392.;
1:33.
5.: .w

HF...“ nirva-

‘Qz‘lqfllt

bank

"a.
z . {5x1}. 9
[In ‘3 A! 1:...an- t
I 1
. u
$1.23..)
32.5.
will ‘

. {Pastor
am .1 :x...52w

1t
.. :an
;

WW3. .

’1qunonaI 341' e \G
,. >, .v. J.
.i. «($31: v3).

. P 2.3:
.u) 11.1.3
7... 3|... :1
:1). .i {.1331
42:. 3....

:2... t.
1.: .7... .§x.&l{»~ali..a.
2.1.371 .1, .i»..§l3.§r{fl ..
V. n H .121; an

.. u?! r...

i ‘ z 3;

 

T .
‘ : '9‘:
.‘r

h... l '
v.1 .511... v.
919; ‘

l
«iiid‘o

THESIS

,i’ a 0

 

LIBRARY

‘ Michigan State
University

 

 

 

PLACE 1N RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with eariier due date if requested.

 

DATE DUE DATE DUE DATE DUE

Ag. $132200?

2 0 A" " r 11115

J

‘ AER ! 1 2007
951‘ 0?

 

 

 

 

;\

 

 

 

 

 

 

 

 

 

 

 

mm W.“

The Influence of Anthropomorphism on Mental Models of Agents
and Avatars in Social Virtual Environments

By

Kristine Nowak

A DISSERTATION

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

DOCTOR OF PHILOSOPHY
Ph.D. Department of Telecommunication

2000

ABSTRACT

The Influence of the Anthropomorphism on Mental Models of Agents and Avatars
in Social Virtual Environments

By

Kristine Nowak

This project examined the influence of anthropomorphism on mental
models of agents and avatars in social virtual environments. This study used a
between subjects experimental design with two factors. The first factor, level of
anthropomorphism of virtual image had three levels, high, low and no virtual
image. The second factor, anthropomorphism of the intelligent other, had two
levels, whether the participants were told they were interacting with a human
(avatar), or a bot (agent). Participants consisted of 134 (94 males and 40
females) undergraduates at a large midwestern university. The results showed
that a virtual confederate (agent or avatar) represented by a high-
anthropomorphic image or no image was perceived to be more credible, Iikeable
and copresent in the interaction as compared to virtual confederate (agent or
avatar) represented by a low-anthropomorphic virtual body. There were no
discernable differences between the perception of virtual confederates when
participants were told they were interacting with an avatar as compared to those

told they were interacting with an agent.

I dedicate this work to my family and friends. They made it possible for me to
achieve so many of my goals. I couldn’t have made it without their patience and
support during the struggles and trials of life and especially those related to
graduate school. First to my family whose support has meant everything -
especially my Mother, brother and grandmother. Also to Dave Logie, my better
half and partner in life. He made me believe I could do anything and his
commitment and dedication has allowed us to survive the years in different states
and helped me keep perspective on life. To my friends who have always been
there and who I am so fortunate to have in my life especially Fatima Kascht and
Erin Hamilton. And to those that came into my life during my time in Lansing-
Brian Winn, Zena Biocca, Scott Connell, Kurt Besecker, Andy Kurtz, Trina
Anderson, Jenn Gregg, James Ramos, Alice Chan, Lynn Rampoldi-Hnilo,
Duncan and Stacy Rowland and so many others-Thank you for introducing me to
llamas and for the wonderful memories from the banks of the Red Cedar. All of
you encouraged and supported me to accomplish things I may not have
attempted on my own.

ACKNOWLEDGMENTS

With any project of this magnitude, it could not have been accomplished
without the assistance of a number of people. First, I could not have done this
without my advisor, mentor and friend, Frank Biocca who supported and
encouraged me through every step of the Ph.D. and especially through the
dissertation process. I would also like to thank the rest of my committee, Carrie
Heeter, Kelly Morrison and Sandi Smith for their suggestions, comments and
support. For the set up, design and creation of stimulus materials, I would like to
thank all the staff and interns at the MIND Lab, but especially Eric Maslowski,
Duncan Rowland, Kwok Hung (Arthur) Tang and Joe Berger. There are a
number of other faculty members who contributed to my positive academic
experience at Michigan State, including Charles Salmon, Chip Steinfield, Pam
Whitten and many others. I would also like to acknowledge Steven Chaffee

because without him, I never would have attempted graduate school.

