‘

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:IProj/Acc&Pres/ClRC/DateDue,indd

 

 

TITLE: THE EFFECT OF LEARNING TYPE AND AVATAR SIMILARITY ON
LEARNING OUTCOMES IN EDUCATIONAL VIDEO GAMES

By

Melissa L. Lewis

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILSOPHY
Communication Arts & Sciences — Media & Information Studies

2009

ABSTRACT

THE EFFECT OF LEARNING TYPE AND AVATAR SIMILARITY ON LEARNING
OUTCOMES IN EDUCATIONAL VIDEO GAMES

By
Melissa L. Lewis

Two theories guide two very different ideas about learning. Social
cognitive theory (Bandura, 1977, 1989) places the greater emphasis on
observational learning, or learning by watching a model produce a behavior
before doing it oneself. Other researchers purport that experiential learning, or
learning by doing, results in stronger learning (Kolb, 1984). Neither of these
theoretical areas have been adequately explored with regard to video games,
and they have never been compared before. As well, the entertainment-
education paradigm (EEP), a third theoretical approach (Singhal & Rogers,
2002), uses many of the same constructs as SCT, and supports the idea that
gaining attention and motivating learning by using intrinsically-motivating games
will then increase learning. This research ﬁrst explores which of these
paradigms—SOT and the EEP, or experiential leaming—is more effective in
producing outcomes such as knowledge retained, involvement, intention to
perform behaviors in the future, enjoyment, and self-efﬁcacy.

Additionally, SCT posits that identiﬁcation with the model one is observing
increases positive learning results, and video games allow for the creation of an
avatar with which one can adjust identiﬁcation. People often create avatars to

represent their ideal selves or their actual selves, but it is not yet known which of

these produces the more effective Ieaming results. Because the player becomes
the model when he or she creates an avatar based on self, there possibly exists
a third condition—the self-as-model, or automodeI—that is distinctly different
from a third party observational Ieaming perspective or an enactive perspective.
Given the history of research on video games and avatar use showing that
people tend to idealize themselves in avatar form, expectations are that the ideal
self avatar would yield the greatest Ieaming outcomes, followed by real self (both
conditions that offer extreme identiﬁcation with the model), and lastly the third-
party avatar. This is the second major exploration this research undertakes.
Results indicate that enactive Ieaming results in higher self-efﬁcacy.
Behavioral intentions and knowledge retained did not reach signiﬁcance, but
means were higher in the enactive condition. Ideal, real, or other observational
conditions also did not reach signiﬁcance in an analysis of variance, but again,
means were highest for the ideal self avatar, and correlations showed
relationships in the hypothesized direction for enjoyment of the game, enjoyment

of learning, and enjoyment of the avatar.

 

Copyright by
MELISSA L. LEWIS
2009

ACKNOWLEDGMENTS

In my case, I can say that writing the dissertation was as hard on me as it
was for the people around me, so those people who have stuck with me for the
long haul deserve some serious thanks.

First, I thank my advisor, supervisor, and mentor, Dr. Robert LaRose.
Wrthout his guidance, feedback, and, most importantly, his unwavering support, I
would not be where I am today, nor who I am today. I want to thank Dr. Nora
Rifon, my committee member, my advisor, and my friend, for always having a

willing ear during those times when I needed to toss around ideas, and for

 

encouraging me to do what it was that I wanted to do. I must also thank my
committee member and colleague, Dr. Thomas Muth, whose encouragement and
perpetual optimism kept me going. There were many times when the support of
these individuals was all that got me through that day.

I thank Dr. David Lawrence, Dr. Ronald Kyhos, and Dr. Robert Drew, who
originally planted the idea of graduate school in my brain. I could not even begin
to put my gratitude into words, except to say, “Thank you for changing my life”.

I have to thank Tern' Lukowski, for all those times she listened to my ideas,
complaining, and tears. I also have to thank her for all those times she said,
“Fine, then don’t ﬁnish it". Nothing works better than reverse psychology, sis.

And ﬁnally, I want to thank Mary and Frank Lewis, my parents, for instilling
in me the drive to succeed and the love of Ieaming, for giving me all the
opportunities I could ever want, and for being willing to spend another ﬁve years

telling people that yes, their daughter was still in school.

TABLE OF CONTENTS

List of Tables ....................................................................................... viii
List of Figures ....................................................................................... ix
Chapter 1: Statement of the Problem ......................................................... 1
Chapter 2: Review of Literature ................................................................. 3
Social Cognitive Theory .................................................................. 3
Observational Learning .................................................................. 3
Enactive Learning ......................................................................... 7

The Role of Self-Efﬁcacy in Learning ................................................ 9

SCP and the EEP ........................................................................ 13

The Case of Avatars: Extending SCP .............................................. 17
Chapter 3: Methodology ........................................................................ 25
Sample ..................................................................................... 25
Measures .................................................................................. 26

Scale Reliabilities ........................................................................ 30

The Stimulus Instrument ............................................................... 31

The Avatar Manipulation ............................................................... 34
Procedure ................................................................................. 35

Chapter 4: Results ................................................................................ 37
Manipulation Check ..................................................................... 37
Statistical Analyses ..................................................................... 38
Hypothesis 1 ..................................................................... 39

Hypothesis 2 ..................................................................... 44

Hypothesis 3 ..................................................................... 44

Hypothesis 4 ..................................................................... 45

Chapter 5: Discussion ........................................................................... 48

Enactive Versus Observational Learning .......................................... 48

Ideal, Real, or Other Avatar ........................................................... 56
Chapter 6: Limitations ........................................................................... 61
Manipulation Check ..................................................................... 61
Hypothesis 1 .............................................................................. 65
Hypothesis 2 .............................................................................. 67
Hypothesis 3 .............................................................................. 67
Hypothesis 4 .............................................................................. 68

Chapter 7: Directions for Future Study ...................................................... 74

vi

Appendices ......................................................................................... 81

References ......................................................................................... 94

vii

LIST OF TABLES

Table 1: Mean Distance Scores and Scale Reliabilities ................................. 35
Table 2: Means, Standard Deviations, and Cell Sizes for Enactive Learning, Ideal
Self, Real SeIf, and Friend ...................................................................... 39
Table 3: F-Tests for Dependent Variables and Condition .............................. 40
Table 4: Signiﬁcance Values for Planned Comparisons Between Enactive and

Observational Conditions ....................................................................... 43
Table 5: Planned Comparisons for Ideal, Real, and Friend Conditions ............ 45
Table 6: Correlations Between Distances and Dependent Variables ................ 46
Table 7: Correlation Matrix for Involvement Scale Items ............................... 81
Table 8: Correlation Matrix for Self-Efﬁcacy Scale Items ............................... 81
Table 9: Correlation Matrix for Behavioral Intentions Scale Items. .................. 82

viii

LIST OF FIGURES

Figure 1: Opening Screen (with Avatar) .................................................... 83
Figure 2: Phishing Email Example ........................................................... 84
Figure 3: Phishing Completion Screen (without Avatar) ................................ 85
Figure 4: Accepting Non-Invasive Cookie (with Avatar) ................................. 86
Figure 5: Instant Message Level (with Avatar) ............................................ 87
Figure 6: Moving On To Final Level (with Avatar) ........................................ 88
Figure 7: Setting Automatic Updates (with Avatar) ....................................... 89
Figure 8: Pop-Up Blocker (with Avatar) ..................................................... 90
Figure 9: PopoUp Blocker (without Avatar) ................................................. 91
Figure 10: Final Screen (with Avatar) ........................................................ 92

ix

Chapter 1: Statement of the Problem

Bandura (1986) discussed in his seminal work on Social Cognitive Theory
that there are two major types of Ieaming: observational Ieaming, and enactive
Ieaming. Enactive Ieaming involves Ieaming by doing it oneself, with a great
deal of trial and error. Observational Ieaming has Ieamers watching the
performance of a behavior and then modeling their own performances on that
which they have watched. Which type of Ieaming is the more effective, however,
has not yet been adequately explored. According to Bandura (1986),
observation is the key to Ieaming (Lerner & Steinberg, 2004). However, other
researchers who support experiential Ieaming theories believe that enactive
Ieaming provides the more effective means of Ieaming (Kolb, 1984). Until now,
there has not been an adequate way to compose an experiment that could
examine this issue while ensuring precise similarities between conditions, but
with the advent of video games, we can create those similar conditions with
different agencies: self (enactive) and other (observational). Thus, it is possible
to begin to answer the question of which type of Ieaming is the more effective
type.

As well, Social Cognitive Theory states that Ieaming will occur when there
is a close identiﬁcation between the model and the individual. But what happens
when we are the model? In video games, “self" is almost always controlling the
action to some degree, and when the self becomes the model, Ieaming type
moves from observational Ieaming to enactive Ieaming. Would the observation

of a character (as opposed to an entirely enactive, ﬁrst-person perspective) be

more effective? And with what would we feel the most similarity: our actual self,
or our ideal self? This new case—being able to see an external version of
ourselves, as opposed to, for instance, a virtual reality simulation where we are
immersed so deeply into the environment that we literally do not see our physical
bodies anymore—should be explored and documented to determine what the
use of avatars might mean for Ieaming.

Moreover, if observational Ieaming is, as Bandura posits, the better
method for Ieaming, then what type of agent works best as the model? A
representation of oneself, or a representation of some external other? And
should these representations be real, or idealized? Social cognitive theory is one
of the most ubiquitous theories when it comes to education and Ieaming, but
researchers have most often focused on one or the other type of learning, and
have not explored SCT with regard to avatars.

In order to begin to answer these questions, this research designs a study
using an electronic video game that manipulates both agency and similarity. The
enactive condition is where players are instructed that they are playing as
themselves. The observational condition is where players are instructed that
they are teaching some “other” being, the game character representing perhaps
a friend of the player's. Then, conditions are assigned as real self, ideal self, and

a third party.

Chapter 2: Review of Literature
Social Cognitive Theory

How do we Ieam? Every human being has at least some capacity for
Ieaming, but how that Ieaming occurs can be explained in different ways. Social
cognitive theory, developed by Albert Bandura, explains Ieaming and behavior as
a combination of behavioral, personal, and environmental interactions (Bandura,
1986). Personal factors include cognitive, affective, and biological events.
Behavioral factors refer to behavioral patterns, and environmental events are
what they sound like: things that occur outside of the individual but affect that
individual’s life.

Learning with social cognitive theory can be either observational or
enactive (Bandura, 1986). Observational learning occurs when individuals watch
someone else and then model the behavior that they have just observed.
Enactive Ieaming occurs when the individual performs the behavior herself, often
through trial and error, without viewing the behavior ﬁrst. Both types of learning
are discussed in greater detail below.

Observational Leaming

Observational Ieaming, according to Bandura (1986), is the key to Ieaming
(Lerner 8 Steinberg, 2004). How well an individual learns depends on how
carefully that person observes what the other person (the model) is doing
(Ferrari, 1996). Observational Ieaming is the ﬁrst type of Ieaming that we
experience. As infants, we learn to mimic what we see our parents (or primary

caregivers) doing, Ieaming language, movement, and emotion simply by

observing someone else (Bennett, Farrington, & Huesmann, 2005). \ﬁsual
demonstration is one of the most powerful ways to communicate behavioral and
cognitive patterns (Wesch, Law, & Hall, 2007).

Observational Ieaming can be more efﬁcient than enactive Ieaming. By
having a model whose behavior one can observe, one is left with a mental image
of how something is to be done (Weeks 8 Anderson, 2000). Learners do not
have to work through a trial-and-error process of learning but rather can access
the cognitive representation formed by modeling and base their behavior on this
preconceived mental model (Adams, 1986; Ferrari, 1996; Lee & White, 1990;
Pollock & Lee, 1992). As well, Ieamers have an idea of what the outcome of
their behavior will be if they have observed a model (Bandura, 1986; Ferrari,
1996).

Research has determined that modeling is effective, especially with regard
to children. Children learn values through modeling and social reproduction
(Bennett, Farrington, & Huesmann, 2005). Children who associate with
delinquent peers may develop anti-social, inappropriate cognitive scripts that are
based on observations of these peers. Associating with delinquent peers is
linked to an increase in criminal or violent potential (Farrington, 1998, 2001 b),
and parents who are criminals are more likely than those who are not to have
delinquent and/or violent children (Farrington, 2001a).

Numerous subjects have been studied with regard to observational
Ieaming, particularly scth subjects. Observational Ieaming is effective for

Ieaming mathematics (Schunk & Hanson, 1985, 1989a, 1989b; Schunk, Hanson,

8 Cox, 1987), writing (Graham 8 Harris, 1994; Graham, Harris, 8 Troia, 1998;
Schriver, 1992), speaking and listening (Sonnenschein 8 Whitehurst, 1983,
1984), and reading and writing (Couzijn, 1999). Modeling works for physical
activity, too, from swimming skills (Weiss, McCullagh, Smith, 8 Bertant, 1998) to
strategy, form, and game plans (Cumming, Clark, Ste-Marie, McCullagh, 8 Hall,
2005; Wesch, Law, 8 Hall, 2007). Modeling prior to practice can also increase
initial performance (Landers, 1975; Thomas, Pierce, 8 Ridsdale, 1977; Weeks 8
Anderson, 2000).

There are several mechanisms guiding observational Ieaming that are
important for this discussion: attention, retention, motor reproduction, and
motivation (Bandura, 1986; Wesch, Law, 8 Hall, 2007). Attention processes
determine to which things the individual pays attention. Retention processes
explain what an individual remembers (or does not remember). To remember
the information, it must be encoded into long-tenn memory. The individual must
then be able to reproduce the behavior—motor reproduction. A person has to '
have the capability to coordinate thought and action to enact the behavior.
Finally, the individual must be motivated to learn the behavior, which affects the
other three processes. A person must be motivated ﬁrst to pay attention, then to
remember, and ﬁnally to practice the behavior. Motivation can be internal or
external. A person might want to learn and practice a behavior because she is
motivated to better herself, or because she has been promised a thousand
dollars. Wherever motivation comes from, however, it must be strong enough to

get the person to reproduce the behavior (Bandura, 1986).

Similarity between the model and the individual affects the modeling
(Schunk, 1987). Modeling is a type of social comparison (Berger, 1977). When
external standards are not readily accessible, people will compare themselves to
others to evaluate themselves (Festinger, 1954). For instance, weak Ieamers
learn better when they have weak models on which to focus. Better Ieamers
learn more by focusing on stronger models (Braaksma, Rijlaarsdam, 8
vandenBergh, 2002). However, similarity does not automatically enhance
learning from modeling (Schunk, 1987). Some variables do not make much of a
difference. Past researchers assumed that gender would make a difference, but
subjects did not learn more from members of their own gender (Leifer, Collins,
Gross, Taylor, Andrews, 8 Blackmer, 1971). Gender does matter, however, with
regard to gender roles. Children may categorize the behavior learned from a
certain gender as being appropriate for that sex but not for the other (Schunk,
1987; Zimmerman 8 Koussa, 1975). Therefore, gender is important when it
comes to task appropriateness. Age is important, in that models that are similar
to children had beneﬁcial effects compared to those who were dissimilar in age
(Schunk, 1987).

Why do people learn from models? There are a number of variables that
can affect the direction of attention. When people pay attention to models, they
learn (Leifer et al., 1971). The closer the perceived similarity between the
individual and the model, the greater the inﬂuence that the model has on the
subject’s behavior (Bandura, 1986). Subjects that are high in similarity to the

model with regard to background produce more modeled behavior (Rosekrans,

1967; Schunk, 1987). Observational Ieaming can affect motivation to change, or
to perform a behavior (Wesch, Law, 8 Hall, 2007). It can also increase
motivation, or involvement, with the activity (Clark 8 Ste-Marie, 2007).
Additionally, observation can increase self-efﬁcacy in observers (Bandura, 1986,
1977; Schunk 8 Hanson, 1985, 1989a). Watching others succeed can increase
one’s own belief that he can succeed, which can in turn inﬂuence Ieaming. Self-
efﬁcacy will be discussed in more depth here later.

Enactive Learning

Whereas observational Ieaming involves watching a model and then using
that information to engage in a behavior, enactive Ieaming refers to the process
of attempting a behavior oneself using one’s own abilities and skills. “Enactive
mastery experiences refer to actual information a person has about their [sic]
ability to execute a particular behavior gathered from their prior experience with
that task” (Wesch, Law, 8 Hall, 2007, p. 221). Many of the experiences that
individuals gain from observational Ieaming are reﬁned through enactive Ieaming
(Bandura, 1986; Carroll 8 Bandura, 1985). We hear this often as “practice
makes perfect”. First the behavior is observed; then it is enacted.

Individuals still experience outcome effects, but they are direct effects.
Intrinsic response information is gained within the individual. If an individual
punches another person, the individual will feel pain in his hand. Extrinsic
response information is a socially arranged effect. These responses are socially
mediated and occur usually after the behavior has been performed (Bandura,

1986). If an individual punches someone, he will be arrested. Individuals learn

with enactive Ieaming, then, by a trial-and-error process where they perform
behaviors, examine the resultant consequences, and then adjust or continue the
behavior based on this feedback.

Whereas Bandura and supporters are proponents of observational
Ieaming, another school of thought has developed regarding enactive
experiences, referred to as experiential Ieaming. In fact, David Kolb developed a
theory of Ieaming called the experiential Ieaming cycle, composed of four stages
(Kolb, 1984). First is the concrete experience stage, where individuals actively
do something, then the reﬂective observation stage (the individual thinks about
and reﬂects on what just happened), the abstract conceptualization stage, and
ﬁnally, active experimentation.

