned stance toward their own moral failings, or perhaps even use it as a license to be morally lax. All
these examples highlight how an acknowledgment of moral fallibility or limitations is not necessarily an indication
that it will promote a more moral, ethical, or prosocial approach in someone.
This raises important questions about the boundary conditions of the implications of moral fallibility, and
perhaps moral humility generally. Does moral fallibility operate differently in the presence of other psychological
factors? For example, does the acknowledgement of one’s moral fallibility produce different behaviors and
inclinations when it occurs in the presence of a stronger vs weaker moral identity (i.e., how important being moral is
to one’s self concept) or higher versus lower levels of moral openness/learning (e.g., another facet of moral
humility). Does high moral fallibility look different in a person high in psychopathy versus someone high in
empathy? These suggest the need for future work to investigate moral fallibility (and by extension, moral humility)
in an integrative fashion to get a nuanced understanding of the nature and implications of moral humility, rather than
assuming moral fallibility to be an unequivocally and uniformly a positive force. Broadly, these insights apply not to
just moral humility but to the study of humility in general (including other types of humility in literature). An
acceptance of one’s fallibility is a core part of the conceptualization of humility and is often assumed to produce
good interpersonal and intergroup outcomes. The theorizing needs to account for cases where the acceptance of
fallibility occurs in people who don’t care about being wrong or flawed or consider fallibility inevitable, as these
kinds of cases might mean differential predictions of fallibility’s consequences based on maybe the presence of
other psychological characteristics. Moral fallibility is then perhaps a necessary but not sufficient condition for
constructive interpersonal and intergroup outcomes.
The second set of unexpected findings emerged when probing the relationship between moral humility and
polarization outcomes, specifically its lack of association with myside sharing in a correlational study. I had
113
anticipated that higher moral humility would be associated with lower sharing of moralized and ideologically
congruent news items on social media. Although the results tended in that direction (marginally significant, weak
effects), moral humility’s role remains unclear. Given that myside sharing was one of the few outcomes across the
many studies and outcomes in this project that was more behavioral in nature, these results highlight the important
need to understand moral humility’s impact on behavior. The studies in this project did provide strong evidence of
moral humility’s relationship with more cognitive (i.e., learning) and emotional (e.g., antipathy) outcomes, but
information on its relationship with behavioral outcomes is lacking. The relationship between self-reports and
behavior is famously weak (Dang et al., 2020). One reason for this is that behavior is multi-determined whereby
other determinants of the behavior can weaken the relationship between certain thoughts/feelings/attitudes and
behavior. In the case of myside sharing, the public or social context of sharing news items on social media might
bring other strong contextual considerations in mind (e.g., social condemnation or reward) that trump or weaken the
effects of one’s moral humility. This possibility highlights that the effects of moral humility on behavior might be
context dependent. Thus, it is essential in future work to systematically examine what kinds of contexts can hinder
or facilitate the impacts of moral humility. For example, is moral humility less influential in environments where the
norms favor moral outrage (e.g., social media), or more influential in contexts where norms favor seeing the good in
others (e.g., community or interpersonal dialogues).
In a similar vein, another possible reason for the weak links with behavior is that self-report scales
(including mine) are typically designed to be context-insensitive, i.e., they capture the participant’s understanding of
themselves across contexts (e.g., how morally humble they are across moral situations). Said differently, they are
designed to capture traits rather than states. Thus, these trait-oriented scales are sometime poor at predicting
behavior in a particular context. One way to address this is to adapt the scales to be more context-sensitive so that
they can capture moral-humility states (e.g., how morally humble people think they are when it comes to interacting
with people on social media). Thus, this raises the need for self-report scales that can capture states as well, which
might then help with predicting morally relevant behavior in particular contexts with greater fidelity. However, it is
possible that the lack of relationship with myside sharing was not because of the scale’s context-insensitivity but
because the current scale captures people’s cognitive orientations more than it captures their behavioral orientations.
Although during scale development, I aimed to have items that captured behavior (e.g., moral choices, decisions) as
well as cognitive orientations (e.g., moral views), in the process of refining the scale, the scale ended up being more
114
dominated by items capturing the latter. Therefore, even a trait-focused moral humility scale that includes more
balanced items measuring both moral behaviors and moral views could be more effective at predicting behavioral
outcomes. This would involve refining the scale in the future to more comprehensively capture the full scope of
moral humility content.