TABLE OF CONTENTS

INTRODUCTION ........................................................................................ 1
CHAPTER 1: TYPES OF SOCIAL VIRTUAL ENVIRONMENTS AND
BODIES ...................................................................................................... 5
Influence of mediation and context .................................................. 5
Virtual Reality and Presence ........................................................... 6
Social Virtual Environments and Copresence ................................ 8
Copresence ................................................................................... 9
The Virtual Embodiment: Anthropomorphic Trends ....................... 12
Agents: Embodied ”bots" ............................................................ 15
Avatars: Humans represented by virtual bodies .......................... 16
CHAPTER 2 FORMING MENTAL MODELS OF OTHERS ................................. 17
Uncertainty Reduction ................................................................... 18
The Categorization Process in the natural world ........................... 19
Defining Humanity: Living but not Animal ................................... 21
Physical Characteristics Used for Categorizing Others ............... 23
Social Judgment from mental models ........................................... 25
Social Attraction .......................................................................... 26
Partner Satisfaction ..................................................................... 27
Confidence of Judgment ............................................................. 28
The Utopian Promise and Reality of interactions in SVEs ............. 29
Categorization Process in SVEs .................................................. 31
The virtual image influences perception ..................................... .34
The Virtual Environment and Context for Interaction ................... 36
Responding to anthropomorphic others ......................................... 40
CHAPTER 3 THEORY OF PERCEIVING ANTHROPOMORPHIC OTHERS .......... 52
Mental Models of Others ............................................................... 53
Effect of Humanity on social judgment of interactant .................... 54
The relationship of Presence and copresence .............................. 59
Effect of embodiment on social judgment of interactant ................ 60
More anthropomorphic virtual IMAGES generate more positive
social Judgments of human and non-human interactants ............. 63
CHAPTER 4 METHOD ............................................................................... 66
Design ........................................................................................... 66
Participants .................................................................................... 66
Stimulus Materials ......................................................................... 67
Measurement Instruments ............................................................. 69
Procedure ...................................................................................... 72
CHAPTER 5 RESULTS .............................................................................. 74
CHAPTER 6 DISCUSSION ......................................................................... 91
On the difference between human and non-human interactants ...91
On the Existence of a Virtual Image .............................................. 93
Not all virtual images are equal ................................................... 93

The feelings of Co-presence and presence are interconnected 95

On Revising the Model and Understanding the influence of the

Virtual Image ................................................................................. 96
Potential Limitations of This Manipulation ................................... 100
Features of the Interface ............................................................ 101
Issues Related to the Interaction................................................101
Features of the Virtual confederate (agent or avatar) ................ 102
The uniqueness of the Virtual Images ................................ 102
The Voice and script .......................................................... 103
Issues related to the manipulation and lab setup ....................... 104
The Agent/Avatar Manipulation ................................................. .106
Environmental Issues: Lab Setup 107

Interacting with Intelligent others in SVEs; Implications of Findings
for Interface Designers and Users ............................................... 107
CHAPTER 7 CONCLUSION ...................................................................... 110
APPENDIX A: AUDIO SCRIPT ....................................................................................... 133
APPENDIX B: PRE- EXERIMENT QUESTIONNAIRE ................................................... 135
APPENDIX C: POST EXPERIMENT QUESTIONNAIRE ............................................... 138
APPENDIX D: AVATAR CONDITION INSTRUCTION SHEET ...................................... 148
APPENDIX E: AGENT CONDITION INSTRUCTION SHEET ........................................ 151
BIBLIOGRAPHY .............................................................................................................. 154

vi

 

LIST OF TABLES

Table I: Types of Virtual Bodies and Indicators Available ....... 114

Table II: Summary table of Hypotheses .................................... 115
Table III: Design of Experiment ................................................. 116
Table IV: Social Attraction Scale ............................................... 117
Table V: Partner Satisfaction Scales ........................................ 118
Table VI: Perceived copresence of the virtual confederate scale119
Table VII: Self reported copresence scale .................................. 120
Table VIII: Uncertainty scale ........................................................ 121
Table IX: Presence Scale .......................................................... 122

vii

 

Figure 1:
Figure 2:
Figure 3:
Figure 4:
Figure 5:
Figure 6:
Figure 7:
Figure 8:

 

LIST OF FIGURES

Theoretical Model on Influence of Anthropomorphism ...................... 124
No Image Condition Scene condition ................................................. 125
Virtual Representation Selection Screen ........................................... 126
High Anthropomorphic image Condition Scene ................................. 127
Low-anthropomorphic image Condition Scene .................................. 128
Results table: Influence of Agency ..................................................... 129
Results table: Influence of Image ....................................................... 130
Revised Theoretical Model on Influence of Anthropomorphism ......... 131

viii

 

LIST OF ABBREVIATIONS

CMC= Computer Mediated Communication
SVEs=Social Virtual Environments

 