Learning by doing is not a new idea, but the explanation behind it is
unique. Behaviorists like Skinner (1953) explain that behavior is reinforced by
consequences, known as shaping. However, social cognitive theory explains the
consequences as sources of information and motivation. People Ieam behaviors
that they believe will reward them (Zimmerman 8 Schunk, 2001). Experiential
Ieaming theory holds, however, that Ieaming is most effective when it is based on
personal experience (Spencer, 2003). This type of Ieaming is often used
effectively in classes such as marketing classes (Bobbitt, Inks, Kemp, 8 Mayo, '
2000; Gremler, Hoffman, Keaveney, 8 Wright, 2000; Hoffman 8 Batman, 1997;
Lovelock 8 Wright, 1999; Wright, Bitner, 8 Zeithaml, 1994) and medical

instruction (Maudsley 8 Strivens, 2000; Stanton 8 Grant, 1999).

However, a structured, guided approach was found to be more effective
than a trial-and-error approach, at least in terms of Ieaming to search for
information electronically (Debowski, 2002). Indeed, enactive Ieaming may allow
people to form adequate skills, but not optimal skills (Bandura, 1986). So which
is the more effective method of Ieaming? Taking the SCT stance, we would
expect observational Ieaming to be the strongest method of instruction.

H1: Observational Ieaming will lead to (a) more knowledge retained, (b)

more enjoyment, and (c) more involvement with the topic than enactive

Ieaming.

The Role of Self-Efﬁcacy in Learning

Self-efﬁcacy is deﬁned speciﬁcally by Bandura as the belief in one’s
capability to engage in a particular course of action to achieve important
attainments (Bandura, 1997). Bandura (1977) describes four sources of efﬁcacy
expectations: verbal persuasion, emotional arousal, vicarious experience, and
performance accomplishments.

Verbal persuasion is the ﬁrst source of efﬁcacy expectation deﬁned by
' Bandura. Persuasion is one of the more commonly used means of getting
people to do and think what is desired by the persuader. Persuasion is easy to
do and readily available; a few well-chosen words over a matter of even a few
seconds is all that is needed, as opposed to the previous sources involving either
personal experience (time and effort are needed to actually work through the
task) or vicarious experience (more time and effort are needed to watch

someone else work through the task). However, persuasion is also a less

effective efﬁcacy source than the others listed so far because there is no
authentic experience, either by oneself or by a third party.

Emotional arousal can also be a source of efﬁcacy expectation. It is much
the same as the “ﬁght or ﬂight” response, however, and can have negative
repercussions as well (discussed later in this dissertation with regard to fear
appeals). Thus, emotional arousal can instigate a “ﬁght” response, where the
individual processes the stressful situation and then responds accordingly, or it
can cause a “ﬂight” response instead, leading to avoidance and even more stress
as the individual is not only upset at the initial arousal but increasingly upset at
thoughts of his or her personal ineptitude (Bandura, 1977).

The two more effective sources of self-efﬁcacy are vicarious experience
and progressive mastery. Vicarious experience explains how we learn by
watching others. If an individual observes someone else being successful at an
endeavor, and observes that individual getting through the experience with no
adverse effects, then that individual may feel as if slhe will also experience
success with continued effort and improvement. Modeling behavior of others,
however, tends to lead to weaker efﬁcacy expectations that are more vulnerable
to change and reduction than would the last kind of self-efﬁcacy source:
performance accomplishment.

Performance accomplishment, or progressive mastery of a task, is the
best source of efﬁcacy information, as it is individually tailored and tends to
reduce future negative effects. Repeated successes emphasize an individual’s

mastery of tasks or experiences. If failure is experienced later, the negative

10

effects are likely to be reduwd because they are offset by the already-present
self-efﬁcacy felt due to the previous successes. Additionally, the positive feelings
and increased self-efﬁcacy associated with one particular task can transfer to
increased self-efﬁcacy with other experiences. With modeling behavior,
individuals may see others succeed, but individual differences and the capacity
for social comparison may impede the sense of self-efficacy. In other words, just
because someone else was able to succeed at a task doesn’t mean I will be; I’m
not nearly as skilled at things as the other person might be.

Playing video games can be a type of enactive Ieaming, and given what
we know of the effectiveness of performance accomplishment as a means of
gaining self-efﬁcacy, we would expect that individuals who play a video game will
have higher self-efﬁcacy regarding the material than those who do not play a
game. Indeed, health behavior is one prime example of successful entertainment
education that can result in an increase in self-efﬁcacy. Lieberman (2001)
described the effectiveness of using video games to edumte children on health
behavior. Click Health, Inc., developed a number of games for children that
provide players the ability to take on the role of the main character (with the
same health condition as the child), and the ability to customize the game
character’s self-dare regimens so that they might mirror (model) those of the
child. Findings included an equal amount of knowledge gained from the video
game condition and the non-interactive condition (a thirty-minute long video tape
that participants viewed), but a higher level of enjoyment for those in the video

game condition, and an increase in self-efﬁcacy among those in the game

11

condition. Those in the video tape condition experienced a decrease in self-
efﬁcacy. The opportunity to rehearse behavior is a strong predictor of self-
efﬁcacy (T homas, Cahill, 8 Santilli, 1997) resulting as well in Ieaming gains
(Bandura, 1990). Video games offer these hands-on chances for behavior
rehearsal.

Additionally, self-efﬁcacy can predict future behavior better than past
performance (Bandura, 1986). There are many examples of self-efﬁcacy
interventions that resulted in a change in future behavior, from improving patient
follow-through regarding adherence to medication (Ludman, VonKorff, Katon,
Lin, Walker, Bush 8 Wahab, 2000), to reducing caregiver burnout (Mackenzie 8
Peragine, 2003) and increasing regular physical activity among cancer survivors
(Bennet, Lyons, Winters-Stone, Nail, 8 Scherer, 2007). Research suggests that
increasing self-efﬁcacy (essentially giving people the conﬁdence to carry out the
desired behaviors) increases the likelihood that people will actually enact these
behaviors (Basen-Engquist, 1994; F ruin, Pratt, 8 Owen, 1992; Rippetoe 8
Rogers, 1987; Rimal 8 Real, 2003; Turner, Rimal, Morrison, 8 Kim, 2006).

To summarize, if enactive mastery is the better source of self-efﬁcacy in
previous research, we would expect this to be true for video games.

H2: Enactive Ieaming will lead to higher levels of self-efﬁcacy than

observational Ieaming.

Consequently, if self-efﬁcacy increases the likelihood of engaging in a
behavior, it would also be logical to assume that an increase in self-efﬁcacy due

to Ieaming type would extend to intention to engage in behavior.

12

H3: Enactive Ieaming will result in greater intentions to perform safe online

behavior than observational Ieaming.
SCT and the EEP

Using video games to teach is not a new idea, but using games to educate
is not something that is readily accepted by all educators. In fact, one of the
major barriers to using interactive games to teach is a lack of understanding of
what makes educational games different from arcade-style games (Rice, 2007).
Some researchers have recently called for integrating electronic games into
Ieaming after determining, through empirical study and reviews of prior research,
that electronic games are effective for Ieaming (Dipietro, Ferdig, Boyer, 8 Black.
2007; Gros, 2007; Mayo, 2007).

Different from traditional media such as books and television, video games
are interactive. While players are in control of what happens in the game, video
games adapt the challenges presented based on player performance and
provide immediate reward and feedback. Further, video games are customizable
and can “remember” a player's progress; they therefore personalize game play to
each individual player. Enjoyment of video games is a result of (a) sensory
delight; (b) feeling a sense of achievement, control, and self-efﬁcacy; and (c)
suspense, thrill, and relief (Vorderer, Klimmt, 8 Ritterfeld, 2004).

One of the major reasons why researchers posit that electronic games
work can be explained using the entertainment education paradigm (EEP), which
links the enjoyment of being entertained with the Ieaming and processing of

education, and is deﬁned as “the intentional placement of educational content in

13

entertainment messages” (Singhal 8 Rogers, 2002). There are three pathways
for Ieaming involved in the EEP: motivation, reinforcement, and blending.

The motivation pathway uses entertainment as a “door opener” for .
Ieaming. Individuals might be playing a game for entertainment, but at the same
time, this necessarily allocates attention to the educational content, which then
leads to an interest in the content and ﬁnally processing of the content (Ritterfeld,
Weber, Fernandes, 8 Vorderer, 2004; Vorderer 8 Ritterfeld, 2003). This parallels
the attention stage of observational Ieaming in social cognitive theory. The
second pathway uses entertainment for the reinforcement of Ieaming which
enhances the motivation to process educational content. This is similar to the
motivation stage in SCT. Game players also need to be able to remember the
game behavior that is rewarded (memory in SCT) and reproduce the behavior
that is reinforced in order to get rewarded (reproduction in SCT). Reinforcement
is almost always integrated into interactive video games, through scores,
feedback of progress in the game, adaption to skill levels, or rewards in the form
of obtaining money, new objects, and other desirable goods. The EEP does not
formally associate itself with social cognitive theory, although given the parallels
between the two paradigms, researchers ought to begin integrating both of these
when studying educational video games.

The sense of self-efﬁcacy and the enjoyment of playing reinforce Ieaming
and enhance the potential for a person to seek out more of these entertaining
and educational experiences, thus leading to more Ieaming opportunities

(Ritterfeld, Weber, Fernandes, 8 Vorderer, 2004). However, the greatest

14

potential for entertainment education occurs when education becomes an
essential part of the entertainment experience (Slater, 2002). Game design
according to the EEP is not simply taking an educational game and adding
entertaining elements; it is a game in which the entertaining experience is
inseparable from the game's educational goal.

Research has demonstrated that turning Ieaming into a game can
encourage less interested individuals to take an active part in Ieaming, as well as
give the already motivated a new and exciting way to study the material (Kafai,
1994; Lieberman, 2006; Ritterfeld 8 Weber, 2006), and that the skills learned in
computer games may transfer to other areas, lending themselves to increases in
overall technological literacy (Gee, 2003; Lieberman, 2006; Prensky, 2005).
Proponents of Ieaming with computer games indicate that indeed, learning does
occur and that use of games can even increase motivation, an integral part of
observational Ieaming (cf. Amory, Naicker, Wncent, 8 Adams, 1999; Bosworth,
1994; Egenfeldt-Nielsen, 2007; Egloff, 2004; Lieberman, 2001; Singhal, Cody,
Rogers, 8 Sabido, 2004; Subrahmanyam 8 Greenﬁeld, 1994; Thomas, Cahill, 8
Santilli, 1997; Walshe, Lewis, Kim, O’Sullivan, 8 Wrederhold, 2003; Yoon 8
Godwin, 2007). While some of these studies have methodological shortcomings,
there are studies with fairly sound methodologies. Lieberman (2001) randomized
in-patient participants in an experiment regarding self-care for children suffering
from asthma. One group played a video game for thirty minutes while another
group watched a thirty-minute video. The video game was an observational

experience. The main character was not the child but rather a dinosaur who

15

suffered from the same health issue as the child. The level of knowledge gained
for both groups was the same, but enjoyment was greater in the video game
condition, and while self-efﬁcacy increased for the game players, it decreased for
the video watchers.

The i-Safety project, in progress since 2005, has been funded by the
National Science Foundation as well as Microsoft (Wrrth, Rifon, LaRose. 8 Lewis,
2007). The purpose of the project, in part, was to develop a curriculum that will
teach individuals of any age how to be safe online. The curriculum has been
tested numerous times, using various subject populations, from teenagers
through adults, and has been veriﬁed through experiments and data analysis to
contain valuable, valid information that does indeed teach people how to protect
themselves against online threats such as phishing attacks, spyware, identity
theft, and more.

The i—Safety project was the impetus for creating the video game used in
this project. Research is indicating a steady trend in adolescents of the lack of
knowledge about online safety. A comparative study of adolescent females in
the United States and New Zealand found that a signiﬁcant percentage of girls
were engaging in risky behavior online, including sharing personal information;
results indicated a lack of education about safety online (Berson 8 Berson,
2005). The UK Children Go Online project determined that only two in ﬁve
adolescents (40%) said they could ﬁx a problem with their computer, and less
than one in ﬁve (18%) can set up a ﬁlter on their computer or remove a virus

(Livingstone, Bober, 8 Helsper, 2005). Four-ﬁfths of those ages 10-17 are

16

concerned about their privacy online, but most are also willing to yield sensitive
information if sufﬁciently incented (T urow 8 Nir, 2000). In a sample of high
school students over 60% failed to regularly look for privacy policies, 58% did not
clear their browser history, 37.7% failed to check if online forms are secure and
50% did not set their browsers to reject cookies (Quilliam, Rifon, LaRose. 8
Carlson, 2006). Therefore, this research created the stimulus game SafetyNet,
not only for the purposes of this research, but with the goal of educating children
as well in an attempt to combat these statistics.

The Case of Avatars: Extending SCT

Al-Natour, Benbasat, and Cenfetelli (2006) (as well as from Reeves and
Nass’ media equation (1996)) have shown that technological entities can be
perceived as being similar to ourselves, but the extent to which these similarities
affect individuals” behavior is unknown and is recommended by these
researchers as a future direction for study. Avatars in electronic environments
give users the opportunity to have experiences they are not able to have in real
life and to “try on” new identities (Allison, von Wahlde, Shockley, 8 Gabbard,
2006) . But how are Ieaming variables affected when we can actually change
ourselves?

SCT explains that people imitate characters they ﬁnd attractive. This is
part of the explanation for those that believe that violent video games cause
increased aggression (Sherry, 2001). But what happens when the model is an
electronic representation of one’s self in a game? SCT has much to say about

modeling, but the case of an avatar is a new horizon in SCT research: Avatars

17

give individuals the opportunity to see themselves externally. While the
experience is still enactive Ieaming, because the individual is controlling the
character, the individual also has the ability to view an external being and
transfer the modeled game behavior into his life external to the game.

While literally any video game has the potential to deliver an educational
message, this research is focused on avatar-based games. These games require
players to take on the persona of a game character, and to create, customize,
personalize, and play as the game character. Usually, an educational message
is integrated in the game’s narrative and learned by vicariously experiencing
game outcomes as a result of a virtual character's actions.

As an example, the role-playing and simulation game “The Sims”
(http:/fthesims.ea.coml) allows players to create virtual characters and
manipulate those characters, like a virtual dollhouse. The goal of the game play
is “survival” in a quotidian environment and rules similar to real life apply. Players
(mainly children for this speciﬁc game) are responsible for telling their “Sim" (the
virtual character) when to eat, when to sleep, when to go to work, how to have
fun, and how to interact with other Sims. If rules are violated (eg. a Sim does not
go to work or does not eat), the virtual character has to suffer the negative
consequences (i.e. runs out of money, becomes sick, etc.). If rules are not
violated, the virtual characters prosper and become rewarded (i.e. make new
friends, receive more money and resources, etc.) By observing rewards and

punishments as a result of the virtual character's actions, children vicariously

18

learn basic life skills by playing “The Sims”. The Ieaming experience is an
essential part of the game design.

“Making History,” a game developed by Muzzy Lane (http://wwwmaking-
historycoml), is a historical role-playing game that asks users to assume the role
of world leaders during World War II (“Making History”, 2007). Players can
choose from the United Kingdom, France, Germany, Italy, Japan, the USSR, the
United States, or China (though a new patch for the game now allows them to
play as the leader of any nation of the time), and integrated into the role-playing
are characteristics of each nation, including economic and military strengths,
diplomatic relations, ideology, and technical advancement. In addition to
teaching players about history, it also offers them the chance to “redo” history
with six separate scenarios to play through. Players get to take on historic roles,
experience new perspectives, and ultimately understand the critical thought
processes behind decision-making as opposed to just Ieaming the facts of World
War II.

Most often, research on video games and avatars have studied the use of
the avatar not as the individual Ieamer but as an electronic instructor of sorts.
However, Lee, Nass, Brave, Morishima, Nakajima, and Yamada (2007) studied
the effect of a co-leamer in an electronic environment. Many children's television
programming such as Sesame Street and Blue's Clues present on-screen
characters not only as instructors but as co-leamers, Ieaming the information
along with the children while at the same time passing on the information. Kim,

Wei, Xu, Ko, and Ilieva found that girls who learned with a pedagogical agent

19

increased in self-efﬁcacy and developed a more positive attitude about the
subject matter (2007), and that pedagogical agents can effectively motivate
children to learn (Kim 8 Baylor, 2007; Kim, Baylor, 8 Shen, 2007). Clearly, the
co-leamer is an external person. In the case of video game avatars, however,
things change because the co-Ieamerﬁnstructor is the external representation of
the player and is controlled by the player.

Of course, Ieaming of a game’s message is more likely if players care
about their character; rewards and punishments that a virtual character has to
experience are meaningful for a player as the player is ultimately rewarded or
punished via the game outcome (winning or losing). Because a game with an
avatar offers this secondary model, it is possible that individuals will retain more
information playing an educational game with an avatar as opposed to playing a
game without an avatar. As well, avatars have been studied with regard to
persuasive communication in the advertising realm (Wood, Solomon, 8 Englis,
2005). Consumers preferred the idealized images of avatars to realistic ones.
Additionally, if an individual disliked the avatar, that dislike extended to the
product. We can relate this to the video game arena and hypothesize that if an
individual dislikes the avatar, then that dislike extends to the game content.
Conversely, if an individual likes the avatar, this like would extend to the game
content and create an environment in the user’s own mind more conducive to
involvement and Ieaming.