Finally, the third set of unexpected findings was the inconsistency of certain moral humility treatments or
interventions across the two experiments. Some moral humility treatments, contrary to my expectations, were
weaker in their effects when tested in a different experiment later or even backfired. As mentioned previously, one
possibility is that the increasingly charged political landscape during Experiment 2 caused participants to react more
defensively or rigidly to treatments that could be perceived as either an attack on or a validation of a particular
political group. Regardless of the exact reason, these findings highlight a crucial gap in the present research—i.e., in
order improve the reliability and effectiveness of moral humility interventions, it is important to gain a deeper
understanding of the psychological mechanisms driving moral humility and its change. While the intervention
studies in this project were primarily outcome-oriented, i.e., focusing on ways to change moral humility, the next
phase of research can complement the understanding of moral humility by taking a more process-oriented approach
that involves understanding the psychological processes through which these changes occur. For example,
understanding how cognitive processes like cognitive reappraisal of one’s moral limitations or perspective taking of
other’s moral strengths, or emotional responses like empathy, guilt, or shame, mediate the effects of interventions
could provide insight into why some approaches work better or worse than others. Additionally, exploring how
identity dynamics, such as moral or ideological identity/worldview threat or salience influences responses to moral
humility interventions may help reveal why certain individuals or groups react negatively or fail to benefit from
interventions. By identifying these underlying psychological processes, future research can identify and design
interventions that are reliable and effective across different conditions (e.g., individuals, groups, and contexts).
115
Conclusion
While the “better angels of our nature” may guide us toward moral progress, there is a dark side to our
morality as well that has long fueled violence, conflicts, hate and, intolerance. This project proposed moral humility
as a possible antidote to these destructive tendencies of our morality. It offered the first comprehensive empirical
examination into the psychological nature and significance of moral humility. Through both correlational and
experimental studies, the studies showed moral humility’s role in mitigating the rigidity and hostility that often
characterize moralized political conflicts. This research also laid the groundwork for future inquiries into moral
humility—amongst others, its relationship with moral action and moral improvement, its developmental pathways,
strategies for cultivating it, and its manifestations across diverse cultural contexts.
116
REFERENCES
Ahler, D. J., & Sood, G. (2018). The parties in our heads: Misperceptions about party composition and their
consequences. The Journal of Politics, 80(3), 964-981.
Ashton, M. C., & Lee, K. (2009). The HEXACO–60: A short measure of the major dimensions of personality.
Journal of personality assessment, 91(4), 340-345.
Ballantyne, N. (2023). Recent work on intellectual humility: A philosopher’s perspective. The Journal of Positive
Psychology, 18(2), 200-220.
Bastian, B., Zhang, A., & Moffat, K. (2015). The interaction of economic rewards and moral convictions in
predicting attitudes toward resource use. PLoS ONE, 10(8), Article e0134863.
https://doi.org/10.1371/journal.pone.0134863
Baumeister, R. F. (1999). Evil: Inside human violence and cruelty. Macmillan.
Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological bulletin, 107(2), 238.
Bowes, S. M., Blanchard, M. C., Costello, T. H., Abramowitz, A. I., & Lilienfeld, S. O. (2020). Intellectual humility
and between-party animus: Implications for affective polarization in two community samples. Journal of
Research in Personality, 88, 103992.
Brandt, M. J., & Reyna, C. (2010). The role of prejudice and the need for closure in religious
fundamentalism. Personality and Social Psychology Bulletin, 36(5), 715-725.
Brandt, M. J., & Vallabha, S. (2022). Intraindividual Changes in Political Identity Strength (but not Direction) are
Associated with Political Animosity in the United States and the Netherlands. Personality and Social
Psychology Bulletin, 01461672231203471.
Brady, W. J., Wills, J. A., Jost, J. T., Tucker, J. A., & Van Bavel, J. J. (2017). Emotion shapes the diffusion of
moralized content in social networks. Proceedings of the National Academy of Sciences, 114(28), 7313-
7318.
Brambilla, M., Sacchi, S., Rusconi, P., & Goodwin, G. P. (2021). The primacy of morality in impression development:
Theory, research, and future directions. In Advances in experimental social psychology (Vol. 64, pp. 187-
262). Academic Press.
Broockman, D., & Kalla, J. (2016). Durably reducing transphobia: A field experiment on door-to-door canvassing.
Science, 352(6282), 220-224.
Burke, B. L., Martens, A., & Faucher, E. H. (2010). Two decades of terror management theory: A meta-analysis of
mortality salience research. Personality and Social Psychology Review, 14(2), 155-195
Cacioppo, J. T., & Petty, R. E. (1982). The need for cognition. Journal of personality and social psychology, 42(1),
116
Cassese, E. C. (2021). Partisan dehumanization in American politics. Political Behavior, 43(1), 29-50.
Clifford, S., & Rainey, C. (2024). Estimators for Topic-Sampling Designs. Political Analysis, 32(4), 431-444.
Clifford S. (2019). How emotional frames moralize and polarize political attitudes. Political Psychology, 40(1), 75–
91. https://doi.org/10.1111/pops.12507
117
Collier‐Spruel, L., Hawkins, A., Jayawickreme, E., Fleeson, W., & Furr, R. M. (2019). Relativism or tolerance?