__—.._ ‘24-.-

INTRODUCTION

Recent technological advances have led to some important changes in
interfaces, including the way the human body is connected to, immersed in and
represented by the computer, as well as the connection of the user or human to
the computer. These changes in the interfaces have led to subsequent influences
on both human-computer interaction as well as computer mediated interpersonal
communication. The proliferation of technology has led to increasingly social and
business related uses of technology with others around the world (Chesebro &
Bonsall, 1989; Rice & Love, 1987; Siegel, Dubrovsky, Kiesler, & McGuier, 1986).
These social and work related uses of technology have led to the question of
how people perceive the "other” in mediated interactions (Lea & Spears, 1992;
Palmer, 1995; Rice, 1993; Rice & Tyler, 1995; Short, Williams, & Christie, 1976;
Steinfield, 1986; Walther, 1996). This research project explored one such
technological advance- the increased anthropomorphism of virtual bodies in
cyberspace and artificial intelligences. Specifically, this project tested the
influence of anthropomorphism of virtual ”others” on both computer mediated
interpersonal interaction as well as human computer interaction.
Anthropomorphic refers to having human-like qualities (such as intelligence) or
physical features (appearing human).

People have automatically responded socially to computers and other

intelligent entities and anthropomorphic interfaces, and anthropomorphism has

enhanced this tendency, whether the interfaces were anthropomorphic in terms
of displaying intelligence or in terms of appearance (Reeves & Nass, 1996). The
anthropomorphism of interfaces has increasingly personified interfaces and
technology and led to increased social responses to computers (Koda, 1996;
Reeves & Nass, 1996; Turkle, 1995). There has been a great deal of concern
about how this tendency to respond socially to computers will influence people's
ability to distinguish humans from computers and to successfully function in the
approaching age of networked computers (Don, 1992; Laurel, 1990; Reeves &
Nass, 1996; Turkle, 1995).

In the natural world‘, humans perceive the features of a person or object
they encounter in the environment in order to place them in categories that have
been meaningful in previous experience in the world (Bruner, 1957; Lakoff,
1987). They also use the characteristics of things to determine whether an object
is living or not and, if living, whether human or not. Following assignment to this
first category, the process of perceiving the object becomes based on
characteristics associated with either inanimate objects, living things or with
human living things (Konner, 1991; Sheehan, 1991a; Sheehan, 1991b; Turkle,
1991 ).

In the process of perceiving humans, the categorization process has relied
on indicators provided by the first attributions perceivers can make given their

physical and psychological salience and are believed to reflect fundamental

 

' The natural unmediated world to be contrasted with the virtual, mediated world.

 

 

distinctions in social behavior (Fiske & Neuberg, 1990; Hamilton & Sherman,
1994). In the natural world, people have primarily used the features of the body
to assign people to membership in categories (Argyle, 1975; Chesebro &
Bonsall, 1989; Fiske & Neuberg, 1990; Goffman, 1963; Hinton, 1993; Ichheiser,
1970). Although many of these processes are automatic and unlikely to change
in SVEs, the very nature of the mediated information is likely to influence the
processing of information and subsequent perception of the other.

The first and most salient influence of the concept of anthropomorphic
others in social virtual environments (SVEs) is that it has served to add a level of
complexity to the question of perceiving others. First, humans have been
represented by virtual bodies (avatars) in SVEs that did not reflect features of
their natUral2 physical bodies (Benford, Greenhalg, Bowers, Snowdon, & Fahlén,
1995; Damer, 1997; Suler, 1996). This makes the process of forming mental
models of others even more difficult than in the natural world. Second, in virtual
worlds, one cannot rely on physical characteristics to distinguish between and
among people and objects. Finally, recent advances in artificial intelligence work
and increasingly anthropomorphic visual representations, entities in SVEs can
speak and move in ways that make it difficult to know if the interaction one is

having is with another human or with a bot3. In the pages that follow, these

 

2 Natural means real, or the body one was born with. This is contrasted with a virtual body that
only exists in cyberspace.

3 An autonomous computer program, allowed to autonomously interact with other people and bots
in the environment.

issues are considered and some of the original predictions and assumptions
about anthropomorphic interfaces are questioned.

To begin, the categorization process humans go through when perceiving
others will be examined, followed by the way these processes are likely to port“,
or translate, to social virtual environments. The pages that follow explore
literature relevant to the topics introduced above, including an exploration of
recent trends in the design of virtual bodies as well as the processes of forming
mental models of others. It then moves on to develop a theory predicting the
likely influence of anthropomorphism5 on the perception of virtual bodies and
intelligent others (whether humans or bots) in social virtual environments (SVEs).
The potential influence anthropomorphism is considered in two ways; those
traditionally used to consider the influence of virtual environments as well as
those traditionally used to measure people’s perceptions of others. Those
considered in the pages that follow include copresence with their interaction
partner and presence in the environment as well as on perceptions of social

attraction and partner satisfaction (See figure I).