The avatar may not even have to physically resemble the individual. Just

believing that the avatar is a representation of self may be enough to reach

20

adequate levels for identiﬁcation (Bailenson, Blascovich, and Guadagno, 2008).
Research by Fox and Bailenson found that participants exercised more,
regardless of whether feedback was positive (reward) or negative (punishment),
when they used a virtual representation of self rather than a virtual representation
of an other (2009). They also found that, when participants viewed a virtual self
running on a treadmill, a virtual other running, or a virtual self loitering, those who
viewed virtual selves running exercised signiﬁcantly more 24 hours later than
those who viewed themselves loitering, and those that viewed an other running.
The use of self as a model in an electronic environment appears to motivate the
individual more than the use of an other as a model. Reward and punishment
had no differences in effect. All that mattered was that the avatar represented
the self.

From here, the question then becomes in a virtual game setting, is it an
idealized version or a realistic version of self that has the bigger effects? The
more an individual perceives a model as being similar to himself, the more likely
learning is to occur (Bandura, 1977, 2001).

The potential exists for self-efﬁcacy and involvement (and subsequently
Ieaming) to increase when an individual sees his actual self being successful, or
for these to decrease if an individual uses his ideal self, potentially viewed as an
unattainable self and thus creating a gap that may lead the individual to think that
although the ideal version of himself can be successful, the real version cannot.
Prior research has determined that a match between a celebrity and a

consumer's ideal self leads to a more positive response to persuasive messages

21

(Choi 8 Rifon, in press). However, even when people are not aware of it, seeing
their actual selves to a degree in another’s image increases positive response
(Bailenson, Iyengar, Yee 8 Collins, 2009). But which is the better self, ideal or
real? Even before examining this, the validity of the similarity assumption must
be examined in this type of digital environment. Do individuals still respond most
favorably to what they see as similar to themselves, or does playing as someone
else still give the same results? How do people react to the external versions of
themselves?

Wood, Solomon, and Englis (2005) studied the avatars given on an online
shopping website that could be personalized to serve as models for the clothing.
Participants in the research were instructed to measure their bodies and then
create a representation of themselves using an online shopping website.
However, participants did not feel that these models were important to their
decision-making. Authors offer dissatisfaction with the projected image due to
poor self-concept or dissatisfaction with one’s own body as reasons for this
result. Perhaps if, rather than instructed to make their real selves participants
had been allowed to make idealized versions of themselves, results may have
been different. In other words, being faced with their actual selves may have
skewed the results negatively.

Part of the attraction of playing a game with a customizable avatar is that
users can create idealized versions of themselves and act through that idealized

version in the video game. Bessiere, Seay, and Kiesler (2007) found that people

22

prefer to make idealized avatars, suggesting that despite their circumstances in .
their real lives, they can still experience success virtually.

If the use of an idealized avatar increases enjoyment (as opposed to the
use of a real-self avatar), then it might also increase attention paid to the game
which includes the information contained within the game (cf. Singhal 8 Rogers,
2002). Perceived similarity increases message effectiveness (Andsager,
Bemker, Choi, 8 TonrveI, 2006). But there has been no research to date on
whether or not the perceived similarity must be between the actual self and some
external other. Perhaps similarity between the ideal self and some external other
is what makes the difference.

If enjoyment increases attention which can also potentially affect retention,
then what is more enjoyable, playing as your ideal self, or playing as your real
self? Do these conditions have an effect on self-efﬁcacy, involvement, and
retention as well? Does similarity to oneself still even hold in the digital world?
Social cognitive theory posits that identiﬁcation with a subject should lead to
greater learning (Bandura, 2001; Weber, Ritterfeld, 8 Kostygina, 2006), but does
not specify whether this refers to seeing similarity with your real self or your ideal
self. Research suggests that the most effective models should be similar but
slightly better, with intelligence, socioeconomic success, and competence
(Bandura, 1969), similar social status (Miller 8 Dollard, 1941), and the power to
reward (Bandura, Ross, 8 Ross, 1963). This seems to point to the creation of an
idealized version of oneself: similar in status but with more power and

intelligence.

23

Research on celebrity models indicates that relevance (how similar the
individual sees the celebrity to himself or herself) of the external celebrity to the
individual increases inspiration (Lockwood 8 Kunda, 1997), and people are most
likely to compare themselves to another when that other resembles them in
features, structure, and purpose (Holyoak 8 Thagard, 1995; Markman 8 Gentner,
1993). Inspiration may be thought of as motivation to succeed, and if we
conceptualize it thusly, we can see again how drawing similarities between
oneself and another can lead to greater motivation as well as greater enjoyment.
If we prefer using idealized avatars that resemble our ideal selves, perhaps it is
this similarity that matters.

It is possible that, for educational games regarding a topic about which the
player is not very involved, if the player has a connection to the source (i.e. the
game character), then the message (the educational portion of the game) may
get through regardless of interest. Given the propensity of individuals to choose
avatars reﬂecting idealized versions of themselves and SCT’s propositions that
similarity increases attention and modeling behavior, it is likely that the ideal self
that will have the greatest impact on the outcome variables, followed by the real
self, and ﬁnally, a third-party avatar that is not the individual will have the
smallest impact, as identiﬁcation with that avatar will be lower.

H4: The ideal-self avatar will be the most effective in terms of knowledge

retained, involvement, enjoyment, and self-efﬁcacy, followed by the real-

self avatar and the third-party avatar.

24

Chapter 3: Methodology

Sample

For this research population, a total of 314 subjects was obtained through
two waves of random sampling of students attending summer sessions I and II at
a large Midwestern university. Email addresses were randomly selected by the
Ofﬁce of the Registrar and provided in an electronic spreadsheet. After the
random samples were obtained, subjects were then contacted via email
regarding participation. The ﬁrst email notiﬁed subjects that they had been
selected and provided them with a consent form as well as a link to a website
that housed the game as well as the survey.

Of the total 314, 180 were female (57.3%), 124 were male (39.5%), and
10 did not indicate a response (3.2%). Ages ranged between 18 years and 48
years, with the greatest percentage being 21 years old (23.6%) and a median
age of 22 years. Two hundred forty-three respondents indicated that they were
white (77.4%), and 76 respondents indicated that they were of Hispanic descent
(24.2%). One hundred forty-six respondents were in their senior year (46.5%),
76 were juniors (23.6%), and 59 were graduate students (18.8%). Eighty-two
point two percent of respondents were PC users, and only 14.6% were
Macintosh users.

Participants were also asked to indicate with which type of avatar they
would most prefer to play a game. The majority of subjects responded that they
would prefer to play with an avatar of their real selves (39.2%). Twenty-three

point nine percent would prefer to play with a third-person character, and only

25

17.8% of respondents would prefer to play as a version of their ideal selves. The
rest of the respondents selected “Other,” chose more than one response, or did
not answer.

Measures

Variables to be measured were behavioral intentions, self-efﬁcacy,
knowledge retained, involvement, enjoyment of the game, enjoyment of learning,
enjoyment of the avatar, general demographic variables, and measures of the
character and the players’ real and ideal selves as well as the players’ friends in
the third-party condition in order to obtain the similarity measure.

Similarity (whether to ideal, actual, or other) was measured using a
distance analysis. The method for measuring this congruence is based on Choi
and Rifon’s measure for estimating ideal congruence (in press) which was in turn
based on other research using the absolute distance formula (Ericksen, 1996;
Graeff, 1996; Sirgy, 1985).

The distance formula is as follows:

n

2‘. l N] - II] |

i= 1

Aij = avatar image (i) of individual (i)
II] = ideal self-image (i) of individual (i)

For example, the similarity of ideal self to the avatar was measured by
asking participants to rate how they viewed the avatar and how they viewed their
ideal selves. The smaller the distance was, the greater the match. This also

provides a manipulation check to ensure that representations are in fact

26

participants” ideals. Using this formula, it is possible to determine congruence
between ideal self and avatar, actual self and avatar, and friend and avatar. The
frfteen semantic-differential items for rating avatar, ideal self, real self, and friend
(on a seven-point scale) are as follows:

. Comfortable/Uncomfortable

. Pleasant/Unpleasant

. Sophisticated/Naive

o Wlse/Stupid

. Socially Responsible/Irresponsible

. StrongANeak

o Conﬁdent/Apprehensive

o Enthusiastic/Not Enthusiastic

o Determined/Undetermined

c Rugged/Delicate

o ExcitablelCalm

. Dominating/Submissive

o Masculine/Feminine

. Public/Private

. Bold/Shy

Knowledge retained was measured using ten objective questions based

on the information contained in the game. Two incorrect answers and one
correct answer were provided for options. Correct answers were later coded with

a 1. Incorrect answers were coded with a 0. Questions were:

27

What is phishing?

Which URL is indicative of a phishing attempt?

Which of the following is a type of phishing attack?

What should you do if you get an email from a company requesting that
you call them?

What is a cookie?

What does a third—party cookie do?

What does “.exe” stand for?

How should you close pop-up windows?

Where do you go to set your system to update automatically?
Turning on your pop-up blocker will do what?

Questions were pretested to determine difﬁculty.

Self-efﬁcacy was measured using items created based on Bandura’s 2006

guide for constructing self-efﬁcacy scales. Subjects are asked to rate how much

they agree with each of the following statements. The scale is a 7-point Likert—

type scale, with 1 being “Strongly Disagree” and 7 being “Strongly Agree”. Items

are as follows:

I am conﬁdent that I can recognize a phishing attack.

I am conﬁdent I can protect myself from phishing attacks.

I am conﬁdent I can close pop-up windows the correct way.

I am conﬁdent I can reject cookies based on my own preferences.
I am conﬁdent I can set my browser to block pop-ups.

I am conﬁdent I can set my computer to update automatically.

28

o I am conﬁdent I can keep myself safe from spyware.

Enjoyment is measured by asking participants to rate how much they
enjoyed playing the game. Again, the scale is a seven-point Likert-type scale,
with 1 being “not at all” and 7 being “a great deal”. Items instruct players to rate
how much they enjoyed playing the game overall, how much they enjoyed
Ieaming the information, and how much they enjoyed being able to play with the
avatar.

Involvement is operationalized and measured using a measurement
based on a scale developed by Ohanian (1989). The 5-item semantic differential
scale items (on a 7-point scale) are listed below. Participants rated “online
safety“ using these 5 items.

. Matters to melDoesn’t matter to me

. Of concern to melOf no concern to me
o Relevant/Irrelevant

. Important/Not important

. Interesting/Not interesting

Time on task was a simple measure of how long it took players to
complete the game, measured using information recorded by the online survey
application used, which tracked both the time of day that participants started the
game and the time of day that they ﬁnished. Intention to enact behaviors was
based on questions previously created for a separate research project (LaRose 8

Rifon, 2007) asking participants to rate, on a 7-point scale from Very Unlikely to

29

Very Likely, how likely it is that they will enact the behaviors learned in the game.
The items are listed below.

. Set my computer to automatically update.

. Change the settings for cookies on my computer.

. Change the settings for popups on my computer.

. Discuss phishing with friends or family.

. Discuss cookies with friends or family.

. Discuss computer settings with friends or family.

. Check URLs in email for phishing attempts before clicking on them.

0 Look up phone numbers given in emails to check for phishing attempts.

. Search Google to verify identities of email senders.

Demographic variables were also recorded from the subjects (age, race,
SES, and gender) to determine if any of these variables had an impact on any of
the predicted relationships.
Scale Reliabilities
Reliabilities were calculated where applicable for the scales used:

character rating, ideal self rating, real self rating, friend rating, self-efﬁcacy,
behavioral intentions, and involvement. The character rating scale resulted in a
Cronbach’s alpha of .87 with a mean of 14.77 and a standard deviation of 11.44.
The ideal self scale had an alpha of .80, mean of 24.59, standard deviation of
8.59. The real self scale had an alpha of .86, mean of 12.45 and standard
deviation of 11.25. The friend scale yielded an alpha of .84 with a mean of 16.04

and a standard deviation of 10.82. Self-efficacy yielded an alpha of .95 with a

30

mean of 42.60 and a standard deviation of 8.01, behavioral intentions resulted in
an alpha of .88 with a mean of 41 .90 and standard deviation of 12.43, and the
alpha for involvement was .85 with a mean of 10.11 and standard deviation of
4.62. Individual items as well as correlation matrices are reported in Appendix A.
The ten knowledge questions resulted in a mean of 7.92, SD = 1.38, and a
Kuder—Richardson value of .50.

The Stimulus Instrument

To cany out this research, it was necessary to use a computer game that
would teach and educate users on testable subject matter that they were not
likely to know and would allow for some kind of identiﬁcation with the avatar.
Unfortunately, this type of game does not exist. Games that allow customization
of avatars are typically not educational video games, and educational video
games are usually targeted toward younger children (grade-school age) or do not
contain the necessary components of avatar inclusion. It was necessary, then, to
create a game from scratch. Because of the timeliness of the i-Safety project
and the prior testing of the effectiveness of the information contained in the
curriculum, the game was based on this information.

The game engine uses a simple combination of Adobe Flash interactions
as well as ActionScript 2, a programming language speciﬁcally for Adobe Flash
CS4 Professional, the program in which the entire video game was created
(Adobe, 2008). All portions of the game were designed and implemented by the
author, allowing for complete control over the game and the information

contained within, as well as the ability to track game players’ actions.

31

To use a game to instruct, the game must be compelling. Creating a
compelling game can involve a number of different things, beginning with a
storyline to draw users in (Green, Brook, 8 Kaufman, 2004). Allowing users to
put themselves in the game character’s shoes (a ﬁrst-person perspective) also
makes a game more compelling because it brings the game closer to home for a
player (Dickey, 2005; Riddle, 2002). Therefore, SafetyNet begins by informing
players that they have just been given a new job: a security guard for their
family’s computer. However, before they ofﬁcially earn their job title of Online
Security Expert and become qualiﬁed to protect the family computer from outside
attacks, they must complete several levels of training in the Cyber Security
Training Course to earn their badge. Players are informed that they must pass
each level (of which there are a total of three) in order to receive their ofﬁcial
security badge. They are told that there will be hints along the way.

The game environment is a 2D representation of the computer desktop.
Entering and exiting various portions of the computer, using a browser,
accessing the Internet, and other navigational activities are done by pointing and
clicking the mouse in much the same way that a user would interact with his or
her computer. Each level of the game corresponds to a different “threat zone“.
Players begin with the more simple tasks and move on to more difﬁcult tasks.
External threats to safety in the form of spyware, phishing attacks, cookies, etc.,
are represented with avatars as well. In order to battle these threats, players

must have knowledge of how to be safe online.

32

The ﬁrst level deals solely with email safety, speciﬁcally phishing attacks.
A phishing attack occurs when an outside source attempts to inﬁltrate an
individual’s private information by impersonating an ofﬁcial source in an email
and requesting that an individual give out information. As an example, the email
might appear to come from a bank, requesting that the individual click on a link in
the email and, once at the new URL, input information such as names,
addresses, passwords, and ﬁnancial information to “update records,” “verify
information,” or other phony (though reasonable-sounding) reasons. However,
this new URL is not afﬁliated with the bank, but has been set up by identity
thieves, and each person who ﬁlls out the information has now left themselves at
risk for identity theft.

Players are taught to recognize phishing attempts using several different
methods of veriﬁcation. They are also advised about the actions they should
take if they do receive a phishing email. Four sample emails are given, and
players must determine if they are phishing attempts.

The second level deals with browser safety. Once they have mastered
email safety, they are then instructed to launch their browser. This level contains
potential threats such as pop-up windows (that may contain spyware when
clicked on), cookies, and instant messages. Players Ieam how to be suspicious
of blindly accepting things, and how to spot problems before they occur.

The ﬁnal level deals with system settings. Setting a pop-up blocker,
determining how their browser will handle those cookies, and setting their system

to automatically update are taught in this level. Although operating systems and

33

browsers differ, the game is set up in as generic a way as possible so that
information gained from the game may be transferable, regardless of browser or
OS in use at the player’s home.

When a decision is made and action taken in the game, either a dialogue
box or the game character will appear on the screen to explain either (a) why
what the player did was the safe thing to do, or (b) why what the player did was
the unsafe thing to do and, more importantly, what the appropriate course of
action would be. Thus, players might experience a set-back, but they are not left
without the proper safety information. Once players have successfully completed
all three levels, they are awarded their Security Expert badge and have won the
game.

The Avatar Manipulation

The avatar manipulation was a textual manipulation shown to players prior
to entering the video game site. The players were asked to imagine that the
avatar they were presented with was one of three things: their ideal self, their real
self, or a friend of theirs. The avatar used was a simple two—dimensional ﬁgure
with human body and limbs and the head of a dog. Six different avatar heads
were pretested to determine which was most effective. An animal was used to
eliminate cultural biases as well as the potential for closer matching in
appearance to some players and not to others that use of a human avatar would
have meant. The author was responsible for all graphics, programming, etc. in
order to keep the project at a reasonable cost as well as to ensure that the game

could be created to very exact speciﬁcations, and if changes needed to be made,

those changes could be easily implemented. SafetyNet was hosted on the i-
Safety server, located in the Communication Arts and Sciences building at
Michigan State University. Mean distances and reliabilities are reported in Table
1.

Procedure

Subjects were randomly assigned to one of the four treatment conditions.

An email containing a link to the website housing the game as well as the survey
was provided to subjects in an email. Subjects were asked to provide an
assigned log-in name for tracking purposes. After entering the log-in
Table 1: Mean Distance Scores and Scale Reliabilities

 

 

'32:. gee?! Friend Avatar RecIiacatrlﬁﬂy

Ideal Self .00 16.27 18.10 19.30 .80

Real Self 16.27 .00 17.68 18.92 .86
Friend 18.10 17.68 .00 17.40 .84
Avatar 19.30 18.92 17.40 .00 .87

h

 

name, the survey application, through basic PHP programming, randomly sent
each subject to a webpage for one of the four conditions that contained
i|f1$tructions regarding how the individual was to play the game. Those in the
e"active condition were not asked to imagine that they were playing as anyone
bUt rather were presented with an avatar-free game. Those in the observational
Corldition were playing as a character and were instructed to either imagine that

this character was their ideal self, their real self, or a friend of theirs. After

35

completing the game, players were then asked to ﬁll out an online questionnaire
during which the dependent measures as well as the measures for similarity
ratings were taken. Random assignment of individuals accounted for a number
of confounds that might othenrvise be present, such as base knowledge, general
educational ability, maturation, and so forth, as well as self-efficacy and
involvement to participate prior to beginning the project. Using the subject area of
online safety education not only taught subjects useful material, but the subject
matter lent itself well to using electronic teaching methods, being electronically-
based itself.