Defining, assessing, connecting, and distinguishing two moral personality features with prominent roles in
modern societies. Journal of personality, 87(6), 1170-1188.
Comrey, A., & Lee, H. (1992). A first course in factor analysis. Hillsdale, NJ: Erlbaum.
Conrique, B. G. I. (2021). Different Values but Similar Backgrounds: How Relativism Influences Naïve Realism in
Everyday Disagreements (Doctoral dissertation, University of Pittsburgh).
Corneille, O., & Gawronski, B. (2024). Self-reports are better measurement instruments than implicit measures.
Nature Reviews Psychology, 1-12.
Curry, O. S. (2016). Morality as cooperation: A problem-centred approach. The evolution of morality, 27-51.
Curry, O. S., Mullins, D. A., & Whitehouse, H. (2019). Is it good to cooperate? Testing the theory of morality-as-
cooperation in 60 societies. Current anthropology, 60(1), 47-69.
Dang, J., King, K. M., & Inzlicht, M. (2020). Why are self-report and behavioral measures weakly correlated?.
Trends in cognitive sciences, 24(4), 267-269.
Davis, D. E., Rice, K., McElroy, S., DeBlaere, C., Choe, E., Van Tongeren, D. R., & Hook, J. N. (2016).
Distinguishing intellectual humility and general humility. The Journal of Positive Psychology, 11(3), 215-
224.
Darwin, C. (1872). The descent of man, and selection in relation to sex (Vol. 2). D. Appleton.
David, C (2023). The Chinese Room Argument, The Stanford Encyclopedia of Philosophy (Summer 2023 Edition),
Edward N. Zalta & Uri Nodelman (eds.), URL =
.
Delton, A. W., DeScioli, P., & Ryan, T. J. (2020). Moral obstinacy in political negotiations. Political Psychology,
41(1), 3-20.
Douglas, B. D., Ewell, P. J., & Brauer, M. (2023). Data quality in online human-subjects research: Comparisons
between MTurk, Prolific, CloudResearch, Qualtrics, and SONA. Plos one, 18(3), e0279720.
Drinkwater, K., Dagnall, N., Denovan, A., Parker, A., & Clough, P. (2018). Predictors and associates of problem–
reaction–solution: statistical bias, emotion-based reasoning, and belief in the paranormal. SAGE Open, 8(1),
2158244018762999.
Druckman, J. N., & Levy, J. (2022). Affective polarization in the American public. In Handbook on politics and
public opinion (pp. 257-270). Edward Elgar Publishing.
Effron, D. A., & Miller, D. T. (2012). How the moralization of issues grants social legitimacy to act on one’s
attitudes. Personality and Social Psychology Bulletin, 38(5), 690-701.
Enders, A. M., & Lupton, R. N. (2021). Value extremity contributes to affective polarization in the US. Political
Science Research and Methods, 9(4), 857-866.
Enders, A. M., & Armaly, M. T. (2019). The differential effects of actual and perceived polarization. Political
Behavior, 41, 815-839.
Faul, F., Earth fields, E., Lang, A.-G., & Buchner, A. (2007). G*Power 3: A flexible statistical power analysis
program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39, 175-191.
118
Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009). Statistical power analyses using G*Power 3.1: Tests for
correlation and regression analysis. Behavior Research Methods, 41, 1149-1160
Finkel, E. J., Bail, C. A., Cikara, M., Ditto, P. H., Iyengar, S., Klar, S., ... & Druckman, J. N. (2020). Political
sectarianism in America. Science, 370(6516), 533-536.
Finkel, E., Landry, A., Hoyle, R. H., Druckman, J., & Van Bavel, J. J. (2024). Partisan Antipathy and the Erosion of
Democratic Norms
Fiske, A. P., & Rai, T. S. (2014). Virtuous violence: Hurting and killing to create, sustain, end, and honor social
relationships. Cambridge University Press.
Fleeson, W., & Jayawickreme, E. (2015). Whole trait theory. Journal of research in personality, 56, 82-92.
Fowers, B. J., Carroll, J. S., Leonhardt, N. D., & Cokelet, B. (2021). The emerging science of virtue. Perspectives on
Psychological Science, 16(1), 118-147.
Frankovic, K., Bialik, C., & Orth, T. (2023, March 30). Which issues Americans care about most and how they
evaluate their current congressional leaders. YouGov. https://today.yougov.com/politics/articles/45495-
which-issues-americans-care-about-most-poll?redirect_from=%2Ftopics%2Fpolitics%2Farticles-
reports%2F2023%2F03%2F30%2Fwhich-issues-americans-care-about-most-poll
Frimer, J. A., Skitka, L. J., & Motyl, M. (2017). Liberals and conservatives are similarly motivated to avoid
exposure to one another's opinions. Journal of Experimental Social Psychology, 72, 1-12.