 

‘ In computer science, port means to transfer from one platform to another. For example, one can
port a program from a PC to a Mac but the two programs may not run identically.

5 Anthropomorphic means more like human, or having human-like qualities (such as intelligence)
or physical features (such as looking human).

Chapter 1

Types of Social Virtual Environments and Bodies

This chapter defines and explains the influence of mediation and context
on people's ability to make mental models of others. It then considers what
makes an environment virtual in terms of virtual reality and the defining sense of
presence. It defines what makes a virtual environment a social virtual
environment and the corresponding sense of copresence with an intelligent
other. It briefly explores the phenomenon of interacting in social virtual
environments including providing examples of different types of SVEs. The
chapter concludes by explaining the difference between agents and avatars and
the potential implications of this distinction for perceiving intelligent others in
social virtual environments

INFLUENCE OF MEDIATION AND CONTEXT

McLuhan (1964) argued that “the medium is the message,’ and that each
new media technology extends our senses further and in new and different
directions than the technology before it. He was essentially arguing that the
medium chosen for an interaction altered or redefined the message. Similarly,
Davis (1995) argued "it is not possible to appreciate the full meaning of a
message without knowing the 'language' of the technology and the culture that

helped develop that technology" (p. 537). This perception has never been more

insightful and potentially powerful than when one considers interface design and
communication in virtual environments.

In considering the influence of computer interactions, especially in SVEs, it
is important to consider not only the medium and its features but also the context
of the interaction as well as the interface. The interface is the connection of the
user to the computer world including the way the user has been connected to,
immersed in, or represented by the computer. The interface is the part of the
medium that influences how our senses are exposed to the message. The
different interfaces present information to a variety of senses in different ways.
These differences are similar to media differences in that they influence not only
the presentation of information, but also how the information is processed and
what is remembered (Drew & Grimes, 1987; Graber, 1990; Katz, Adoni, &
Parness, 1977; Lang, 1995). This means it is likely that the interface will influence
the presentation of information and interactions in the environment as well as
how people are perceived and categorized.

VIRTUAL REALITY AND PRESENCE

This section defines what a virtual environment is, as well as the
corresponding sense of presence. Many technologies are defined by particular
features or physical properties of the medium, or interface. For example, a
television has a screen and displays programming. Virtual reality has not been
defined in terms of particular features of the medium, the interface, or even in

terms of its function. Virtual reality instead has been defined in terms of the

experience or perception of the user of the system. A group of input and output
devices would be considered virtual reality if they responded appropriately to the
input of the user (be interactive) and if they provided a sense of presence in the
environment (see Biocca, 1997; Biocca, Nowak, & Lauria, 1997; Steuer, 1994).
The term virtual reality would be misleading if one assumed it would accurately
reflect the properties of the natural world, or reality. Some of the most interesting
uses of virtual reality have been in its ability to provide experiences that would
not be possible in the natural world, such as augmented reality or a simulation of
possible future events.

Presence has been defined in terms of the sensation of ‘being’ in the
virtual or mediated environment; a feeling that one has left the confines of the
natural world and entered or become immersed in the virtual world. Biocca
(1997) explained, "Users experiencing presence report having a compelling
sense of being in a mediated space other than where their physical body is
located” (p. 9). So virtual environment would be an environment that provides a
sense of presence. An environment was virtual if an individual feet a sense of
presence in the environment (Biocca, 1997; Biocca & Nowak, 1999b; Biocca et
al., 1997; Lombard & Ditton, 1997).

Now that virtual reality and the corresponding sense of presence have
been defined, the next section explains the characteristics that make a virtual

environment social.

SOCIAL VIRTUAL ENVIRONMENTS AND COPRESENCE

As with virtual reality, a social virtual environment has been defined by the
sense or perception of the user(s). There is no technological or physical definition
of social virtual environments and no technology or group of technologies by
themselves could be definitional of social virtual environments. The word social
adds a dimension to the definition of a virtual environment in that it must have
more than one person interacting in it. An environment is a social virtual
environment when two or more individuals are interacting, are not in the same
physical space in the natural world, and feel present in the same virtual
environment. Technically, this could even include phone conversations but
usually this implies communication where individuals feel they are in the same
virtual place, resulting in a sense of copresence as defined below. By this
definition, a social virtual environment can be a place where people are
interacting in real time, but do not require it to be a social interaction (although it
can be) It also includes interactions where people are working together in more
task-oriented interactions.