Effect sizes in video game studies regarding video games as educational
tools have yielded effect sizes of small to medium. Meta-analysis of high-quality
studies of the efﬁcacy of education software resulted In an effect size of .38
(Gentile 8 Anderson, 2006). Power calculation using the G*Power software
program (G*Power, 2008) suggested for an alpha of .05 (power of .95), an effect
size of .25, and four treatment conditions, 280 subjects should be used. This

research resulted in a total of 314 subjects across the four treatment conditions.

36

Chapter 4: Results

Manipulation Check

The manipulation check instituted in this project was to determine that the
textual manipulation for assigning conditions of real self, ideal self, and friend
(the third-person condition) functioned properly. To check that the manipulation
functioned correctly, respondents were asked to respond to items about how they
viewed the avatar, and then to respond to those same items regarding how they
view their real selves and their ideal selves. Respondents in the friend condition
were also asked to rate the friend they chose. The distance formula below

(which uses the ideal condition as an example) was used to calculate a measure

for each category.
n
2: l Ar'j- II] |
i= 1

AI] = avatar image (I) of individual (i)

If] = ideal self-image (r) of individual (I)

By taking the sum of the absolute values of the differences between each
item for two conditions, a measure may be calculated. In this case, the
difference between each item for the individual’s ideal self image, as well as the
avatar image (the way the individual viewed the avatar), is taken and added
together. Next, a measure would be calculated for the avatar image and the
individual’s real self image, and in conditions where respondents were assigned
to imagine the avatar as a friend of theirs, a measure would be calculated for
avatar image and friend image as well. These numbers may then be compared

to each other. The smallest distance should be the distance between avatar
37

image and the image to which the respondent was assigned. For example, if
respondent A were asked to imagine that the avatar was respondent A’s friend,
then the sum of distances between A’s friend image measure and avatar image
measure would be smaller than the sum between A’s ideal self image measure
and avatar image measure, as well as A’s real self image measure and avatar
image measure. As would be expected, the distance was the smallest between
real and ideal selves.

The textual manipulation was successful in 184 cases (41 cases of the 80
in the ideal group, 41 of 77 in the real group, and 28 of 77 in the friend group, as
well as the 74 cases in the enactive group which did not contain an avatar).
Success was determined by examining distance measures between
standardized scores (calculated by subtracting the overall mean score from each
individual’s score and dividing by the standard deviation) for ideal self and
character, real self and character, and friend and character. If the standardized
distance was smallest in the pair that was the assigned condition, the
manipulation was deemed successful.

A correlation of successful versus unsuccessftrl manipulations with the
various dependent variables yielded no signiﬁcant results, indicating that success
of manipulation was not the cause for results found.

Statistical Analysis
All statistical analyses were conducted with SPSS Statistics 17 (SPSS,

2009). The four main hypotheses to be tested were:

38

0 H1: Observational Ieaming will lead to more knowledge retained, more

enjoyment, and more involvement with the topic than enactive Ieaming.

0 H2: Enactive Ieaming will lead to higher levels of self-efﬁcacy than

observational Ieaming.

. H3: Enactive Ieaming will result in greater intentions to perform safe online

behavior than observational Ieaming.

. H4: The ideal-self avatar will be the most effective in terms of knowledge

retained, involvement, enjoyment, and self-efﬁcacy, followed by the real-

self avatar and the third-party avatar.

Hypothesis 1. To compare the two groups, enactive and observational

Ieaming, a multivariate analysis of variance using planned contrasts was

conducted to check for differences between each group. Means, standard

deviations, and cell numbers are reported in Table 2. Between-subjects effects

are reported in Table 3, and planned contrasts are reported in Table 4.

Table 2: Means, Standard Deviations, and Cell Sizes for Enactive Leaming, Ideal

Self, Real Self, and Friend

 

 

. . . Standard

Dependent Vanable Condrtron N Mean Deviation
Knowledge Enactive 74 8.30 1 .28
(total correct, 10-item Ideal 80 7.81 1.35
scale) Real 75 7.75 1 .44
Friend 77 7.84 1.40
Self-Efﬁcacy Enactive 73 .80 .07
(scale of 1 to 7, 7-item Ideal 77 .77 .15
scale, transformed) Real 75 .76 .18
Friend 77 .75 .17
Behavioral Intentions Enactive 73 4.81 1.30

39

Table 2 (cont’d)

(scale of 1 to 7, 9—item
Scale)

Involvement
(scale from -3 to +3, 5-item
scale)

Enjoyment of Game
(from -3 to +3)

Enjoyment of Learning
(from -3 to +3)

Enjoyment of the Avatar
(from -3 to +3)

Ideal
Real
Fnend

Enactive
Ideal
Real

Friend

Enactive
Ideal
Real

Fﬁend

Enactive
Ideal
Real

Friend

Enactive
Ideal
Real

Fnend

79
75
77

71
79
75
77

73
79
75
77

73
79
75
77

nla
79
75
77

4.66
4.60
4.57

2.15
2.07
1.88
1.99

1.33
1.13
1.05
1.18

1.51
1.37
1.27
1.17

nla
.62
.68
1 .08

1.45
1.26
1.50

.83

.86

.97

1.02
1.31
1.23

1 .47
1 .47

1.19
1.13
1.36
1.47

nla

1 .47
1.69
1.65

 

A multivariate analysis of variance was conducted using the General
Linear Model menu in SPSS with dependent variables of enjoyment of Ieaming,
enjoyment of game, behavioral intentions, total number of correct answers, and
self-efﬁcacy, involvement as a covariate (as involvement is likely to remain
constant), and condition as the independent variable. Results are reported in
Table 3.

Table 3: F- Tests for Dependent Variables and Condition

 

 

Type "I Partial
3.... mar.“ .. .. F E. ms:
Swares Squared
Corrected Behavioral
Model Intentions 25"” 3 4'64 .00 ~05 -89

40

Table 3 (cont’d)

 

 

 

 

ng'ﬁg: 7.76 1.34 .26 .02 .36
Enjoyment of

Game 93.02 20.10 .00 .21 1.00
Enjoyment of

Learning 79.16 18.36 .00 .20 1.00
Self-Efﬁcacy .06 .75 .00 .01 .21
Behavioral 357.4

Intercept Intentions 657.92 4 .00 .62 1.00
Knowledge 1222.

Gained 2352.90 48 .00 .85 1.00
E"’°Ym§2:n‘g 1.59 1.03 .31 .01 .17
E"’°{;":r:tirg .13 .09 .77 .00 .06
Self-Efﬁcacy 26.56 977'? .00 .81 1.00

lnvolve- Behavioral

men, Intentions 24.83 13.49 .00 .06 .96
ng'ﬁg: 7.62 4.00 .05 .02 .51
Enjoyment of

Game 92.48 59.92 .00 .21 1.00
Enjoyment of

Learning 77.68 54.04 .00 .19 1.00
Self-Efﬁcacy .05 1.71 .19 .01 .26

Group 3:333: .42 .12 .89 .00 .07
ng'ﬁ'gj .11 .03 .97 .00 .05
E"j°ym§2:n‘: .20 .06 .94 .00 .06
Enhgaimg 1.46 .51 .60 .00 .13
Self-Efﬁcacy .02 .33 .72 .00 .10
Behavioral

Error Intentions 412.31 224
Knowledge

Gained 431.13 224

41

Table 3 (cont’d)

Enjoyment of

 

 

Game 34516 224
Enjogggtin‘g 321.98 224
Self-Efﬁcacy 6.08 224
Knoglaeigge: 14351.8 228
Enioymggg’g 733.00 228
Self-Efﬁcacy 137.72 228

““111
Knoglﬁg: 438.89 227
Enlomggtm‘: 438.79 227
Eniogaﬁrg 401.14 227
Self-Efﬁcacy 6.15 227

 

Although type of condition did not result in any signiﬁcant ﬁndings (though
involvement had a signiﬁcant effect on all dependent variables except self-
efﬁcacy), a planned contrast between observational Ieaming and enactive
Ieaming was still examined. According to Kwon (1996), planned comparisons
can offer more power than traditional MANOVA followed by post hoc tests,
guarding against Type II error, and planned comparisons force the researcher to
think more critically about research design. Planned comparisons can also

detect signiﬁcant differences between groups of interest that a classical omnibus

42

test would not. Therefore, a planned comparison between the enactive condition
and the observational conditions was examined. The planned comparison
involved assigning the enactive condition a contrast coefﬁcient of 3, and the three
observational conditions a -1 for a total of 0 when all conditions were examined
together. Because this research had predetermined directional hypotheses, one-
tailed test results for the contrasts were use. Consideration was given to the
exploratory nature of this research, and in an effort to reduce Type II error (failure
to reject the null hypothesis when the null hypothesis is false), the acceptable
alpha level was raised to .10.

All signiﬁcance values for the planned comparison are reported in Table 4,
whether or not they reached signiﬁcance in the expected direction.

Table 4: Significance Values for Planned Comparisons Between Enactive and

 

 

Observational Conditions

Dependent Value of t df Signiﬁcance

Vanable Contrast (one-way)

Knowledge Gained 1.49 2.72 302 1.00
Self-Efﬁcacy .10 1.64 298 .05
Involvement .51 1 .35 298 .91
Future Behavior .59 1.06 300 .15
Enjoyment of Game . .62 1.13 300 .86
Enjoyment of Learning .72 1.38 300 .92

 

43

CU

ha

SE

hlg

ob

Although it was hypothesized that knowledge, enjoyment, and involvement
would be higher in the observational condition, none of these reached
signiﬁcance. Therefore, hypothesis 1 was not supported.

Hypothesis 2. Hypothesis 2 indicated that self-efﬁcacy would be higher for
individuals in the enactive condition than for those in the observational condition.
Again, to compare the two groups, a planned contrast was utilized. The self-
efﬁcecy scale was not normally distributed and a transformation of log10 was
made to the data. Self-efﬁcacy was found to be higher in the enactive condition
than in the observational condition. Effect size for self-efﬁcacy using Cohen’s d
was .24, a small effect size using the standard convention for Cohen’s d where .2
is a small effect size, .5 is a medium effect size, and .8 is a large effect size.

Hypothesis 2, that those in the enactive Ieaming condition would have
higher levels of self-efﬁcacy regarding online safety behaviors than those in the
observational Ieaming condition, was therefore supported.

Hypothesis 3. Hypothesis 3 posited that individuals in the enactive
Ieaming condition would report greater intentions of performing safe online
behaviors that they had learned in the video game than would those in the
observational condition. The multivariate analysis of variance did not result in a
signiﬁcant difference between groups on the behaviors overall, although the
means were in the hypothesized direction. The mean for behavioral intentions of
the enactive condition was 4.81 (SD = 1.30) and the means of the observational
conditions were 4.66 (SD = 1.45) for ideal, 4.60 (SD = 1.26) for real, and 4.57

(SD = 1.50) for friend.

Hypothesis 4. Hypothesis 4 stated that the ideal-self avatar will be the
most effective in terms of knowledge retained, involvement, enjoyment, and self-
efﬁcacy, followed by the real-self avatar and the third-party avatar. Again,
planned comparisons were used to detect speciﬁc group differences between
ideal, real, and friend conditions. For each possible pair of comparisons, the two
conditions being examined were assigned contrast coefﬁcients of 1 and -1, while
the condition that was not being included in the comparison was assigned a 0.

Despite the numerous opportunities for signiﬁcant results, the planned
contrasts did not show signiﬁcant differences between the three groups for any of
the outcome variables.

Table 5: Planned Comparisons for Ideal, Real, and Friend Conditions

 

Dependent Contrast Value of df

 

Variable Contrast Sig.
Knowledge
Gained Ideal-Real .07 .29 229 .77
Ideal-Friend -.03 -. 14 229 .89
Real-Friend -.10 -.43 229 .67
Self-Efficacy Ideal-Real .01 .53 226 .60
Ideal-Friend .02 .69 226 .49
Real-Friend .00 .16 226 .88
Involvement Ideal-Real .18 1.19 228 .23
Ideal-Friend .07 .48 228 .62
Real-Friend -.1 1 -.71 228 .48
Future Behavior Ideal-Real .05 .24 228 .81
Ideal-Friend .09 .38 228 .72
Real-Friend .03 .14 228 .89
Enjoyment of
Game Ideal-Real .07 .33 228 .74
Ideal-Friend -.06 -.25 228 .81
Real-Friend -. 1 3 -.60 228 .57

45

dl:

Er
ga

Table 5 (cont’d)

Enjoyment of
Learning Ideal-Real .10 .47 228 .64
Ideal-Friend .20 .93 228 .35
Real-Friend .10 .45 228 .65

Enjoyment of
Avatar Ideal-Real -.06 -.23 228 .82
Ideal-Friend -.46 -1.78 228 .08
Real-Friend -.40 -1.53 228 .13

 

Again, though, the means on several variables were heading in the proper
direction. For enjoyment of Ieaming, involvement, and behavioral intentions,
means were all highest in the ideal condition.

To further examine this hypothesis, the standardized distance measures
between avatar and other (ideal self, real self, and friend) were used as
continuous variables and were then correlated with the dependent variables in a
post hoc analysis. Correlations and Steiger's 2 values, based on the test
designed by Meng, Rosenthal, and Rubin (1992) for testing the signiﬁcance of
correlations between correlated variables, are reported below in Table 6 between
the ideal and real conditions, as standardized scores for real and friend
conditions were nearly identical and resulted in the same correlation values.

Table 6: Correlations Between Distances and Dependent Variables

 

 

”*Steiger's
Ideal-Avatar Real-Avatar Friend-Avatar Z value
(Ideal-Real)
Enj°yment °f **-.41 (p < .01) "-33 (p < .01) **-.33 (p < .01) 1.39

game

46

Table 6 (cont’d)

Enjoyment of ..._ .._ .*_

Ieaming .31 (p < .01) .24 (p < .01) .24 (p < .01) 1.11

Enjoyment of ,,,_ ._ = ._ =

avatar .34 (p < .01) .15 (p .02) .15 (p .02) 3.08

Involvement **-.28 (p < .01) **-.20 (p < .01) **-.20 (p < .01) _ 1.28

Knowledge Both

”tamed .12 (p = .06) .10 (p = .14) .10 (p = .14) °°"e'a"%';st
signiﬁcant

Self-efﬁcacy One

(t'a"3f°""ed) .13 (p = .64) *.03 (p = .64) *.03 (p = .06) ”mag;
signiﬁcant

Behavioral .. _ ._ = ._ =

intentions -.17 (p - .01) .16 (p .02) .16 (p .02) .29

 

*p is signiﬁcant at the .05 level
** p is signiﬁcant at the .01 level
”One-tailed Z-critical is 1.28 for p < .10 and 1.65 for p < .05

Examining all of these results, hypothesis 4 is therefore partially
supported. Enjoyment of the game, enjoyment of Ieaming, enjoyment of the
avatar, and involvement were all signiﬁcantly higher in the ideal condition than in
the real or friend conditions. However, knowledge retained and self-efﬁcacy
were not signiﬁcantly correlated with condition, and there was no difference
between friend and real conditions

Trme spent playing was not signiﬁcantly different between any of the
groups. As well, demographic factors did not have any signiﬁcant inﬂuences on

results.

47

Chapter 5: Discussion

In summary, enactive Ieaming was related to higher levels of self-efﬁcacy
than observational Ieaming, and avatar type (ideal self, real self, or friend)
showed the strongest correlations between ideal self and the Ieaming outcomes.
Although signiﬁcant differences were not found between involvement, behavioral
intentions, and enjoyment between observational and enactive Ieaming, the
means were higher for enactive Ieaming than for observational Ieaming,
indicating that there may be some support for these, as well. These results are
in line with Kolb’s Experiential Learning Theory while on the surface contradicting
Bandura’s Social Cognitive Theory.
Enactive Versus Observational Learning

Social Cognitive Theory posits that observational Ieaming should be the
stronger form of Ieaming. However, this research did not ﬁnd signiﬁcant
differences in knowledge gained. Generalizations cannot be broad, given that
this project only tested one Ieaming area, but it is possible that in the case of
video games, enactive Ieaming (Ieaming without a model) is the better way to
instruct. This is also contradictory to the idea of video game design in which
including customizable avatars, particularly in massively multiplayer online role-
playing games (or MMORPG’s), is one of the more important aspects of game
play (Yee, 2004). Respondents indicated that they most enjoyed the game play
using their ideal selves, but enjoyment means were highest in the enactive

condition without an avatar at all.

48

Perhaps, though observational Ieaming may be more effective in certain
cases (Bennett, Farrington, 8 Huesmann, 2005; Farrington, 1998; Farrington,
2001a; Farrington, 2001b; Wesch, Law, 8 Hall, 2007), when Ieaming activities
that do not have immediate, real-world, negative consequences (for instance,
when playing a video game about online safety, a mistake or error does not
result in a loss of privacy or data—an immediate negative consequence likely to
happen in real like—but instead, simply a loss of points in the game), method of
Ieaming does not matter. More likely, though, the lack of ﬁndings may be a case
of a one-shot experimental design lacking not statistical power but the power of
connection with a character. It is difﬁcult for gamers to form an instant bond with
a video game character; the relationship between player and connection to a
character strengthens as the player spends more time with the character (Lewis,
Weber, 8 Bowman, 2007). Given this, repeated game play or even extending
the game from a 15-minute version to a lengthier version with more levels may
allow connections to develop more signiﬁcantly between the individual and the
assigned condition.