Funder, D. C., & Ozer, D. J. (2019). Evaluating effect size in psychological research: Sense and nonsense. Advances
in methods and practices in psychological science, 2(2), 156-168.
Furnham, A., & Thorne, J. D. (2013). Need for cognition: Its dimensionality and personality and intelligence
correlates. Journal of Individual Differences, 34(4), 230–240. https://doi.org/10.1027/1614-0001/a000119
Garrett, K. N. (2019). Fired up by morality: The unique physiological response tied to moral conviction in politics.
Political Psychology, 40(3), 543-563..
Garrett, K. N., & Bankert, A. (2020). The moral roots of partisan division: How moral conviction heightens
affective polarization. British Journal of Political Science, 50(2), 621-640.
Goering, M., Espinoza, C. N., Mercier, A., Eason, E. K., Johnson, C. W., & Richter, C. G. (2024). Moral identity in
relation to emotional well-being: a meta-analysis. Frontiers in Psychology, 15, 1346732.
Goodwin, G. P., & Darley, J. M. (2008). The psychology of meta-ethics: Exploring objectivism. Cognition, 106(3),
1339-1366.
Goodwin, G. P., Piazza, J., & Rozin, P. (2014). Moral character predominates in person perception and evaluation.
Journal of personality and social psychology, 106(1), 148.
Gowans, C. (2021). Moral Relativism, The Stanford Encyclopedia of Philosophy (Spring 2021 Edition), Edward N.
Zalta (ed.), URL = .
Graham, M. H., & Svolik, M. W. (2020). Democracy in America? Partisanship, polarization, and the robustness of
support for democracy in the United States. American Political Science Review, 114(2), 392-409.
Green, P., & MacLeod, C. J. (2016). SIMR: An R package for power analysis of generalized linear mixed models
by simulation. Methods in Ecology and Evolution, 7(4), 493-498.
119
Grubbs, J. B., Warmke, B., Tosi, J., James, A. S., & Campbell, W. K. (2019). Moral grandstanding in public
discourse: Status-seeking motives as a potential explanatory mechanism in predicting conflict. PloS one,
14(10), e0223749.
Grubbs, J. B., Warmke, B., Tosi, J., & James, A. S. (2020). Moral grandstanding and political polarization: A multi-
study consideration. Journal of Research in Personality, 88, 104009.
Haidt, J. (2012). The righteous mind: Why good people are divided by politics and religion. Vintage.
Haidt, J., & Kesebir, S. (2010). Morality. In S. T. Fiske, D. T. Gilbert, & G. Lindzey (Eds.), Handbook of social
psychology (5th ed., pp. 797–832). John Wiley & Sons, Inc..
https://doi.org/10.1002/9780470561119.socpsy002022
Halevy, N., Kreps, T. A., Weisel, O., & Goldenberg, A. (2015). Morality in intergroup conflict. Current Opinion in
Psychology, 6, 10-14.
Hare, J (2019). Religion and Morality, The Stanford Encyclopedia of Philosophy (Fall 2019 Edition), Edward N.
Zalta (ed.), URL = .
Hill, E. D., Cohen, A. B., Terrell, H. K., & Nagoshi, C. T. (2010). The role of social cognition in the religious
fundamentalism‐prejudice relationship. Journal for the Scientific Study of Religion, 49(4), 724-739.
Hobolt, S. B., Lawall, K., & Tilley, J. (2023). The polarizing effect of partisan echo chambers. American Political
Science Review, 1-16.
Hooper, D., Coughlan, J. and Mullen, M. R. “Structural Equation Modelling: Guidelines for Determining Model
Fit.” The Electronic Journal of Business Research Methods Volume 6 Issue 1 2008, pp. 53 - 60, available
online at www.ejbrm.com
Hoyle, R. H., Davisson, E. K., Diebels, K. J., & Leary, M. R. (2016). Holding specific views with humility:
Conceptualization and measurement of specific intellectual humility. Personality and Individual
Differences, 97, 165-172.
Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional
criteria versus new alternatives. Structural equation modeling: a multidisciplinary journal, 6(1), 1-55.
Imai, K., Keele, L., & Tingley, D. (2010). A general approach to causal mediation analysis. Psychological methods,
15(4), 309.
Iyengar, S., Lelkes, Y., Levendusky, M., Malhotra, N., & Westwood, S. J. (2019). The origins and consequences of
affective polarization in the United States. Annual review of political science, 22, 129-146.
Jonason, P. K., & Webster, G. D. (2010). The dirty dozen: a concise measure of the dark triad. Psychological
assessment, 22(2), 420.
Jung, J. H., & Clifford, S. (2024). Varieties of Values: Moral Values Are Uniquely Divisive. American Political
Science Review, 1-17.