Mediated interactions have had the capacity to provide a very strong and
distinct sense of other people. Social virtual environments have allowed people
to form very supportive and close friendships with others they have encountered
(see Fisher, 1997; Parks 8 Floyd, 1996; Parks & Roberts, 1997; Turkle, 1995).
People who have frequently interacted in SVEs have objected to the notion that

their interactions in cyberspace are less than real; that their communities are

’pseudo-communities’ as Beniger (1987) referred to them. When people began
interacting in chat rooms, experiences outside the chat room were referred to as
IRL (In Real Life). Now that their experiences in cyberspace are becoming more
meaningful to them, the use of “IRL” and the use of the term ”real” to refer to
experiences and relationships that exist outside (as opposed to those inside)
virtual environments has become objectionable6 (see Fisher, 1997; Turkle, 1995;

Watson, 1997).

COPRESENCE

 

The term copresence originated in he work of Goffman (1963), who
explained that copresence existed when people sensed that they were able to
perceive others and that others were perceiving them. Further, he explained that
in its true meaning, “copresence renders persons uniquely accessible, available,
and subject to one another” (Goffman, 1963, p. 22), or when people were able to
find the other ‘within range (Goffman, 1963).’ Ciolec (1982) also considered
copresence and emphasized the importance of attention or responsiveness to
others. He explained that it is not only necessary for one to be within the
observable range of another, but also the observer must be aware of their activity
within their sensory zones to achieve copresence (Ciolek, 1982). Copresence as

is a combination of a number of concepts and it is likely to be multidimensional. It

 

5 This is why the term "natural" world is used instead of instead of real world.

 

was divided into two different constructs for measurement in this project, as
discussed below.

The sense of copresence shares some concepts with interpersonal
constructs of intimacy, involvement and immediacy and this section considers
how copresence is similar to and distinct from these constructs. Conversational
involvement has been defined as “the degree to which participants are
enmeshed in the topic, interpersonal relationship, and situation” (Coker &
Burgoon, 1987, p. 463).

An important and unique characteristic of immediacy has been that it
involves a combination of nonverbal and verbal behaviors working together to
increase or decrease the degree of physical, temporal, and psychological
closeness between individuals (Burgoon, Buller, & Woodall, 1996). Further,
immediacy has been used as a way to measure the intensity and directness of
the interaction between entities (Mehrabian, 1967). lmmediacy has traditionally
relied on a system of visual nonverbal behaviors, with each behavior co-
occurring and becoming only a small part of a large system of immediacy
behaviors (Burgoon et al., 1996). The behaviors that are perceived to
communicate immediacy have carried multiple meanings, and more than one of
the nonverbal cues has usually functioned to convey the same sentiment so the
different cues have reinforced the message of immediacy in the relationship or

interaction (Burgoon, 1991).

10

Although copresence could be seen to conceptually share some of these
issues, there are important distinctions. Mehrabian (1967) argued that when
immediacy is adapted to nonverbal communication, it could be considered
analogous to the concept of proxemix. First, copresence does not necessarily
include this concept of proxemix or nonverbal intimacy. Instead, it solely refers to
a psychological connection to and with another person.

Second, copresence requires that interactants feel they were able to
perceived their interaction partner, and that their interaction partner actively
perceived them. This reciprocal nature of the construct makes it unique and
interesting and creates the dual nature of the concept.

Copresence is also distinct from conversational involvement because
involvement has been operationally defined in terms of physical behavioral cues
that indicate people’s involvement in the conversation (see Coker & Burgoon,
1987). Furthermore, involvement focuses more on the relationship and the
situation, and not necessarily on whether or not one is attending to their
interaction partner. These distinctions make copresence an ideal way to ask
about the mental connection between people not in the same geographic
location, such as communicators using telecommunication technology. The
nature of copresence includes both the extent to which the participant feels
involved in the interaction as well as the extent to which they perceive their

partner is involved in the interaction.

11

This section considered the notion of virtual environments and presence,
social virtual environments and copresence. The next section considers the types
of virtual bodies and environments that can provide this sense of copresence and
how the various interfaces, environments and embodiments (especially the
current trends toward more anthropomorphic virtual bodies) are likely to influence
the processing of information and person perception.

THE VIRTUAL EMBODIMENT: ANTHROPOMORPHIC TRENDS

This section begins with a look at different types of virtual environments
and discusses the implications of recent technological advances on people’s
interactions in these environments.