The results of this research, however, could also be due to discrepancies
between Social Cognitive Theory’s assumptions and the necessary environment
of the video game. Though some of the aspects of SCT remained intact (self-
efﬁcacy being higher in the enactive condition which led to greater, though not
signiﬁcant, intentions to enact the behaviors learned), some did not. These

discrepancies must be examined in detail.

49

First, the results indicate that it did not matter whether the model was the
self or the other", it was still seen as being something other than the self. Even in
cases where participants were playing the game as their manipulated selves
(ideal or real) there were no differences between either of these conditions with
the friend condition. Social Cognitive Theory states that Ieaming is most likely to
occur if, ﬁrst, there is a close identiﬁcation between the learner and the model. In
this research, the Ieamer was the model, which would necessarily indicate an
extreme identiﬁcation. The manipulation check indicated the participants were
indeed playing as the type of individual they were instructed to use, ideal or real.
But modeling one’s self (with what can be assumed to be a very high level of
identiﬁcation) was not more effective than modeling an external other (where
identiﬁcation would be lower).

Secondly, Social Cognitive Theory posits that there are four steps in the
modeling process, beginning with attention, then retention (remembering what
has been seen), reproduction (being able to do the behavior that has just been
seen), and motivation (past, promised, and vicarious reinforcements or
punishments). The Entertainment-Education paradigm has a similar basis,
stating that entertainment (the video game) can serve as a door-opener to the
Ieaming (education) that will follow. Here, however, it appears that several of
these premises and steps do not hold true. Entertainment and attention should
be garnered by putting the information in a true game format, complete with
avatar. Motivation in the form of vicarious reinforcement or punishment should

have an effect on Ieaming, as well. But presenting the information about online

50

safety in a video game format with an avatar present to display vicarious
reinforcement/punishment was the less effective method. Presenting a more
traditional training simulation (the only difference being the absence of the avatar
as all other aspects of the game remained exactly the same) resulted in greater
self-efﬁcacy and more information learned. Perhaps the presence of the avatar
again distracted participants and unintentionally placed more focus on the fact
that it is the game players who control the character and form its intelligence,
leading to a feeling of non-transferable skills from avatar in a game to individual
in real life.

Furthermore, the mean of involvement was still higher in the enactive
condition than in the observational condition even though the difference did not
reach signiﬁcance. This means that the identiﬁcation between the player and the
avatar, while it was present, did not make the expected difference. Motivation
(involvement) should have been higher when there was identiﬁcation with the
avatar (the ideal and real selves) and lower in the friend and the avatar-free
condition, but it was not signiﬁcantly different. It could be argued that players felt
more identiﬁcation with no avatar than with any avatar at all, leading to higher
involvement. Perhaps this is because the outcomes in the game appeared to
more directly affect the game player. Participants were not watching a model of
themselves or a friend. but rather were engaged in a ﬁrst-person view of the
game environment. In the avatar condition, even though the only difference was
the presence of an avatar moving about the screen and “talking” to players (in

the no—avatar condition, the text merely appeared on the screen outside of a

51

speech bubble), the avatar still gave the impression that the consequences
would happen to someone other than the player. Regardless of the reason,
involvement was higher in the enactive condition, another surprising ﬁnding.
Research exists supporting the idea that enactive exploration increases intrinsic
motivation (Wood, Kakebeeke, Debowski 8 F reese, 2000) but both groups
compared in this study were exposed to an observational Ieaming component
prior to the experimental differences: exposure to an enactive exploration
(participants were encouraged to practice and explore a computer task on their
own prior to doing it) versus no exposure to an enactive condition.

This research supposes that there may be a third kind of learning,
something between pure enactive and pure observational: the Ieaming that
occurred when the self was the model. However, the conditions in which there
was a self model were not as successful in any of the Ieaming outcomes as the
enactive condition. Social Cognitive Theory states that observing a model with
which the individual identiﬁes will result in greater Ieaming, and individuals with a
great deal of self-efﬁcacy will Ieam more. If observing behaviors is the stronger
way to Ieam them, then the avatar conditions (ideal, real, and other) should have
provided the better results.

To explain why the assumptions of Social Cognitive Theory about
observational and enactive Ieaming may not hold here, the nature of video
games must be examined. Because of the natural set-up of the video game,
observational Ieaming is simply too enmeshed with enactive Ieaming. Enactive

Ieaming and observational Ieaming were, in fact, both present in the avatar

52

conditions, because subjects were viewing the avatar engaging in behavior
(observational) as well as Ieaming how to do the behavior themselves (enactive).
The new type of model, the self-model, was present in all conditions in this
research. Observational and enactive Ieaming were occurring at the same time,
regardless of avatar condition.

The difference between purely observational Ieaming and self-
observational Ieaming is the inherent, possibly even subconscious knowledge
that the character is in fact the individual. A game player knows that he is the
one that is ultimately guiding the character and forming the resultant actions. It
does not matter whether an avatar is present or not; it is always going to be
enactive Ieaming because the player is in control. True observational Ieaming
would require the absence of control of any kind over the avatar.

This leads to an argument of what may, at least at ﬁrst, appear to be the
reverse of the discussion that all game-playing is enactive, because it may be
argued that all game-playing is observational. Enactive Ieaming is Ieamingby
doing, Ieaming that takes place directly with reinforcements of rewards or
punishments. But a game player is still playing a role when interacting with a
video game, and necessarily experiencing those rewards or punishments
indirectly. For instance, when watching a character in a video game, the
character experiences the losses and gains of rewards (points, lives, artifacts
pertaining to the game) as well as the success of completing the mission. The
game player experiences these vicariously. When playing a game about World

War II, the player himself is not directly responsible for the outcome. He does

53

not directly injure or kill opponents. The game is a simulation. Similarly, when
playing a game about online safety, the player himself does not experience direct
gains or losses. His computer is not at greater risk should he not make the
correct choices in the game. This removal from direct consequences deﬁnes
observational Ieaming.

Enactive Ieaming means that there is not this removal from reality. A
player cannot put the blame on the avatar and must accept responsibility directly
for the results, which likely results in increased involvement. Video game playing
takes, essentially, the best of both words. Individuals are directly responsible but
the safety factor is still in place. Observational Ieaming may occur in a game if
the individual observes an opponent, either real (as in an MMORPG) or
computer-generated, facing consequences of decisions. But video games
cannot be purely enactive, either, because there is no one-to—one connection.
What one does in the game does not translate into real-world consequences.

Therefore, avatars in video games, whether visible as in a third-person
perspective or invisible as in the ﬁrst-person perspective comprise a new type of
model—tentatively labeled the automodel—combine elements of both. Although
this research attempted to separate the automodel from the other conditions in
video game research, it appears theoretically and practically impossible to create
a video game that excludes one or the other types of Ieaming. The game
created satisﬁed all the conditions for a video game—players, rules, and goals
(Smed 8 Hakonen, 2003) with interactive capabilities wherein the player’s input

has an effect on the game and even a brief opportunity for customization.

Consumers of more traditional media, such as ﬁlm, are passive consumers, but
video game players become a part of the narrative (Grodal, 2000). Eliminating
this input to create observational Ieaming, or eliminating indirect rewards,
punishments, rules, and goals to create enactive Ieaming both take the video
game and change it into something that is by deﬁnition no longer a video game.

Though it makes intuitive sense that experiencing two means of gaining
self-efﬁcacy would create the highest levels of self-efﬁcacy, even in cases where
the avatar is a third-party avatar, it is still controlled by the game player. If
players go into a game to learn, they are Ieaming as they go, but in modeling, the
model is expected to have more expertise than the Ieamers do. In a video game,
this isn’t true: the model (avatar) has whatever information the players bring to it.
In fact, Fox and Bailenson (2009) discuss this very issue: self models are limited
by player skill. Using a self model would mean that self-efﬁcacy is not as high
because individuals do not put as much faith into the observations. They know
the model’s level of expertise is equal to their own, even if it is only on a
subconscious level. Therefore, perhaps this dual source of self-efﬁcacy,
progressive mastery and observation of the avatar together, is the less effective
means of increasing self-efﬁcacy in video game instruction, because although the
mastery increases self-efﬁcacy, it is offset by the knowledge that the model and
the knowledge of the model is limited by the individual controlling the model.
Although players can observe consequences as they happen to other characters
in the game (either controlled by other individuals such as in an MMORPG or

controlled by the game drivers—a nonplayer character or NPC), purely

55

observational Ieaming regarding one’s own avatar is not possible and again, this
is what makes video games an auto-modeled experience.

As for the lack of signiﬁcant differences in behavioral intentions, there was
the possibility that the game did a better job of covering some behaviors than
others, which might have masked true differences. More likely, however, is that
the study did not contain enough participants to reach signiﬁcance. Although the
intentions as a whole were not signiﬁcant, means for each item in the scale were
higher in the enactive condition than the observational conditions, so perhaps
additional subjects might have revealed differences.

Ideal, Real, or Other Avatar

This research next attempted to answer the question of whether it is
idealized versions or realistic versions of self that have the bigger effects. The
correlation analysis showed that enjoyment of the game, enjoyment of Ieaming
the material, enjoyment of the avatar, and involvement were all higher when
players were playing as their ideal selves. There was not, however, the expected
differentiation between real self and the third party condition. In fact, the
correlations between the real self scores and the friend scores with the
dependent variables yielded the same values. This is likely a limitation of the
study design and will be discussed in Limitations below. However, the ideal
condition consistently indicated more enjoyment and greater involvement with the
material. Thus, even though observational Ieaming did not differ from enactive
Ieaming as deﬁned in this project, when considering only the avatar conditions,

Social Cognitive Theory was supported. Research on modeling in Social

56

Cognitive Theory suggests that for models to be the most effective, they should
ﬁrst be similar to the individual, but they should also be slightly improved versions
of the individual. They should be intelligent, have symbols of socioeconomic
success, and possess a high level of competence (Bandura, 1969). At the same
time, individuals are more likely to imitate models that are in adjacent positions in
the social hierarchy rather than positions that are far beyond the individual’s
position (Miller 8 Dollard, 1941). Children identify more with models who have
the power to reward behavior rather than models who are competing for rewards
(Bandura, Ross, 8 Ross, 1971). Thus, the model that has the greatest effect on
an observer is the model that is like the observer, with positive characteristics,
but also accessible. In essence, the ideal self forms this better model. When
individuals create an idealized version of themselves, the ideal version tends to
be smarter, have a higher status, be more competent, and so forth.

However, when models are too far above what is attainable by the
individual, outcomes may suffer. Although results of the correlation were
consistent with hypothesized relationships for enjoyment and involvement,
knowledge and self-efficacy in the correlation, as well as the analyses of
variance, were not. This may have been caused by a lack of true similarity
between avatar and player. Similarity in Social Cognitive Theory is usually
physical similarity (Bandura, 1989, 2001; Wood, Solomon, 8 Englis, 2005). This
research used a verbal manipulation of psychological similarity. Although
physical similarity may lead to a belief that psychological similarity exists as well,

it is likely that physical similarity is what makes the difference in terms of

57

identiﬁcation. This study used the verbal manipulation to attempt to avoid
confounds by using various aspects of physical customization, but perhaps, in
line with similarity research, physical similarity would have yielded different
results. The Uncanny Valley (Mori, 1970) provides an extension to this
explanation. When non-human things come closer to seeming human, there
comes a point where similarity is almost (but not quite) achieved, and the non-
human aspects stand out, causing a negative response from humans. The
avatar used in this experiment was a dog. Although the avatar pretested better
than human-looking avatars and was used to eliminate cultural biases, perhaps
that dog was not similar enough to a human being and thus failed to elicit the
empathy needed for identiﬁcation.

Different results have been obtained using virtual reality-based
observational Ieaming, although there was no apparent enactive condition.
There are several reasons that observational Ieaming might work in virtual reality
settings but fail in video game settings. First of all, virtual reality has the ability to
make far better representations of ourselves, leading to a greater degree of
visual similarity which, according to various studies (Bandura, 1977; Jose 8
Brewer, 1984), leads to a greater degree of identiﬁcation. Because of this
potential for greater identiﬁcation (the model that subjects see is more
representative of the subjects’ views of self), observational Ieaming may be more
effective.

Additionally, virtual reality has subjects performing the behavior in real

time. While the video game in this project offered participants the ability to

58

practice doing behaviors that they would then perform for real, subjects were not
Ieaming movement and motion, but rather information and instruction about
where to go and what to click to remain safe. Virtual reality studies have
primarily utilized the teaching of movement (such as, for instance, Tai Chi
motions (Bailenson, Patel, Nielsen, Bajcsy, Jung, 8 Kurillo, 2008)) rather than the
teaching of more cognitive information. Learning how to be safe online still
requires movement and recall of the steps learned, but rather than “muscle
memory,” the Ieaming relies on more in-depth cognitive function. Although it is
tempting to say that observational Ieaming works only in cases where the
modeling is physical, there are instances where observational Ieaming was used
to Ieam mathematics rather than large muscle movements (Schunk, Hanson, 8
Cox, 1987).

Bailenson, Blascovich, and Guadagno (2008) found that people
responded more positively regarding intimacy—consistent behaviors to
representations of themselves than others. There were no differences between
those who interacted with physically similar representations of themselves and
those who interacted with physically dissimilar representations, although the
authors were hesitant to draw conclusions from this. They note that perhaps “the
belief that an embodied agent is a representation of the self is more important
than whether this embodied agent is photographically similar to an individual” (p.
2685). It is possible in the research reported in this paper that participants

viewed any condition as being a representation of a third party, due to the

59

physical dissimilarity between participants and the dog, and this is why there
were not signiﬁcant differences in the planned contrasts.

Although differences in information retained, involvement, behavioral
intentions, and self-efﬁcacy were not noted between conditions, enjoyment of the
game, enjoyment of Ieaming, and enjoyment of the avatar between ideal, real,
and other were correlated in the proper direction, with the ideal self condition
resulting in the highest levels of enjoyment, indicating that ideal self has the
stronger relationship with these variables. Even if game players using their ideal
self avatar do not experience more Ieaming and higher levels of self-efﬁcacy, the
fact that they enjoy the experience more suggests that they will return to the
material (Ritterfeld, Weber, Fernandes, 8 Vorderer, 2004; Vorderer 8 Ritterfeld,
2003). If two participants score equally well, but one had a more enjoyable time,
this individual will, at least in terms of the game contained in this research, have
a better chance of keeping his computer safe than the individual who viewed the
experience as less enjoyable.

In summary, this research explored which type of Ieaming is the more
effective in educational video games, and whether or not avatar type has any
effect on Ieaming outcomes. Results indicated that the enactive Ieaming
condition resulted in higher self-efﬁcacy, and that, in the case of textually-
manipulated avatar type, the ideal self avatar type had the more successful

results than an avatar representing one’s real self or a third party.

60

Chapter 6: Limitations

Manipulation Check

The manipulation itself for establishing real, ideal, and friend (or other)
worked in approximately 60% of the total cases. These results indicate that the
manipulation did not have the desired effect on as many participants as would
have been ideal. Most successful were the manipulations for real and ideal
selves, as would be expected. Asking participants to view an avatar as a
representation of how they actually see themselves would be the easiest leap to
make. Asking participants to view the avatar as their ideal selves would be
slightly more difﬁcult, but still, the only knowledge necessary is knowledge that is
present inside the minds of the participants. They know their real selves, and
they know their ideal selves. Indeed, results show that when using the
standardized scores for the distance analysis, correlations between real self and
friend with the dependent variables yielded the same results. Therefore, use of a
friend condition rather than a neutral third party may have negatively affected
results. Most individuals choose friends based on attraction to that which is
similar to themselves (Byme, 1971), so any differences between one’s real self
and the chosen friend would likely be minimal (as the data show) compared to
potential differences if a third party unknown to the player prior to game play was
used. However, the use of a friend was chosen to attempt to balance the
limitation created by only having players play the game once, creating a situation
that would be relatively equivalent across the three avatar conditions. People

know who they are and who they would like to be, as well as knowing what their

61

friends are like. Because it takes time to form a connection with a game
character, using a friend that the game player already has a connection with
would have allowed players to more adequately have an impression of the
character. Rather than risking random answers to the questions for the similarity
distance measure due to the player not having a strong connection to the
character, participants were asked to think about a friend and rate the friend with
the assumption that they would have a better grasp of the friend's personality.

The manipulation with the lowest success rate was that for the friend
condition. Instead of one connection to be made (putting one’s own self concept
inside the body of a digital character), two were required: imagining what the
friend they chose is like (essentially putting themselves in the minds of their
friends), and then putting that concept into the avatar. Part of the reason that the
verbal manipulation may have been so weak was simply that physical similarity is
far more important than mental similarity. Without the physical cues, participants
could not connect with the avatar. However, it may also be a case for the
necessity of repeat game play.

There were no signiﬁcant correlations between whether or not the
manipulation check worked and the dependent variables of the study. This
means that the success or failure of the manipulation did not somehow inﬂuence
the results, but that the manipulation itself was more probably the cause for
differences found. In other words, were there to be signiﬁcant correlations
between success and the dependent variables, differences found may be

attributed to the type of individual—perhaps participants for whom the

62

manipulation was successful are more conscientious individuals in general, will
read more carefully, pay more attention, and be more involved with the material.
This would lead to signiﬁcant correlations between success and the dependent
variables, and would lead to erroneous conclusions. However, success or failure
of manipulation did not have an effect, and thus, this does not create a confound
for results.