Kalmoe, N. P. & Mason, L. Radical American Partisanship: Mapping Violent Hostility, Its Causes, and the
Consequences for Democracy (Univ. of Chicago Press, 2022).
Kingzette, J. et al. How affective polarization undermines support for democratic norms. Public Opin. Q. 85, 663–
677 (2021).
120
Kodapanakkal, R. I., Brandt, M. J., Kogler, C., & van Beest, I. (2022). Moral frames are persuasive and moralize
attitudes; nonmoral frames are persuasive and de-moralize attitudes. Psychological Science, 33(3), 433-
449.
Kovacheff, C., Schwartz, S., Inbar, Y., & Feinberg, M. (2018). The problem with morality: Impeding progress and
increasing divides. Social Issues and Policy Review, 12(1), 218-257.
Kraaijeveld, S. R., & Jamrozik, E. (2022). Moralization and mismoralization in public health. Medicine, Health
Care and Philosophy, 25(4), 655-669.
Krumrei-Mancuso, E. J., & Newman, B. (2020). Intellectual humility in the sociopolitical domain. Self and Identity,
19(8), 989-1016.
Kuhne, P. & Kamin, J. (n.d.) Social Cohesion Impact Measure (SCIM): A tool to measure the impact of bridging
interventions. Retrieved from
https://docs.google.com/document/d/1X2l379C7zYTH4EGOLilyMIig8QG7EdstVigvO7wGpFs/edit
Leach, C. W., Bilali, R., & Pagliaro, S. (2015). Groups and morality. In M. Mikulincer, P. R. Shaver, J. F. Dovidio,
& J. A. Simpson (Eds.), APA handbook of personality and social psychology, Vol. 2. Group processes (pp.
123–149). American Psychological Association
Leary, M. R., Diebels, K. J., Davisson, E. K., Jongman-Sereno, K. P., Isherwood, J. C., Raimi, K. T., ... & Hoyle, R.
H. (2017). Cognitive and interpersonal features of intellectual humility. Personality and Social Psychology
Bulletin, 43(6), 793-813.
Lees, J., & Cikara, M. (2020). Inaccurate group meta-perceptions drive negative out-group attributions in
competitive contexts. Nature human behaviour, 4(3), 279-286.
Lelkes, Y., & Westwood., S (2023). Global Political Pulse. https://americaspoliticalpulse.com/global
Levy, R. E. (2021). Social media, news consumption, and polarization: Evidence from a field experiment. American
economic review, 111(3), 831-870.
Lins de Holanda Coelho, G., HP Hanel, P., & J. Wolf, L. (2020). The very efficient assessment of need for
cognition: Developing a six-item version. Assessment, 27(8), 1870-1885.
Liu, Q., & Nesbit, J. C. (2023). The relation between need for cognition and academic achievement: A meta-
analysis. Review of Educational Research, 00346543231160474.
Luttrell, A., Petty, R. E., Briñol, P., & Wagner, B. C. (2016). Making it moral: Merely labeling an attitude as moral
increases its strength. Journal of Experimental Social Psychology, 65, 82-93.
MacAskill, M., Bykvist, K., & Ord, T. (2020). Moral uncertainty (p. 240). Oxford University Press.
Marie, A., Altay, S., & Strickland, B. (2023). Moralization and extremism robustly amplify myside sharing. PNAS
nexus, 2(4), pgad078.
Martherus, J. L., Martinez, A. G., Piff, P. K., & Theodoridis, A. G. (2021). Party animals? Extreme partisan
polarization and dehumanization. Political Behavior, 43, 517-540.
Mason, L. (2018). Ideologues without issues: The polarizing consequences of ideological identities. Public Opinion
Quarterly, 82(S1), 866-887.
121
McElroy-Heltzel, S. E., Davis, D. E., DeBlaere, C., Worthington Jr, E. L., & Hook, J. N. (2019). Embarrassment of
riches in the measurement of humility: A critical review of 22 measures. The Journal of Positive
Psychology, 14(3), 393-404.
McElroy, S. E., Rice, K. G., Davis, D. E., Hook, J. N., Hill, P. C., Worthington Jr, E. L., & Van Tongeren, D. R.
(2014). Intellectual humility: Scale development and theoretical elaborations in the context of religious
leadership. Journal of Psychology and Theology, 42(1), 19-30
McCoy, J., & Somer, M. (2019). Toward a theory of pernicious polarization and how it harms democracies:
Comparative evidence and possible remedies. The Annals of the American Academy of Political and Social
Science, 681(1), 234-271.
McCoy, J., Rahman, T., & Somer, M. (2018). Polarization and the global crisis of democracy: Common patterns,
dynamics, and pernicious consequences for democratic polities. American Behavioral Scientist, 62(1), 16–
42. https://doi.org/10.1177/0002764218759576
McGarry, P. P., Shteynberg, G., Hulsey, T. L., & Heim, A. S. (2023). The Great Divide: Neither Fairness Nor
Kindness Eliminates Moral Derogation of People With Opposing Political Beliefs. Social Psychological
and Personality Science, 19485506231194279.