The first virtual places and environments were based on text only
interactions. Users of these early SVEs could “enter" worlds existing only in text-
based descriptions of the surroundings and people in the “room”. When people
signed on, they typed in whatever information they wanted others to know about
themselves. This traditionally included a physical description and demographic
information, although this represented what the person wanted others know and
did not necessarily reflect their appearance or existence in the natural world.
Although these text only interactions and virtual environments are still very
popular (Schiano, 1999), technological innovation has given rise to new types of
worlds with visual images and characters moving through graphical virtual
environments which continue to grow in popularity (Cassell & Vilhjalmsson, 1999;

Damer, 1997). One factor that is consistent across these environments is that

12

people still choose the description or image others have of them, whether the
source of it is text based or resulting from a graphical image (Paulos & Canny,
1997)

Virtual bodies and images7 have been seen in all visual virtual
environments in one form or another. The increased use of virtual bodies may
reflect a recognition of the importance of the different communicative functions of
the body during an encounter (Cassell & Vilhjalmsson, 1999). The virtual
representations have been images created or designed by the user8 with little or
no resemblance to the user’s natural physical embodiment (Benford et al., 1995;
Damer, 1997; Suler, 1996). Theoretically, these environments could allow people
to interact and “exist in a world free from earthly physical constraints” (Paulos &
Canny, 1997). Individuals know they may not have an accurate picture of their
interaction partner’s natural embodimentg, yet, people have formed impressions
of others, as well as relationships and friendships in these environments (Parks &
Roberts, 1997).

The morphology, movement, and behavior of the virtual humanoid image
are far more subject to design and manipulation than at any point in the past (for
a discussion of this, see Badler, Phillips, & Webber, 1993; Benford et al., 1995;

Capin, Pandzic, Magnenat-Thalmann, & Thalmann, 1998; Damer, 1997;

 

’ Not all virtual characters have bodies, some are just faces and others represent inanimate
objects.

° There are ’stock' avatars supplied by most worlds and users can buy an avatar created by
someone else.

9 Natural embodiment is the body one is born with. It is contrasted with virtual embodiment or
avatar.

13

Magnenat-Thalmann, Kalra, & Escher, 1998). The design of effective and
interesting virtual bodies has required an examination of how various cues
generated by virtual bodies influences the perception of the “other.” The types
and levels of virtual bodies have been as varied and diverse as the people who
were embodied by them. The images vary on a number of levels, from
representations of people (highly anthropomorphic) to animals or even inanimate
objects (low-anthropomorphic). Their virtual images have also varied from
unmoving two-dimensional pictures of a character to fluidly moving 3-D
embodiments that walk, fly or float through the environment in a variety of ways.
It is likely that choice of image type as well as the interface properties and
background environment would influence both the perception of the other and the
processing of information.

An issue that may potentially lead to interesting differences in the process
of person perception in SVEs is that there are no longer physical differences
between a computer or a bot and a human interacting in cyberspace. Both a
human and a bot could be represented by a highly anthropomorphic image or by
an image that is less anthropomorphic, anything from a human-like character to a
rock or a box. A bot could easily be represented by the same virtual image as a
human. Although physical differences may not be apparent in virtual
environments, terms have evolved to express this difference. A "bot" is an agent

and a human is an avatar.

14

AGENT: EMBODIED “BOTS”

An agent has been defined as a computer program that is designed to
interact with, or on behalf of a human. Agents have been capable of autonomous
decision—making, and even ‘learning’ with an algorithm. Those capable of
achieving goals autonomously are ‘autonomous agents’ (Franklin, 1997) or
‘intelligent agents’ (Hedberg, 1996). An agent could also be defined as a
character enacted by the computer who would act on behalf of a human, like an
assistant in a virtual environment (Petrie, 1996).

Agents have been used for everything from “managing mundane tasks like
scheduling to handling customized information searches that combine both
filtering and the production (or retrieval) of alternative representations to
providing companionship, advice, and help throughout the spectrum of known
and yet-to-be-invented interactive contexts” (Laurel, 1990, p. 356). The term
agent means a “bot” which indicates intelligence and autonomous behavior. In
this meaning, an agent may have a virtual image, which is not necessary to
traditional definitions. But Franklin (1997) argued, an agent with no image can be
intelligent, but, with no embodiment they cannot be situated in their environment.
The situated body is essential to the concepts explored in the pages that follow.

The same is true for avatars, which will be defined in the next section.

15

AVATAR: HUMANS REPRESENTED BY VIRTUAL BODIES

 

The word avatar originated in Hinduism. In Hindu, an avatar has been
defined as an incarnation or the embodiment of a deity or a spirit in an earthly
form. This Hindu notion is part of the belief that when creators of the universe
desired to experience earth from the perspective of its inhabitants or to speak
with mortals, they were presented as a material body (Vilhjalmsson, 1996). In
SVEs, an avatar10 has been defined as a physical or graphic image that allows
the user to be embodied by a graphical representation in a virtual environment.

This chapter explained how the medium has influenced message
perception and defined the term social virtual environment (SVE). It outlined the
types of virtual worlds and images that have been seen in them. It explained that
both agents and avatars are embodied characters and the difference between
them was explained as whether the entity behind the character was a human or
an autonomous computer program.