Text was used to manipulate ideal, real, and friend for several reasons.
The major reason for not using avatar characteristics was to avoid confounding
any results with the various customization options. For instance, if participants
were told to make their ideal selves, with choices for hair color, hair style, eye
color, clothing, gender, and age, this leaves six variables to contend with. It is
entirely possible that one of these would be responsible for the adequate
manipulation in one case (say, hair style), and the other characteristics were not
at all important in determining whether or not the participant thought of the
character as an ideal self. For other participants, the deﬁning variable may be
gender, or age, or any combination of variables. By eliminating confounds, this
research examined whether avatar type made a difference; future research can
determine which visual characteristics make the largest differences.

Secondly, the important variable for this research was avatar type: real
self, ideal self, or third party. If participants had been allowed to choose the look
of their avatar, using any of the above options, customization choice would have
created confounds. Choice boosts conﬁdence in one’s ability to perform a task

(T afarodi, Mehranvar, Panton, 8 Milne, 2002) which would have likely had an

63

effect on self-efﬁcacy, and customization would result in a great many choices
(skin color, hair style, eye color, and so on) that would make it exceedingly
difﬁcult to determine which characteristic(s) made the difference in participants’
imagining of the avatar as the proper type.

Thirdly, individuals from a variety of cultures and races were included in
the research. To remain culturally neutral, six different non-human avatars were
pretested. The dog avatar resulted in a greater number of successful
manipulations than did the other ﬁve. Unfortunately, the use of the dog avatar
might have created a confounding variable in an attempt to eliminate other
confounds. For those who like other animals better, or who do not like dogs at
all, these individuals may have had a more difﬁcult time identifying with the dog
character. As well, the dog’s fur was brown, which may have been a race
confound.

However, in other research studying avatars and Ieaming, there are
inevitably confounds and this research attempted to eliminate as many
confounds as possible. As previously discussed, by using a non-human avatar
that remained consistent and only altering the textual information cues, cultural
biases were reduced, as were gender biases. By using a consistent avatar, the
potential for one characteristic (eye color, hair style, etc.) to affect one participant
in terms of belief that the avatar was the participant’s ideal self and to not affect
another was also eliminated. Even in studies where actual representations of

participants were used, the possibility for experimenter and observer effects,

particularly in studies where participants needed to spend time having their
virtual selves created, exists.

The game also needed to be comparable in the avatar and the no-avatar
(observational and enactive) conditions so that observations could be made
regarding the effectiveness of the conditions. Keeping the avatar simple and
constant across conditions and merely changing the text prompts to reﬂect either
the presence or absence of an avatar allowed for consistency that might
othenririse have been absent. (See Appendix B for a collection of screenshots
both with and without the avatar.)

Hypothesis 1

It was hypothesized that the observational condition would result in
greater Ieaming than the enactive condition, as modeling is theorized to be the
strongest method of Ieaming. However, signiﬁcant differences in Ieaming
outcomes between the two conditions were not found. Although discussion
offered several explanations for these results, of course they may also be due to
design and experimental limitations.

It is possible that participants knew the material prior to playing the game.
However, pretesting eliminated questions and responses that were too easy or
too difﬁcult and random assignment of participants would have ensured that
regardless of prior knowledge, differences found between conditions would still
be valid. Each participant was asked ten questions about what they had just
learned in the video game during the posttest, but perhaps ten questions were

not enough to adequately determine a difference in knowledge.

65

This research did not include a control condition because it would have
been impossible to ensure precise similarity between the control group and the
rest of the groups and because this research was done entirely online. Future
research, however, should include a non-interactive presentation of the
information. However, this had the potential to confound results because time on
task can vary widely. Those individuals playing the game would likely spend
more time with the material than those reading the copy because ﬁrst of all,
people read faster on paper than they do online. People also skim material
online (Nielsen, 2006). Secondly, it takes more time for users to click the mouse
and move about the screen than it does for individuals to read a printed page.
Because of the potential for a time confound, a printed-material control group
was not used. Perhaps one consideration would have been to put the material
on a webpage and randomly assign participants. Those in the control condition
would simply visit the website, scroll through the material, and move on.
However, the time discrepancy still stands. The research as it is now shows that
the groups all spent the same amount of time interacting with the material and
that time was not affected by the random group assignment.

Because time remained comparable among all groups (analysis of
variance showed no signiﬁcant differences), one of the typical explanations for
differences between groups with regard to video games can be discarded. It
might be tempting to say that the reason for any difference, whether it is a
positive or negative (in the sense of positive effects that we would like to see,

and negative effects that we wish we did not see), can be attributed to time. The

66

—==- *1"

more time an individual spends with any material, let alone a video game, the
more likely he or she is to remember the information (Ebbinghaus, 1913).
Hypothesis 2

This research showed that the enactive condition resulted in higher levels
of self-efficacy than did the observational condition for cases that passed the
manipulation check. Rather than watching someone else succeed, individuals in
the enactive condition were themselves experiencing the consequences.
Therefore, self-efﬁcacy levels should be higher for individuals in that condition,
but this was only true when analyzing only those who passed the manipulation
check. Perhaps the individuals who passed the manipulation check were more
careful readers and took the experiment more seriously, resulting in greater self-
efﬁcacy gains than those who were not able to suspend disbelief, who did not
read the instructions carefully enough, and who ultimately did not take the game
as seriously.
Hypothesis 3

Although means were in the correct direction, intentions to perform
behaviors in the future were not affected by the enactive versus observational
conditions. Two of the behavioral intention items exhibited extremely signiﬁcant
differences between the groups as individual items rather than taking the items
as a whole. These were two things related to phishing emails, which many
people do not know about. These future behaviors were likely things that
participants did not know about prior to playing the video game and were

behaviors that the game covered repeatedly, more so than the other behaviors.

67

It might have been that the game spent more time with these, and therefore, they
showed signiﬁcant differences while the other behaviors that received less time
did not.

As well, these behaviors were extremely easy to do. They required no
real additional effort (no downloading, etc.) and could be accomplished in a
matter of seconds. Participants in the observational condition were exposed to
consequences through vicarious Ieaming, but they may have also had things
happen to them related to breaches of online safety, meaning they may have had
personal (enactive) experience. This may have interfered.

Changing settings and the inner workings of a computer may be daunting.
However, the two signiﬁcant behaviors were things that did not require any
settings be changed. There was no potential for error with these two behaviors,
as they involved checking links and searching using Google.

Finally, perhaps the number of participants in each cell was not high
enough to ﬁnd effects that are actually there (Type II error). Although alpha was
raised to assist in elimination of Type II error, when taking the entire sample and
using the distance measure as the independent variable, relationships start to
develop in predicted directions. It is possible that the N for this experiment was
simply not high enough to detect signiﬁcant differences. The effects that may be
present might be larger than the effect size originally used to estimate cell size.
Hypothesis 4

Although the planned comparisons did not show any signiﬁcant

differences between groups, the correlation analysis found the strongest

68

signiﬁcant correlations in the ideal self condition. Again, this could be due to the
smaller cell sizes (Type II error). Although power analysis showed that 70
subjects per condition should be adequate for ﬁnding a medium effect size, it is
possible that the effect in question is a smaller effect size of approximately .2
(instead of the medium effect size of .3) and a larger N would make a difference.

Participants responded that they would most prefer to play as their real
selves, followed by a third party (a friend or other external individual), and lastly
their ideal selves. This is incredibly surprising, given that much of what is written
about the appeal of video games is the ability of individuals to try on new
personas, to do things they could not (or would not normally) do in real life. The
means for enjoyment of the game, of Ieaming, and of the avatar were higher in
the ideal condition than in the other two, and although they did not reach
signiﬁcance, perhaps this is indicative that people like to pretend to be something
they aren’t, but when it comes down to doing something more than just enjoying
themselves, such as Ieaming new material, the type of avatar isn’t important.
Indeed, it would appear that the avatar is not necessary at all.

The lack of signiﬁcant planned contrast differences may have been due to
a smaller subject sample (approximately 75 individuals in each group). When N
increased using the continuous variable of distance as the independent variable,
the correlation matrix showed signiﬁcant results. It is also possible that the
subject sample was too old to beneﬁt from any manipulation of sense of self or
other. Although the sample falls within the range of the typical video and

computer game player—the average age of game players is 35, according to the

69

Entertainment Software Association (2008) with about 49% of gamers falling
between the ages of 18 and 49, individuals who use a video game to learn are
likely younger, so the age of the game players may have skewed the results.

Moreover, many times, respondents do not carefully read text presented
to them online (Nielsen, 1997). Rather than carefully reading the instructions to
imagine that the avatar was a certain type (real self, ideal self, or friend), the
participants for whom the manipulation was not effective may have simply
skimmed this instruction page (or even skipped it entirely) and clicked next. To
play a video game, suspension of disbelief is necessary so that players can
become immersed in the game world (Green, Brook, 8 Kaufman, 2004; Oatley,
1994). For some of the participants, suspending disbelief and being able to
imagine that they were a two-dimensional canine character on the screen may
have been too difﬁcult to overcome at all, let alone to imagine that the character
possessed traits of some version of themselves or a friend, even if they did read
the instruction page.

Perhaps, however, avatars are not necessary when using games as a
Ieaming tool. Though avatars can make the experience more enjoyable (Chan 8
Vorderer, 2006; May, 1994), it is possible that their presence detracts from the
game material. An item on the questionnaire asking respondents if they would
rather play as their ideal self, real self, friend/other person, or something else
resulted in a number of responses indicating “something else,” and of those,
several reported that they would prefer no avatar at all. Those who selected

“other” (the fourth response) were provided space in which to explain their

70

 

1F» .

preference. One respondent wrote, “it [the avatar] takes away from the fact that
computer useage (sic) isn't some cute little game. Most ’avatars’ are associated
with annoyance and hindering useability (sic), not helpful guides. IE: Clippy".
Again, this could also be an issue of age; adults do not typically play games to
learn and have been exposed to previously-created avatars with negative
experiences regarding those avatars (as the respondent noted, the Microsoft
Ofﬁce assistant).

As indicated in the results, involvement was lowest in the real condition,
whereas behavioral intentions were lowest in the friend condition, but both were
highest in the ideal condition, though no signiﬁcant differences were found
between the three groups. An analysis of involvement examining only
differences between ideal and real self, however, found a signiﬁcant difference.
If participants were imagining their ideal selves while playing, it would stand to
reason that they were still imagining their ideal selves while answering the
questionnaire. Socially desirable behaviors, then, such as considering online
safety to be important, could be attributed to idealized versions of self, because if
respondents could be their ideal selves, they would consider online safety
important. In the real condition, however, respondents answered honestly, and
this resulted in a lower level of involvement. The desire to make your actual self
reﬂect the goals and beliefs of your ideal self is known as the social desirability
bias, and even though the only recipient of each individual’s answers was the

researcher, this tendency appears to hold true, regardless of the “social” aspect.

71

expl
phyi
bee
the

the

all

lie

pl

Though Ieaming with electronic environments is just beginning to be
explored, research in press by Bailenson et al. indicates that participants learning
physical actions learn better in virtual reality than through a video condition
because of the ability to see themselves represented on-screen (2008). Perhaps
the lack of signiﬁcant ﬁndings between conditions in this research is related to
the type of avatar used. Had this experiment been able to place exact versions
of individuals in a three—dimensional game space for the real self condition,
allowed participants to adjust their image for their ideal self, and put the exact

version of the subject’s friend in the friend condition (rather than using a

 

previously-created, ﬁxed avatar), results could have been different. Part of the a
appeal of an avatar is the feeling of connection (Allison, von Wahlde, Shockley, 8
Gabbard, 2006; Chan 8 Vorderer, 2006; May, 1994). In an attempt to remain
neutral and consistent, this research utilized one constant avatar, but it is unlikely
that this connection would exist strongly in every participant.

Because the content of the game revolved around online safety, a topic
that can have very serious consequences and that is, in and of itself, serious, the
addition of the avatar may have detracted and distracted participants from the
actual material. Although only one respondent made a comment about this, it is
quite possible that many individuals felt the same way. In fact, game players
may have been enjoying the avatar and enjoying the ability to imagine their ideal
selves, but the enjoyment of the novelty may have overshadowed the content.
The simplistic avatar and graphics may have made the topic seem less serious,

so respondents were not as motivated to take the lesson seriously, which was

72

then reﬂected in the lack of signiﬁcance for various results. When it comes to
enjoyment of a game, clearly, people enjoy pretending to be something they are
not. But when adults are Ieaming with a video game, this makes no difference.
As summarized by the respondent above, perhaps there is no room for

pretending, because Ieaming is serious business.

73

Chapter 7: Directions for Future Study

The ﬁrst and most obvious step to take next in order to further examine
this line of research would be to design a study that contained observational
Ieaming as well as the new potential type discovered here: auto-observational
learning. The conditions could still be tested using a video game, but in the pure
observational Ieaming condition, participants would simply watch the game play
itself, as if watching a video. The auto-observational condition participants would
play the game. This could be compared to a third condition of purely enactive
learning, although this could not be integrated into a video game due to the
enmeshing of enactive and observational Ieaming in a game. Instead, enactive
Ieaming would require participants to be on a computer with instructions to
change settings, etc. 1

Putting in this external observational condition (where participants observe
the game without interacting with it) would test in more depth differences
between observational and auto—observational Ieaming. This concept of self-
modeling or auto-observational Ieaming needs further exploration as well. To
test whether or not the appearance of the avatar (physical similarity) affects the
effectiveness of Ieaming using self-modeling could be accomplished by putting
images of game players directly onto avatars and assigning players their self-
modeled avatars, or by putting images of individuals that are not the game
players onto avatars and assigning those to players as well. As discussed
previously, similarity deals more with physical characteristics rather than

cognitive ones, and it is possible that simply manipulating participants” cognitive

74

images of the avatar was not enough to elicit strong identiﬁcation due to
similarity. Perhaps the physical characteristics must match as well.

Because participants only had one opportunity to play the relatively short
video game, exploring results when individuals are allowed repeat exposure to
the avatar in the game is needed. Adding three more conditions, still with the
three types of avatar, but allowing the ﬁrst triad to play once and instructing the
second triad to play for a certain length of time or to play through a certain 2
number of times would allow comparison between the connection with type of
self (or third party) and avatar gained in a short time versus a longer time.

Related to this, if the use of a longer exposure to the stimulus is utilized, it is also

 

possible to use a neutral third party avatar with which the players are not familiar.
Extended time would allow for a stronger connection to develop and would allow
players to develop their own ideas about what this third party character is like.

Future research should also better explore the use of various types of
avatars, especially using technology that allows for better quality avatars that are
personalized for each respondent. It would be particularly useful to have access
to technology that allows for an exacting image of the individual to be placed
within the game, such as using virtual reality. Subjects in the ideal self condition
could then alter their images to look like what they would like to look like and pick
personality traits that would represent their ideal selves.

As discussed earlier, it is possible that the use of text rather than image
did not create a strong enough link between avatar and player. A stronger

priming effect might make a greater impact on results. Rather than simply telling

75

players how to view the avatar, participants could be asked a series of questions
about how they (as either their ideal self or real self) would respond in various
situations and then shown the generic avatar with instructions to imagine the
avatar as having all the characteristics and qualities they just described in their
answers. Those in the third party condition may be shown an image of an avatar

and asked a series of questions about how they believe that avatar might

respond in the same situations described in the two self conditions. If order of 5'
exposure is problematic, participants in the third party condition could be asked
how they think someone well-known but unlikely to be similar to either their real
or ideal self (the governor of the state, for instance), or even someone they L

 

actively dislike, would respond prior to showing them the avatar and giving them
the instructions.

Using a dog as an avatar rather than a human being likely limited
identiﬁcation with the avatar, so designing research studies to replicate this one
but utilizing humans as the avatars could lead to more signiﬁcant results in line
with the theoretical grounds. In order to determine which characteristics ought to
be included or offered as choices for avatar customization to maximize similarity
between ideal and real selves, exploratory research needs to be conducted ﬁrst.
One potential method could involve using an existing avatar creation system,
such as the concept behind the Nintendo Wii’s “Mii”. Individuals could be
assigned to create their real or ideal selves using the system, and then rate or
rank which of the characteristics were most important to them in determining

what made the avatar representative: hair style, hair color, gender, skin tone,

76

age, height, weight, and so forth. Limiting options (as this existing system does)
would eliminate some of the confounds that might be found in something like
Second Life or The Sims, where customization and hacks lead to almost
innumerable and difﬁcult-to-classify characteristics.

The less involved method of creating a system to offer human cues for
avatar similarity may be done by simply interviewing game players in open
interview sessions. From there, avatars can be designed that manipulate only
one of the common cues, and each manipulated characteristic forms its own
study of differences. For instance, a female and a male avatar can be offered in

the game with the same outcome variables measured to determine the

 

signiﬁcance of gender, although investigation of each particular characteristic in
this manner would require numerous studies and participants. Thus, the former
method is preferable.

Wood, Solomon, and Englis (2005) conducted a study regarding avatars
and online shopping clothing models. Subjects preferred photographic, rather
than characterized, avatars (realistic) but they also preferred idealized avatars
(the equivalent of a supermodel in real life). In essence, then, people prefer
realistic-looking avatars, but avatars that are idealized. The realistic qualities
might assist in the video game world of being able to make the connection
between the game and real life, but the idealization of an avatar could provide
the inherent “expertise” that players don’t have. In the Wood, Solomon, and
Englis study, perhaps it was the fact that the models used looked as if they were,

in fact, knowledgeable about fashion. Participants were more familiar with some

77

of the images than others (for instance, Meg Ryan’s image was used) and
perhaps that familiarity caused participants in their study to prefer the images
with which they were familiar. This, combined with the fact that the models were
those that were typically related to fashion and clothing, could have led subjects
to prefer them because they would have seen these individuals as having
expertise. It would be interesting to conduct this video game study with avatars
utilizing the images of Bill Gates and Steve Jobs to investigate familiarity and
expertise further in terms of avatars and learning using video game
environments.