McLaughlin, A. T., Van Tongeren, D. R., McElroy-Heltzel, S. E., Bowes, S. M., Rice, K. G., Hook, J. N., ... &
Davis, D. E. (2023). Intellectual humility in the context of existential commitment. The journal of positive
psychology, 18(2), 289-303.
Mernyk, J. S., Pink, S. L., Druckman, J. N., & Willer, R. (2022). Correcting inaccurate metaperceptions reduces
Americans’ support for partisan violence. Proceedings of the National Academy of Sciences, 119(16),
e2116851119.
Miller, J. D., Back, M. D., Lynam, D. R., & Wright, A. G. (2021). Narcissism today: What we know and what we
need to learn. Current Directions in Psychological Science, 30(6), 519-525.
Moore-Berg, S. L., & Hameiri, B. (2024). Improving intergroup relations with meta-perception correction
interventions. Trends in Cognitive Sciences
Moore-Berg, S. L., Ankori-Karlinsky, L.-O., Hameiri, B. & Bruneau, E. (2020).
Exaggerated meta-perceptions predict intergroup hostility between American political partisans. Proc. Natl Acad.
Sci. USA 117, 14864–14872
Nicolas, G., Fiske, S. T., Koch, A., Imhoff, R., Unkelbach, C., Terache, J., Carrier, A., & Yzerbyt, V. (2022).
Relational versus structural goals prioritize different social information. Journal of Personality and Social
Psychology, 122(4), 659–682. https://doi.org/10.1037/pspi0000366
Olson, K., Camp, C., & Fuller, D. (1984). Curiosity and need for cognition. Psychological reports, 54(1), 71-74
Owens, B. P., Yam, K. C., Bednar, J. S., Mao, J., & Hart, D. W. (2019). The impact of leader moral humility on
follower moral self-efficacy and behavior. Journal of Applied Psychology, 104(1), 146.
Paluck, E. L., Porat, R., Clark, C. S., & Green, D. P. (2021). Prejudice reduction: Progress and challenges. Annual
review of psychology, 72, 533-560.
Pasek, M. H., Ankori-Karlinsky, L. O., Levy-Vene, A., & Moore-Berg, S. L. (2022). Misperceptions about out-
partisans’ democratic values may erode democracy. Scientific Reports, 12(1), 16284.
122
Peer, E., Rothschild, D., & Gordon, A. (2022). Data quality of platforms and panels for online behavioral research.
Behav Res 54, 1643–1662 (2022).
Pekrun, R., Vogl, E., Muis, K. R., & Sinatra, G. M. (2017). Measuring emotions during epistemic activities: the
Epistemically-Related Emotion Scales. Cognition and Emotion, 31(6), 1268-1276.
Pett, M. A., Lackey, N. R., & Sullivan, J. J. (2003). Making sense of factor analysis. SAGE Publications, Inc.,
https://doi.org/10.4135/9781412984898
Pew Research Center (2021, May 4). 70% of U.S. social media users never or rarely post or share about political,
social issues. Pew Research Center. https://www.pewresearch.org/short-reads/2021/05/04/70-of-u-s-social-
media-users-never-or-rarely-post-or-share-about-political-social-issues/
Pew Research Center (2019, October 23). National politics on Twitter: Small share of U.S. adults produce majority
of tweets. Pew Research Center. https://www.pewresearch.org/politics/2019/10/23/national-politics-on-
twitter-small-share-of-u-s-adults-produce-majority-of-tweets/
Pinker, S. (2008). The Moral Instinct. New York Times Magazine (2008, January 13).
https://www.nytimes.com/2008/01/13/magazine/13Psychology-t.html
Plante, T. G., & Boccaccini, M. T. (1997). The Santa Clara strength of religious faith questionnaire. Pastoral
Psychology, 45(5), 375-387.
Porter, T., Baldwin, C. R., Warren, M. T., Murray, E. D., Cotton Bronk, K., Forgeard, M. J., ... & Jayawickreme, E.
(2022). Clarifying the content of intellectual humility: A systematic review and integrative framework.
Journal of Personality Assessment, 104(5), 573-585.
Prentice, M., Jayawickreme, E., Hawkins, A., Hartley, A., Furr, R. M., & Fleeson, W. (2019). Morality as a basic
psychological need. Social Psychological and Personality Science, 10(4), 449-460.
Pretus, C., Ray, J. L., Granot, Y., Cunningham, W. A., & Van Bavel, J. J. (2023). The psychology of hate: Moral
concerns differentiate hate from dislike. European Journal of Social Psychology, 53(2), 336-353.