The following sections will continue this exploration but expand the issues
to consider the influence of these differences on people’s perception of virtual
others in SVEs. It examines cognitive processes in depth to ask if the influence of
the virtual image and the message is strong enough to compensate for the

difference in perceiving an agent and an avatar.

 

‘° Neil Stephenson is credited with originally using the term in this way in his science fiction book,
SnowCrash.

16

Chapter 2

Forming Mental Models of Others

This section briefly introduces the process of categorization in the natural
world and how humans assign other people and objects to categories. It is
important to understand that people process each object that enters their range
of senses to varying degrees while navigating a complex and ever changing
world (Fiske & Neuberg, 1990; Lakoff, 1987; Moore & Cavanagh, 1998). Familiar
objects and objects similar to things encountered previously would be classified
as “similar to” other things that already have already been processed. In this way,
categorizing familiar objects would require fewer processing resources than
categorize new or unique objects. New or unique objects must be processed
individually, but humans have become very good at using the physical
characteristics of an object to perceive not only what a new object is similar to,
but also what it is different from.

Classifying and perceiving people requires the same process as
perceiving objects, but activates different categories (Asch, 1958; Hastorf,
Schneider, & Polefl<a, 1970; Heider, 1958; Sheehan, 1991b). People categorize
others into familiar groups and conserve processing resources that free them to
navigate the environment (Fiske & Neuberg, 1990). After a brief consideration of

these issues, this section turns to some of the features included in and potentially

17

unique to a mental model of another human. Finally, the implications of these
processes and perceptions for human’s ability to function effectively with humans
and objects in SVEs are considered, including a discussion of salient categories
and indicators in these environments as well as the likely influence of
anthropomorphism.

UNCERTAINTY REDUCTION THEORY

People go through a number of processes when perceiving people and
objects in the world. The purpose of this section is to introduce a metatheory to
explain why people engage in the processes of categorization and social
judgment of others. This process has been explained by uncertainty reduction
theory. This theory was based on the belief that when strangers meet, their
primary goal is to create causal structures to explain their behavior and the
behavior of others (Berger & Calabrese, 1975). In essence, people’s primary goal
would be to reduce uncertainty about people, to allow them to better predict
future behavior as well as better understand motives and context for current or
past behavior (Berger & Calabrese, 1975; Infante, Rancer, & Womack, 1997).

Uncertainty reduction theory was originally conceived of by Berger and
Calabrese (1975) and began with a series of axioms and theorems about how
people experience strangers during initial interactions. The level of uncertainty
about a stranger is largely influenced by the context of the interaction and people
like others more when uncertainty is reduced (Berger & Calabrese, 1975). The

reduction of uncertainty is axiomatically defined to lead to interpersonal attraction

18

(Berger & Calabrese, 1975; Clatterbuck, 1979). Uncertainty can only be reduced
when people feel they have enough information to make desired judgments
within an interaction, In other words, it is not a sum total of information but only
the perceived quality of the information that influence uncertainty of judgment
(Berger & Calabrese, 1975; Clatterbuck, 1979). Thus, it is likely that interactions
in social virtual environments will be a highly unfamiliar and uncertain context
which will lead to reduced certainty and lower liking of strangers in SVEs,
especially for people with little or no experience in these environments.

Berger and Calabrese (1975) argued that there were three phases
involved in these initial interactions. The entry phase, which included the
exchange of demographic information. In the next phase, people began to make
judgments of values. Finally, during the exit phase, people decided whether or
not to pursue future meetings and signal the end of the interaction with certain
behaviors, as necessary (Berger & Calabrese, 1975).

THE CATEGORIZATION PROCESS IN THE NATURAL WORLD

"People categorize other people. social roles, social events, and even themselves to simplify the
complex task of interpersonal understanding. Categorization is the process of identifying a
stimulus as a member of its class, similar to other members and dissimilar from nonmembers.
Categories, then, are cognitive structures that contain instances of the class. Categorizing an
instance allows the perceiver to apply to the instance general knowledge and expectations about
that category without having to ascertain that those features indeed apply to that instance" (Fiske
8- Pavelchak, 1986, p. 170).

An understanding of how humans categorize people and objects is central
to any understanding of how they think and function (Lakoff, 1987, p. 6). The
process of categorization begins when an object enters a person’s range of

senses. The first step would be to perceive the characteristics of an object and

19

place it into categories that fit with a person's knowledge based on the traits and
characteristics that past experiences have shown them to be meaningful and
useful (Hamilton & Shaman, 1994; Lakoff, 1987; Wyer & Carlston, 1994). In this
way, the memories of past experience with people and objects have enabled
humans to interpret new people and objects they encountered in terms of
previously acquired concepts and knowledge (Fiske & Neuberg, 1990; Moore 8.
Cavanagh, 1998; Rock, Schreiber, & R0, 1994; Ullman, 1998; Wyer & Carlston,
1994). Categorization has allowed people to apply expectations about the
category without ascertaining that these features apply to the instance, thus
allowing the perceiver to make inferences that go beyond the information
available (Bruner, 1957; Fiske & Pavelchak, 1986).