Finally, and perhaps most importantly, if self-efﬁcacy was signiﬁcantly
higher in the enactive (no avatar) condition, but enjoyment of the game,
enjoyment of Ieaming, enjoyment of the avatar, and involvement were correlated
most strongly with distance in the ideal self condition as compared to real self
and other, maybe this means that there is no need for an avatar at all, but for
merely the suggestion of an avatar. Perhaps players do not need to be
physically convinced about the similarity of their avatars to themselves, since any
attempts to use physical similarity elicit numerous confounds, but instead,
players need to be psychologically linked to their avatar in order to experience
the greatest positive Ieaming outcomes, and linked using their ideal selves.
Suggesting to players that their ideal self is the main character in the game and
simply providing textual cues to this end could potentially result in a combination
of the positive effects from both conditions, and better encompass the reasons

for creating educational video games in the ﬁrst place. Further exploration of this

78

comparing ideal, real, and other manipulations using text only without an avatar
is necessary to better understand the psychology behind avatar use, but perhaps
the Entertainment-Education Paradigm is not a lost cause after all, but needed
only to be placed on the shoulders of Social Cognitive Theory in order to better

see the future for educational video games.

 

79

APPENDICES

80

 

Appendix A

Table 7: Correlation Matrix for Involvement Scale Items

 

 

 

 

 

Item Item 1 Item 2 Item 3 Item 4 Item 5
1. Matters to me Doesn’t matter to me 1.00 .79 .69 .85 .42
ﬁeOf concern to me Of no concern to .79 1.00 .70 ‘75 '45
3. Relevant Irrelevant .69 .70 1.00 .82 .40
4. Important Not important .85 .75 .82 1.00 .41
5. Interesting Not interesting .42 .45 .40 .41 1.00
Table 8: Correlation Matrix for Self-Efﬁcacy Scale Items
Item Item Item Item Item Item Item Item
1 2 3 4 5 6 7

1. lam conﬁdent that I can

niz e a phishing attack. 1.00 .94 .80 .66 .81 .76 .69
2. I am conﬁdent I can protect
myself from phishing attacks. .94 1.00 .80 .68 .81 .76 .72
3. I am conﬁdent I can close
pop-up windows the correct .80 .80 1.00 .66 .88 .83 .65
way.
4. I am conﬁdent I can reject
invasive cookies. .66 .68 .66 1.00 .70 .65 .75
5. I am conﬁdent I can set my
browser to block pop-ups. .81 .81 .88 .70 1.00 .90 .70
6. I am conﬁdent I can set my
computer to update .78 .76 .83 .65 .90 1.00 .67
automatically.
7' ' am °°"ﬁde"t ' ca" keep .69 .72 .65 .75 .70 .67 1.00

myself safe from spyware.

 

81

Table 9: Correlation Matrix for Behavioral Intentions Scale Items

 

Item

Item
1

Item
2

Item
3

Item
4

Item Item
5 6

Item
7

Item
8

Item
9

 

1. Set my computer
to automatically
update.

2. Change the
settings for cookies
on my computer.

3. Change the
settings for popups
on my computer.

4. Discuss phishing

with friends or family.

5. Discuss cookies

with friends or family.

6. Discuss computer
settings with friends
or family.

7. Check URLs in
email for phishing
attempts before
clicking on them.

8. Look up phone
numbers given in
emails to check for
phishing attempts.

9. Search Google to
verify identities of
email senders.

1.00

.66

.61

.35

.25

.35

.33

1.00

.85

.52

.53

.42

.31

.40

.33

.61

.85

1.00

.45

.37

.24

.37

.30

.52

1.00

.81

.24

.42

.35 .25

.53 .42

.45 .37

1.00 .83

.83 1.00

.24 .21

.41 .37

.41 .41

.31

.24

.24

.24

.21

1.00

.48

.47

.35

.40

.37

.41

.37

.48

1.00

.71

.33

.33

.30

.42

.41

.41

.47

.71

1.00

 

82

 

Appendix B
Figure 1: Opening Screen (with Avatar)

Online safety should be everyonc's job.
Can It be yours?

Enter the appropriate names below, and then click on
the "Next" button to begin!

Eﬂbf your DEMO!

Enter the name of a
friend here:

 

N EXT

83

Figure 2: Phishing Email Example

 

Email 2:

SUBJECT: Account updates I
Dear eBay member:
As part of our continuing commitment to protect your account and to reduce the instance

of fraud on our website, we are undertaking a period review of our member accounts.
You are requested to visit our site by following the link given below:

hlt :l/arribba. i3.eba .com/aw- i/eba ISAPl.dll? atelnforrnationConﬁrm8b user=1

         

Please ﬁll in the required information. This Is required for us to continue to offer you a
safe and risk free environment, and maintain the eBay Experience.

Thank you,
Accounts Management

 

http://81.180.59.10/index.htm
Is this a phishing email?

 

 

 

Figure 3: Phishing Completion Screen (without Avatan

 

INBOX *7 HINTS g

Congratulations!

You spotted all of the phishing emails and cut their lines
before they could hook you! You have now eamed your
“Anti-Phishing License“.

Your next job is to make sure your computer will be safe
while your family browses the Internet. You will learn about
cookies and pop-up blockers, and how to use them to stay
safe.

Click the 'Next' button to continue.

NEXT

 

 

 

85

Figure 4: Accepting Non-Invasive Cookie (with Avatar)

 

«. ﬁremeaﬂmwmwi

 

f; http:i',.'lsalety msuedu

”ﬁaﬁ'ﬂn "<4 w.‘ . I .

 

Goodl This cookie is not
invasive. This means that it will
not do anything you don't know
it‘s ' .

Click 'Next' to continue.

 

 

Us: .,‘ .
if C
I

 

 

 

86

 

 

Figure 5: Instant Message Level (with Avatar)

 

 

What should
I do?

 

 

I.
.p
. . '.
A .Q?

t .23.“: 'w" - r:",;{
.3, ,

mm N 14% 3.318.281
. httpzt'isafety msu edu

- i‘u’ﬂTM' 1.314.") u Tt-IILJJVE‘MHLqu-C‘J a - . ‘ s

a. ‘
H.338.- ‘99

 

 

 

Message from I j

 

Ijust saw the most hilarious video ever!

I‘m sending you the ﬁle. You can
download it and watch it yourself!

F unnyVideo.exe

ACCEPT REJECT

   

 

 

87

 

 

 

Figure 6: Moving On To Final Level (with Avatar)

 

88

Figure 7: Setting Automatic Updates (with Avatar)

Control Panel — X

   

 

 

Figure 8: Pop-Up Blocker (with Avatad

 

httpzs'tisaletymsu edu

, {‘19: 'I‘L’Q-J’LAL ;t.'r'e'.‘1l;\‘i‘: $3,; '31 J. ZEEFLW. I -

 

 

 

 

 

 

 

 

 

What should I choose
for the pop-up blocker?
Pop-up Blocker

Tum ON

Pop-up Blocker
Tum OFF

Pop-up Biocker

L y

 

 

 

90

 

 

Figure 9: Pop-Up Blocker (without Avatar)

 

 

' y} c ‘ '7 V . . 7A???) "
1.4 I marsmememma armamentamv
’ ' http: 'sisafety msu edu

.‘. "fﬁéfrmi i; J- T’k‘v‘SE. C

 

 

 

 

(r

Tum ON

 

 

What should you choose
for the pop—up blocker?

Pop-up Blocker

Turn OFF
Pop-up Biocker

 

 

91

 

 

Figure 10: Final Screen (with Avatar)

 

 

Congratulations! You have ofﬁcially completed
the training designed to prepare you for your
job as Online Security Expert!

 

 

Click ”'Next'” to continue to the questionnalrc.

 

 

 

92

REFERENCES

93

 

References

Adams, J. A. (1986). Use of the model’s knowledge of results to increase the
observer’s performance. Journal of Human Movement Studies, 12, 89—98.

Adobe Flash CS4 Professional. (2008), for Windows. [Computer Software]. San
Jose, CA: Adobe Systems Inc.

Allison, S. E., von Wahlde, L., Shockley, T., 8 Gabbard, G. O. (2006). The
development of the self in the era of the intemet and role-playing fantasy games.
The American Journal of Psychiatry, 163(3), 381 -385.

AI-Natour, 8., Benbasat, l., 8 Cenfetelli, R. T. (2006). The Role of Design
Characteristics in Shaping Perceptions of Similarity: The Case of Online
Shopping Assistants. Joumal of the Association for Information Systems, 7(12),
821. .

Amory, A., Naicker, K., \ﬁncent, J., 8 Adams, C. (1999). The use of computer
games as an educational tool: Identiﬁcation of appropriate game types and game
elements. British Journal of Education Technology, 30(4), 311-321.

Andsager, J.L., Bemker, V., Choi, H.L., 8 Torwel, V. (2006). Perceived similarity
of exemplar traits and behavior: Effects on message evaluation. Communication
Research, 33(3), 3-1 8.

Bailenson, J.N., Blascovich, J., 8 Guadagno, RE. (2008). Self representations in
immersive virtual environments. Journal of Applied Social Psychology, 38(11),
2673-2690.

Bailenson, J.N., Iyengar, S., Yee, N., 8 Collins, N. (2007, in press). Facial
similarity as a voting heuristic. Public Opinion Quarteriy.

Bailenson, J.N., Patel, K., Nielsen, A., Bajcsy, R., Jung, 8., 8 Kurillo, G. (2008).
The effect of interactivity on Ieaming physical actions in virtual reality. Media
Psychology, 11(3), 354-376.

Bandura, A. (1969). Principles of behavior modiﬁcation. New York: Holt,
Rinehart, 8 Vlfinston.

Bandura, A. (1977). Self-efﬁcacy: Toward a unifying theory of behavioral change.
Psychological Review, 84(2), 191-215.

Bandura, A. (1986). Social foundations of thought and action: A social cognitive
theory. New Jersey: Prentice-Hall.

94

 

Bandura, A. (1989). Human agency in social cognitive theory. American
Psychologist, 44(9), 1175-1 184.

Bandura, A. (1990). Perceived self-efﬁcacy in the exercise of personal agency.
Journal of Applied Sport Psychology, 2, 128-163.

Bandura, A. (1997). Self-efﬁcacy: the exercise of control. New York: W.H.
Freeman.

Bandura, A. (2001). Social cognitive theory of mass communication. Media
Psychology, 3, 265-299.

Bandura, A. (2006). Guide for constructing self-efﬁcacy scales. In T. Urdan 8 F.
Pajares (Eds), Self-efficacy beliefs of adolescents (pp. 307-337). Charlotte, NC:
lnfonnation Age Publishing.

Bandura, A., Ross, 0., 8 Ross, SA. (1963). A comparative test of the status
envy, social power, and secondary reinforcement theories of identiﬁcatory
Ieaming. Journal of Abnormal and Social Psychology, 67, 527-534.

Basen-Engquist, K. (1994). Evaluation of a theory-based HIV prevention
intervention for college students. AIDS Education and Prevention, 6(5), 412424.

Bennett, 3., Farrington, DP, 8 Huesmann, LR. (2005). Explaining gender
differences in crime and violence: The importance of social cognitive skills.
Aggression and Violent Behavior, 10, 263-288.

Bennet, J., Lyons, K.S., Winters-Stone, K., Nail, L.M., Scherer, J. (2007).
Motivational interviewing to increase physical activity in long-term cancer
survivors: A randomized controlled trial. Nursing Research, 56(1), 18-27.

Berger, SM. (1977). Social comparison, modeling, and perseverance. In J.M.
Suls 8 R. L. Miller (Eds), Social comparison processes: Theoretical and empirical
perspectives (pp. 209-234). Washington, DC: Hemisphere.

Berson, IR, 8 Berson, M.J. (2005). Challenging online behaviors of youth:
Findings from a comparative analysis of young people in the United States and
New Zealand. Social Science Computer Review, 23(1), 29-38.

Bessiere, K., Seay, A.F., 8 Kiesler, S. (2007). The ideal elf: Identity exploration in
World of Warcraft. CyberPsychoIogy 8 Behavior, 10(4), 530-535.

Bobbitt, L.M., Inks, S.A., Kemp, K.J., 8 Mayo, OT. (2000) Integrating marketing

courses to enhance team-based experiential Ieaming. Journal of Marketing
Education, 22(1), 15-24.

95

Bosworth, K. (1994). Computer games and simulations as tools to reach and
engage adolescents in health promotion activities. Computers in Human
Services, 11(2), 109-119.

Braaksma, M.A.H., Rijlaarsdam, G., 8 van den Bergh, H. (2002). Observational
Ieaming and the effects of model-observer similarity. Joumal of Educational
Psychology, 94(2), 405-415.

Byme, D. (1971). The attraction paradigm. New York: Academic Press.

Carroll, W. R., 8 Bandura, A. (1985). Role of timing of visual monitoring and
motor rehearsal in observational Ieaming of action patterns. Journal of Motor
Behavior, 17(3), 269—281.

Chan, E., 8 Vorderer, P. (2006). Massively multiplayer online games. In P.
Vorderer 8 J. Bryant (Eds), Playing computer games: motives, responses, and
consequences. Mahwah, NJ: Lawrence Erlbaum Associates.

Choi, SM, 8 Rifon, NJ. (in press). It’s a match: The impact of congruence
between celebrity image and consumer ideal self on endorsement effectiveness.
Psychology 8 Marketing.

Clark, SE, 8 Ste-Marie, OM. (2007). The impact of self-as-a-model
interventions on children's self-regulation of Ieaming and swimming performance.
Joumal of Sports Sciences, 25(5), 577-586.

Couzijn, M. (1999). Learning to write by observation of writing and reading
processes: Effects on Ieaming and transfer. Learning and lnstnrction, 2, 109-142.

Cumming, J., Clark, 8., Ste-Marie, D., McCullagh, P., 8 Hall, C. (2005). The
functions of observational Ieaming questionnaire (F OLQ). Psychology of Sport
and Exercise, 6, 517-537.

Debowski, S. (2002). Developing effective electronic information seekers.
Australian Journal of Management, 27, 21-29.

Dickey, MD. (2005). Engaging by design: How engagement strategies in popular
computer and video games can inform instructional design. Educational
Technology Research and Development, 53(2), 67-83.

Dipietro, M., Ferdig, R. E., Boyer, J., 8 Black, E. W. (2007). Towards a

Framework for Understanding Electronic Educational Gaming. Joumal of
Educational Multimedia and Hypermedia, 16(3), 225.

96

Ebbinghaus, H. (1913). Memory: A contribution to experimental psychology (H.A.
Ruger 8 CE. Bussenius, Trans.) New York: Columbia University. (Original work
published in 1885).

Egenfeldt-Nielsen, S. (2007). Third Generation Educational Use of Computer
Games. Journal of Educational Multimedia and Hypermedia, 16(3), 263.

Egloff, T. H. (2004). Edutainment: A case study of interactive cd-rom playsets.
Computers in Entertainment, 2(1), 13-1 3.

Entertainment Software Association. (2008). Essential facts about the computer
and video game industry. Retrieved March 13, 2009, from
http://www.theesa.com/facts/pdfleSA_EF_2008.pdf

Ericksen, M. K. (1996). Using self—congruity and ideal congruily to predict
purchase intention: A European perspective. Joumal of Euromarketing, 6(1), 41-
56.

Farrington, D. P. (1998). Predictors, causes, and correlates of male youth
violence. In M. Tonry, 8 M. Moore (Eds), Youth violence (pp. 421—475).
Chicago: University of Chicago Press.

Farrington, D. P. (2001a). Predicting adult ofﬁcial and self-reported violence. In
G. F. Pinard, 8 L. Pagani (Eds.), Clinical assessment of dangerousness:
Empirical contributions (pp. 66—88). Cambridge, England: Cambridge University
Press.

Farrington, D. P. (2001b). The causes and prevention of violence. In J. P.
Shepherd (Ed.), Violence in health care (2nd ed.) (pp. 1—27). Oxford, England:
Oxford University Press.

Ferrari, M. (1996). Observering the observer: Self-regulation in the observational
learning of motor skills. Developmental Review, 16(2), 203-240.

Festinger, L. (1954). A theory of social comparison processes. Human
Relations, 7(2), 117-140.

Fox, J., 8 Bailenson, J.N. (2009). \ﬁrtual self-modeling: The effects of vicarious
reinforcement and identiﬁcation on exercise behaviors. Media Psychology, 12(1),
1-25.

Fruin, D. J., Pratt, C., 8 Owen, N. (1992). Protection motivation theory and

adolescents' perceptions of exercise. Joumal of Applied Social Psychology,
22(1), 55-69.

97

Gee, JP. (2003). What video games have to teach us about Ieaming and
literacy. New York: Palgrave Macmillan.

Gentile, DA, 8 Anderson, CA. (2006). Video games. In N.J. Salkind (Ed),
Encyclopedia of Human Development (pp. 1303-1307). Thousand Oaks, CA:
Sage Publications.

G*Power. (2008), for Windows (Version 2) [Computer Software]. Duesseldorf,
Germany: Heinrich Heine University. Available from http://www.psycho.uni-
duesseldorf.de/aap/projectslgpower/

Graeff, T. R. (1996). Using promotional messages to manage the effects of brand
and self-image on brand evaluations. Joumal of Consumer Marketing, 13(3), 4-
1 8.