Priem, R. L., Wenzel, M., & Koch, J. (2018). Demand-side strategy and business models: Putting value creation for
consumers center stage. Long range planning, 51(1), 22-31.
Puryear, C., Kubin, E., Schein, C., Bigman, Y., & Gray, K. (2022). Bridging political divides by correcting the basic
morality bias. Retrieved from: https://osf.io/preprints/psyarxiv/fk8g6
Reicher, S., Haslam, S. A., & Rath, R. (2008). Making a virtue of evil: A five‐step social identity model of the
development of collective hate. Social and Personality Psychology Compass, 2(3), 1313-1344.
Robins, R. W., Hendin, H. M., & Trzesniewski, K. H. (2001). Measuring Global Self-Esteem: Construct Validation
of a Single-Item Measure and the Rosenberg Self-Esteem Scale. Personality and Social Psychology
Bulletin, 27, 151-161.
Rozin, P. (2006). Domain denigration and process preference in academic psychology. Perspectives on
Psychological Science, 1(4), 365-376.
Ruberton, P. M., Kruse, E., & Lyubomirsky, S. (2016). 18Boosting State Humility via Gratitude, Self-Affirmation,
and Awe: Theoretical and Empirical Perspectives. In Handbook of humility (pp. 276-289). Routledge.
Ruggeri, K., Većkalov, B., Bojanić, L., Andersen, T. L., Ashcroft-Jones, S., Ayacaxli, N., ... & Folke, T. (2021). The
general fault in our fault lines. Nature Human Behaviour, 5(10), 1369-1380.
123
Ryan, T. J. (2014). Reconsidering moral issues in politics. The Journal of Politics, 76(2), 380-397.
Ryan, T. J. (2017). No compromise: Political consequences of moralized attitudes. American Journal of Political
Science, 61(2), 409-423.
Salgado, J. F. (2017). Bandwidth-fidelity dilemma. Encyclopedia of personality and individual differences, 1-4.
Schein, C., & Gray, K. (2018). The theory of dyadic morality: Reinventing moral judgment by redefining harm.
Personality and Social Psychology Review, 22(1), 32-70.
Schmidt, K., Jones, A. L., Szabelska, A., Ebersole, C. R., Hawkins, C. B., Graham, J., & Nosek, B. A. (2023,
October 17). The Ideology 2.0 Study and Dataset. Retrieved from osf.io/2483h
Schroeders, U., Wilhelm, O., & Olaru, G. (2016). Meta-heuristics in short scale construction: Ant colony
optimization and genetic algorithm. PloS one, 11(11), e0167110.
Schumacker, R. E., & Lomax, R. G. (2004). A beginner's guide to structural equation modeling. psychology press.
Sedikides, C., Meek, R., Alicke, M. D., & Taylor, S. (2014). Behind bars but above the bar: Prisoners consider
themselves more prosocial than non‐prisoners. British Journal of Social Psychology, 53(2), 396-403.
Sibley, C. G., Osborne, D., & Duckitt, J. (2012). Personality and political orientation: Meta-analysis and test of a
Threat-Constraint Model. Journal of Research in Personality, 46(6), 664-677.
Simonovits, G., McCoy, J., & Littvay, L. (2022). Democratic hypocrisy and out-group threat: explaining citizen
support for democratic erosion. The Journal of Politics, 84(3), 1806-1811.
Soto, C. J., & John, O. P. (2017). Short and extra-short forms of the Big Five Inventory–2: The BFI-2-S and BFI-2-
XS. Journal of Research in Personality, 68, 69-81.
Skitka, L. J., Bauman, C. W., & Sargis, E. G. (2005). Moral conviction: Another contributor to attitude strength or
something more?. Journal of personality and social psychology, 88(6), 895.
Skitka, L. J., Hanson, B. E., Morgan, G. S., & Wisneski, D. C. (2021). The psychology of moral conviction. Annual
Review of Psychology, 72, 347-366.
Skitka, L. J., & Mullen, E. (2002). The dark side of moral conviction. Analyses of Social Issues and Public Policy,
2(1), 35-41.
Smith, I. H., & Kouchaki, M. (2018). Moral humility: In life and at work. Research in Organizational Behavior, 38,
77-94.
Strohminger, N., & Nichols, S. (2014). The essential moral self. Cognition, 131(1), 159-171.
Stanley, M. L., Bedrov, A., Cabeza, R., & De Brigard, F. (2020). The centrality of remembered moral and immoral
actions in constructing personal identity. Memory, 28(2), 278-284.
Tangney, J. P. (2009). Humility. In S. J. Lopez & C. R. Snyder (Eds.), Oxford handbook of positive psychology (2nd
ed., pp. 483–490). Oxford University Press.
Tannenbaum, M. B., Hepler, J., Zimmerman, R. S., Saul, L., Jacobs, S., Wilson, K., & Albarracín, D. (2015).