Graham 8., 8 Harris, K. R. (1994). The role and development of self-regulation in
the writing process. In D. H. Schunk 8 B. J. Zimmerman (Eds), Self-regulation of
Ieaming and perfonnance: Issues and educational applications (pp. 203-228).
Hillsdale, N.J.: Erlbaum.

Graham, 8., Harris, K. A., 8 Troia, G. A. (1998). Writing and self-regulation:
Cases from the self regulated strategy development model. In D.H. Schunk 8
B.J. Zimmerman (Eds), Self-regulated Ieaming, from teaching to self-reﬂective
practice (pp. 20-41). New York: The Guilford Press.

Green, M. C., Brock, T. C., 8 Kaufman, G. F. (2004). Understanding media
enjoyment: The role of transportation into narrative worlds. Communication
Theory, 14(4), 311-327.

Gremler, D.D., Hoffman, K.D., Keaveney, SM, 8 Wright, L.K. (2000)
Experiential Ieaming exercises in services marketing courses. Joumal of
Marketing Education, 22(1), 35-44.

Grodal, T. (2000). Video games and the pleasures of control. In Zillmann, D., 8
Vorderer, P. (Eds), Media entertainment: The psychology of its appeal (pp. 1 97-
214). Mahwah, NJ: Lawrence Erlbaum.

Gros, B. (2007). Digital Games in Education: The Design of Games-Based
Learning Environments. Journal of Research on Technology in Education, 40(1),
23.

Hoffman, K.D., 8 Bateson, J.E.G. (1997). Essentials of services marketing. Fort
Worth, TX: Dryden.

Holyoak, K.J., 8 Thagard, P. (1995). Mental leaps: Analogy in creative thought.
Cambridge, MA: MIT Press/Bradford Books.

98

 

Jose, P. E., 8 Brewer, W. F. (1984). Development of story liking: Character
identification, suspense, and outcome resolution. Developmental Psychology,
20(5), 91 1-924.

Kafai, Y.B. (1994). Minds in play: Computer game design as a context for
children’s Ieaming. Mahwah, NJ: Lawrence Erlbaum.

Kim, Y., 8 Baylor, A. L. (2007). Pedagogical agents as social models to inﬂuence
Ieamer attitudes. Educational Technology, 47(1), 23-28.

Kim, Y., Baylor, A. L., Sheri, E., 8 PALS Group (2007). Pedagogical agents as
Ieaming companions: The impact of agent affect and gender. Journal of
Computer-Assisted Learning, 23(3), 220-234.

Kim, Y., Wei, 0., Xu, 8., K0, Y., 8 Ilieva, V. (2007, July). MathGiris: Increasing
Giris’ Positive Attitudes and Self-Efficacy through Pedagogical Agents. Paper

presented at the 13th International Conference on Artiﬁcial Intelligence in
Education (AIED 2007 ): Los Angeles, CA. July 9-13, 2007.

Kolb, D. A. (1984). Experiential Ieaming: Experience the source of Ieaming and
development. Englewood Cliffs, NJ: Prentice Hall.

Kwon, M. (1996, January). The use of planned comparisons in analysis of
variance research. Paper presented at the Annual Meeting of the Southwest
Educational Research Association: New Orleans, LA. January, 1996.

Lenders, D. M. (1975). Observational Ieaming of a motor skill: temporal spacing
of demonstration and audience presence. Joumal of Motor Behavior, 7, 281-287.

LaRose, R., 8 Rifon, NJ (2007). Online safety begins with you and me: Getting
lntemet users to protect themselves. Paper presented at the 57th Annual

Conference of the lntemational Communication Association: Montreal, Quebec,
Canada. May, 2007.

Lee, J.E.R., Nass, C., Brave, S.B., Morishima, Y., Nakajima, H., 8 Yamada, R.
(2007). The case for caring coleamers: The effects of a computer-mediated
coleamer agent on trust and learning. Journal of Communication, 57, 183-204.

Lee, T. D., 8 White, M. A. (1990). Inﬂuence of an unskilled model’s practice
schedule on observational motor Ieaming. Human Movement Science, 9, 349—
367.

Leifer, A.D., Collins, W.A., Gross, B.M., Taylor, P.H., Andrews, L., 8 Blackmer,

ER. (1971). Developmental aspects of variables relevant to observational
Ieaming. Child Development, 42(5), 1509-1516.

99

Lerner, RM, 8 Steinberg, L.D. (2004). Handbook of Adolescent Psychology.
New Jersey: John Wiley and Sons.

Lewis, M. L., Weber, R., 8 Bowman, N. D. (2008). They may be pixels, but
they‘re MY Pixels: Developing a metric of character attachment in role-playing
video games. CyberPsycho/ogy and Behavior, 1 1(4), 515-518.

Lieberman, D. (2001). Management of chronic pediatric diseases with interactive
health games: Theory and research ﬁndings. Journal of Ambulatory Care
Management, 24(1), 26-38.

Lieberman, D. (2006). What can we Ieam from playing interactive games? In P.
Vorderer 8 J. Bryant (Eds), Playing computer games: Motives, responses, and
consequences. Mahwah, NJ: Lawrence Erlbaum Associates.

Livingstone, S., Bober, M., 8 Helsper, E. (2005). lntemet literacy among children
and young people: ﬁndings from the UK Children Go Online Project [online].
London: LSE Research Online. Available at:
http://eprints.lse.ac.uklarchivel00000397.

Lockwood, P., 8 Kunda, Z. (1997). Superstars and me: Predicting the impact of
role models on the self. Journal of Personality and Social Psychology, 73(1), 91-
103.

Lovelock, CH, 8 Wright, L.K. (1999). Principles of service marketing and
management. Upper Saddle River, NJ: Prentice Hall.

Ludman, E., VonKorff, V., Katon, W., Lin, E., Walker, E., Bush, T. 8 Wahab, S.
(2000). The design, implementation, and acceptance of a primary care-based

intervention to prevent depression relapse. lntemational Joumal of Psychiatry in
Medicine, 30(9), 229-245.

Mackenzie, CS, 8 Peragine, G. (2003). Measuring and enhancing self-efﬁcacy
among professional caregivers of individuals with dementia. American Journal of
Alzheimer’s Disease and Other Dementias, 18(5), 291 -299.

Making History. (2007). Retrieved July 14, 2007, from http://wwwmaking-
history.comledufindex.php

Markman, AB, 8 Gentner, D. (1993). Structural alignment during similarity
comparisons. Cognitive Psychology, 25, 431-467.

Maudsley, G., 8 Strivens, J. (2000). Promoting professional knowledge,

experiential Ieaming and critical thinking for medical students. Medical Education,
34, 535-544.

100

 

May, S. A. (1994). The best in role-playing games. Compute, 16(1), 810-611.

Mayo, M. J. (2007). Games for Science and Engineering Education. Association
for Computing Machinery. Communications of the ACM, 50(7), 30.

Meng, X.L., Rosenthal, R., 8 Rubin DB. (1992). Comparing correlated
correlation coefﬁcients. Psychological Bulletin, 111(1), 172-175.

Miller, NE, 8 Dollard, J. (1941). Social Ieaming and imitation. New Haven: Yale
University.

Mori, M. (1970). The uncanny valley (K. F. MacDorman 8 T. Minato, Trans).
Energy, 7(4), 33-35.

Nielsen, J. (1997, October 1). How users read on the web. Alertbox. Retrieved
March 2, 2009, from http://www.useit.comlalertboxl9710a.html

Nielsen, J. (2006, April 17). F-Shaped pattern for reading web content. Alertbox.
Retrieved March 2, 2009, from
http://www.useit.com/alertboxlreading_pattem.html

Oatley, K. (1994). A taxonomy of the emotions of literary response and a theory
of identiﬁcation in ﬁctional narrative. Poetics, 23, 53-74.

Ohanian, R. (1989). Ego centrality as an indicator of enduring product
involvement. Joumal of Social Behavior and Personality, 4(3), 443-455.

Pollock, B. J., 8 Lee, T. D. (1992). Effects of the model’s skill level on
observational motor Ieaming. Research Quarterly for Exercise and Sport, 63, 25-
29.

Prensky, M. (2005). Computer games and Ieaming: Digital game- based
Ieaming. In J. Raessens 8 J. Goldstein (Eds), Handbook of Computer Game
Studies (pp. 97-122). Cambridge, MA: The MIT Press.

Quilliam, E. T., Rifon, N. J., LaRose, R. J., 8 Carlson, L. (2006). Losing my good
name: Teens, online privacy, and identity theft, Marketing 8 Public Policy
Conference. Washington, DC.

Reeves, B., 8 Nass, C. (1996). The Media Equation — How People Treat
Computers, Television and New Media Like Real People and Places. Cambridge:
Cambridge University Press.

Rice, J. W. (2007). New Media Resistance: Barriers to Implementation of

Computer Video Games in the Classroom. Joumal of Educational Multimedia
and Hypermedia, 16(3), 249.

101

 

Riddle, J. (2002). Cameras and Point-of-view in the gamespace. In
SIGGRAPH2002 Proceedings, ACM, 155. San Antonio, TX.

Rimal, R. N., 8 Real, K. (2003). Perceived risk and efﬁcacy beliefs as motivators
of change: Use of the risk perception attitude (RPA) framework to understand
health behaviors. Human Communication Research, 29(3), 370-399.

Rippetoe, P. A., 8 Rogers, R. W. (1987). Effects of components of protection-
motivation theory on adaptive and maladaptive coping with a health threat.
Journal of Personality and Social Psychology, 52(3), 596-604.

Ritterfeld, U., 8 Weber, R. (2006). Video games for entertainment and education.
In P. Vorderer 8 J. Bryant (Eds), Playing computer games: Motives, responses,
and consequences (pp. 399-413). Mahwah, NJ: Lawrence Erlbaum.

Ritterfeld, U., Weber, R., Fernandes, S., 8 Vorderer, P. (2004). Think sciencelz
Entertainment education in interactive theaters. Computers in Entertainment,
2(1), 1 1-11.

Rosekrans, MA. (1967). Imitation in children as a function of perceived similarity
to a social model and vicarious reinforcement. Joumal of Personality and Social
Psychology, 7(3), 307-315.

Schriver, K. A. (1992). Teaching writers to anticipate readers' needs: A
classroom-evaluated pedagogy. Written Communication, 9 (2), 179-208.

Schunk, D. H. (1987). Peer models and children’s behavioral change. Review of
Educational Research, 5 7, 149-174.

Schunk, DH, 8 Hanson, AR. (1985). Peer models: Inﬂuence on children’s self-
efﬁcacy and achievement. Journal of Educational Psychology, 77(3), 313-322.

Schunk, D. H., 8 Hanson, A.R. (1989a). Influence of peer-model attributes on
children’s beliefs and Ieaming. Journal of Educational Psychology, 81, 431-434.

Schunk, D. H., 8 Hanson, A.R. (1989b). Self-modeling and children’s cognitive
skill Ieaming. Journal of Educational Psychology, 81 , 155-163.

Schunk, D.H., Hanson, AR, 8 Cox, PD. (1987). Peer model attributes and
children’s achievement behaviors. Journal of Educational Psychology, 79(3), 54-
61.

Sherry, J.L. (2001). The effects of violent video games on aggression. Human
Communication Research, 27(3), 409-431.

102

Singhal, A.S., Cody, M. J., Rogers, E. M. and Sabido, M. (2004). Entertainment
education wortdwide: History, research, and practice. Mahwah, NJ: Lawrence
Erlbaum.

Singhal, AS, 8 Rogers, EM. (2002). A theoretical agenda for entertainment-
education. Communication Theory, 12(2), 117-135.

Sirgy. M.J. (1985). Using self-congruity and ideal congruity to predict purchase
motivation. Journal of Business Research, 13(3), 195-206.

Skinner, B.F. (1953). Science and human behavior. New York: Macmillan.

Slater, M. D. (2002). Entertainment Education and the persuasive impact of
narratives. In M. C. Green (Ed), Narrative Impact: Social and Cognitive
Foundations. Mahwah, NJ: Lawrence Erlbaum Associates.

Smed, J., 8 Hakonen, H. (2003). Towards a deﬁnition of a computer
game. TUCS Technical Report No. 553). Turku Centre for Computer Science.

Sonnenschein, S., 8 Whitehurst, G. J. (1983). Training referential communication
skills: The limits of success. Joumal of Experimental Child Psychology, 35, 426-
436.

Sonnenschein, S., 8 Whitehurst, G. J. (1984). Developing referential
communication: A hierarchy of skills. Child Development, 55, 1936-1945.

Spencer, J. (2003). Learning and teaching in the clinical environment. British
Medical Journal, 326(7389), 591-594.

SPSS, Inc. (2009), for Windows (Version 17.0) [Computer Software]. Chicago, IL:
SPSS, Inc.

Stanton, F., 8 Grant, J. (1999). Approaches to experiential learning, course
delivery and validation in medicine: A background document. Medical Education,
33, 282-297.

Subrahmanyam, K., 8 Greenﬁeld, PM. (1994). Effect of video game practice on
spatial skills in girls and boys. Journal of Applied Developmental Psychology, 15,
1 3-32.

Tafarodi, R. W., Mehranvar, S., Panton, R. L., 8 Milne, A. B. (2002). Putting

oneself in the task: Choice, personalization, and conﬁdence. Personality 8 Social
Psychology Bulletin, 28(5), 648-658.

103

Thomas, R., Cahill, J., 8 Santilli, L. (1997). Using an interactive computer game
to increase skill and self-efﬁcacy regarding safer sex negotiation: Field test result.
Health Education and Behavior, 24(1), 71 -86.

Thomas, J.R., Pierce, C., 8 Ridsdale, S. (1977). Age differences in children’s
ability to model motor behavior. Research Quarterly, 48, 592-597.

Tumer, M. M., Rimal, R. N., Morrison, D., 8 Kim, H. (2006). The role of anxiety in
seeking and retaining risk lnfonnation: Testing the risk perception attitude
framework in two studies. Human Communication Research, 32(2), 130-156.

Turow, J., 8 Nir, L. (2000). The intemet and the family 2000: The view from
parents, the view from kids. Philadelphia: The Annenberg Public Policy Center.

Vorderer, P., Klimmt, C., 8 Ritterfeld, U. (2004). Enjoyment: At the heart of media
entertainment. Communication Theory, 14(4), 388-408.

Vorderer, P., 8 Ritterfeld, U. (2003). Children’s future programming and media
use between entertainment and education. In E.L. Palmer 8 B. Young (Eds),
The faces of televisual media: Teaching, violence, selling to children (2"‘1 ed., pp.
241-262). Mahwah, NJ: Lawrence Erlbaum Associates.

Walshe, D.G., Lewis, E.J., Kim, 8)., O’Sullivan, K., 8 Wrederhold, B.K. (2003).
Exploring the use of computer games and virtual reality in exposure therapy for
fear of driving following a motor vehicle accident. CyberPsychology 8 Behavior,
6(3), 329-334.

Weber, R., Ritterfeld, U., 8 Kostygina, A. (2006). Aggression and violence as
effects of playing violent video games? In P. Vorderer 8 J. Bryant (Eds), Playing
computer games: Motives, responses, and consequences. Mahwah, NJ:
Lawrence Erlbaum.

Weeks, D.L., 8 Anderson, LP. (2000). The interaction of observational Ieaming
with overt practice: Effects on motor skill Ieaming. Acta Pscyhologica, 104, 259-
271.

Weiss, M.R., McCullagh, P., Smith, A., 8 Berlant, A. (1998). Observational
learning and the fearful child: Inﬂuence of peer models on swimming skill
performance and psychological responses. Research Quarteriy for Exercise and
Sport, 69, 380-394.

Wesch, N.N., Law, 8., 8 Hall, CR. (2007). The use of observational Ieaming by
athletes. The Joumal of Sport Behavior, 30(2), 219-231.

\Nirth, C.B., Rifon, N.J., LaRose, R., and Lewis, ML. (2007). Promoting teenage
online safety with an i-Safety intervention: Enhancing self-efﬁcacy and protective

104

behaviors. Paper presented at the Annual Meeting of the lntemational
Communication Association: Montreal, Quebec, Canada. May, 2007.

Wood, N.T., Solomon, MR, 8 Englis, 8.6. (2005). Personalisation of online
avatars: Is the messenger as important as the message? lntemational Joumal of
lntemet Marketing and Advertising, 2(1/2), 143-161.

Wood, R.E., Kakebeeke, B.M., Debowski, S., 8 Frese, M. (2000). The Impact of
enactive exploration on intrinsic motivation, strategy, and performance in
electronic search. Applied Psychology, 49(2), 263-283.

Wright, L.K., Bitner, M.J., 8 Zeithaml, VA. (1994). Paradigm shifts in business
education: Using active Ieaming to deliver services marketing content. Joumal of
Marketing Education, 16(fall), 5-19.

Yee, N. (2004). Unmasking the Avatar: The Demographics of MMO Player
Motivations, In-Game Preferences, and Attrition. Gamasutra. Retrieved March 1,
2009, from

http://wwwgamasutra.comlview/feature/Z1 39lunmasking_the_avatar_the_.php

Yoon, S. L., 8 Godwin, A. (2007). Enhancing Self-Management in Children with
Sickle Cell Disease Through Playing a CD-ROM Educational Game: A Pilot
Study. Pediatric Nursing, 33(1), 60.

Zimmerman, B.J., 8 Koussa, R. (1975). Sex factors in children’s observational
Ieaming of value judgments of toys. Sex Roles, 1(2), 121-132.

Zlmmerrnan, B.J., 8 Schunk, DH. (2001). Self-regulated Ieaming and academic
achievement: Theoretical perspectives. New Jersey: Lawrence Erlbaum
Associates.

105