Appealing to fear: A meta-analysis of fear appeal effectiveness and theories. Psychological bulletin,
141(6), 1178.
124
Tingley, D., Yamamoto, T., Hirose, K., Keele, L., & Imai, K. (2014). Mediation: R package for causal mediation
analysis.
Tersman, Folk (2022). Moral Disagreement, The Stanford Encyclopedia of Philosophy (Fall 2022 Edition), Edward
N. Zalta & Uri Nodelman (eds.), URL = .
Tosi, J., & Warmke, B. (2020). Grandstanding: The use and abuse of moral talk. Oxford University Press, USA.
Tomasello, M., & Vaish, A. (2013). Origins of human cooperation and morality. Annual review of psychology, 64,
231-255.
Tracy, J. L., Cheng, J. T., Robins, R. W., & Trzesniewski, K. H. (2009). Authentic and hubristic pride: The affective
core of self-esteem and narcissism. Self and identity, 8(2-3), 196-213.
Vallabha, S. & Brandt, M. J. (2024, April 5). Moral Humility Reduces Political Divisions.
https://doi.org/10.31234/osf.io/5r6fw
Van Bavel, J. J., Packer, D. J., Haas, I. J., & Cunningham, W. A. (2012). The importance of moral construal: Moral
versus non-moral construal elicits faster, more extreme, universal evaluations of the same actions. PloS
one, 7(11), e48693.
Van Bavel, J. J., Robertson, C. E., Del Rosario, K., Rasmussen, J., & Rathje, S. (2024). Social media and morality.
Annual review of psychology, 75, 311-340.
Van Tongeren, D. R., Stafford, J., Hook, J. N., Green, J. D., Davis, D. E., & Johnson, K. A. (2016). Humility
attenuates negative attitudes and behaviors toward religious out-group members. The Journal of Positive
Psychology, 11(2), 199-208.
Van Tongeren, D. R., Davis, D. E., Hook, J. N., & Witvliet, C. vanOyen. (2019). Humility. Current Directions in
Psychological Science, 28(5), 463-468. https://doi.org/10.1177/0963721419850153
Van Tongeren, D. R., Ng, V., Hickman, L., & Tay, L. (2023). Behavioral measures of humility: Part 2. Conceptual
mapping and charting ways forward. The Journal of Positive Psychology, 18(5), 722-732.
Van Zomeren, M., Postmes, T., & Spears, R. (2012). On conviction's collective consequences: Integrating moral
conviction with the social identity model of collective action. British Journal of Social Psychology, 51(1),
52-71.
Voelkel, J. G., Chu, J., Stagnaro, M. N., Mernyk, J. S., Redekopp, C., Pink, S. L., ... & Willer, R. (2023).
Interventions reducing affective polarization do not necessarily improve anti-democratic attitudes. Nature
human behaviour, 7(1), 55-64.
Voelkel, J. G., Stagnaro, M., Chu, J., Pink, S. L., Mernyk, J. S., Redekopp, C., … Willer, R. (2023, March 20).
Megastudy identifying effective interventions to strengthen Americans’ democratic attitudes.
https://doi.org/10.31219/osf.io/y79u5.
Wilson, T. D., Aronson, E., & Carlsmith, K. (2010). The art of laboratory experimentation. In S. T. Fiske, D. T.
Gilbert, & G. Lindzey (Eds.), The handbook of social psychology (5th ed., Vol. 1, pp. 51-81). Hoboken, NJ:
Wiley.
Williams, E. G. (2015). The possibility of an ongoing moral catastrophe. Ethical Theory and Moral Practice, 18,
971-982.
125
Wilson, T. D., Aronson, E., & Carlsmith, K. (2010). The art of laboratory experimentation. In S. T. Fiske, D. T.
Gilbert, & G. Lindzey (Eds.), The handbook of social psychology (5th ed., Vol. 1, pp. 51-81). Hoboken, NJ:
Wiley.
Wright, J. C., Nadelhoffer, T., Perini, T., Langville, A., Echols, M., & Venezia, K. (2017). The psychological
significance of humility. The Journal of Positive Psychology, 12(1), 3-12.
Wright, J. C., & Pölzler, T. (2022). Should morality be abolished? An empirical challenge to the argument from
intolerance. Philosophical psychology, 35(3), 350-385.
Yoder, K. J., & Decety, J. (2014). Spatiotemporal neural dynamics of moral judgment: A high-density ERP study.
Neuropsychologia, 60, 39-45.
Zimbardo, Philip G. "A situationist perspective on the psychology of evil: Understanding how good people are
transformed into perpetrators." The social psychology of good and evil (2004): 21-50.
Zmigrod, L., Rentfrow, P. J., Zmigrod, S., & Robbins, T. W. (2019). Cognitive flexibility and religious disbelief.
Psychological research, 83(8), 1749-1759.
126