There is little difference between the nature and process of the categorical
inferences based on defining attributes used in object recognition and the
process of forming mental models of living objects, especially another human
(Bruner, 1957; Fiske & Neuberg, 1990; Srull, Lichtenstein, & Rothbart, 1985).
The difference in perception does not exist in the process, but in the range of
categories that become accessible when a living thing is encountered in the
environment as compared to encountering an inanimate object.

Further, there is a distinction between categories that becomes accessible
for inanimate objects and all other living things and humans. When perceiving
other humans, people must consider intentionality and emotion as well as

behavior and features (Asch, 1958; Heider, 1958; Ritvo, 1991; Sheehan, 1991b;

20

Turkle, 1991; Williams, 1991). There are distinctive properties of living things that
do not apply to non-living things. For example, living things reproduce, move
autonomously through the environment and have physical properties with
specific uses to serve them (Keil, 1994). In other words, there would be more to
the experience of perceiving other people than to the experience of perceiving

objects (Hastorf et al., 1970). Consider the scenario explained by Heider (1958)

Let us assume that we enter an unfamiliar room for the first time and in it we find a few people we
have never met before. A glance around the room will suffice to get an approximately correct idea
of the shape of the room and of the objects in it. We shall be much more insecure about our
judgments of people (p. 23).

DEFINING HUMANITY: LIVING BUT NOT ANIMAL

 

Humans have long desired to distinguish themselves from both objects
and animals, defining humanity as distinct from, if not superior to, all other things
whether living or not (Sheehan, 1991a; Sheehan, 1991b). The human experience
in the natural world has led to the perception that if it displays intelligence, it is
human. This notion, that intelligence and social interaction are reserved for
humans, has been considered diagnostic of humanity.

However, the rise of artificial intelligence and experience with intelligent
agents may lead to a reconceptualization of this definition of humananity. Each
technological advance has led to a reconsideration of what is diagnostic of
humanity, with people struggling to define humanity in terms of behaviors that
computers cannot do, trying to stay ahead of the latest computers that can

perform these diagnostic behaviors (Bolter, 1984; Turkle, 1991).

21

Recent technological advances have begun to push the next
reconsideration and yet another paradigm shift including new diagnostic
behaviors. It may be time to reconsider the notion that behaviors can be
diagnostic of humanity and perhaps turn to emotions and feeling for diagnostic
indicators, but there we risk an inability to distinguish humanity from animals
(Ritvo, 1991; Sheehan, 1991a; Turkle, 1991). Some continue to define humanity
as distinct from either animal or computer by discussing the “spark” of people
along with the notion that to be human is to be unprogrammable (Turkle, 1991).
With "romantic machines” able to ”organize themselves to produce intelligent
behavior" and reveal a complexity of behavior, computers may not be far from
expressing something very similar to or indistinguishable from human emotion
(Turkle, 1991, p. 226). Even Hampshier’s (1991) claim that the distinction
between humans and computers is that humans have bodies and computers do
not is becoming less true as computers can have virtual bodies, although his
argument that they do not understand drives such as hunger and thirst may still
be distinguishing characteristics.

Regardless of what one considers diagnostic of humanity, there are a
number of categories that are relevant and accessible when perceiving an
intelligent other, many of which will also be relevant when interacting with an
agent. On the other hand, there may be categories that are salient only when

perceiving a non-human other. Future experience with intelligent others who are

22

not human may give rise to categories brought about only when interacting with
an intelligent other that is not human.

A main distinction between humans and all other things, whether living or
not, has been intelligence and autonomy, but current advances in artificial
intelligence are making that distinction increasingly difficult. The defining
distinctions for humanity vary with the thing we are distinguishing humans from.
Emotion may distinguish humans from computers, but it has been the ability to
self-consciously reflect and rationality that has distinguished humans from
animals (Konner, 1991).

The next section considers the question of how humans acquire such
information about other people and the implications of these problems peculiar to
interpersonal cognition when nonhuman entities look and act as humans and
may be perceived to have intent and motives in their behavior.

PHYSICAL CHARACTERISTICS USED FOR CATEGORIZING OTHERS

 

In the natural world, it has been the physically manifested and 'relatively
unchanging’ features that were of primary importance in the process of
categorizing other people (Argyle, 1975; Chesebro 8 Bonsall, 1989; Fiske 8
Neuberg, 1990; Goffman, 1963; Hinton, 1993; lchheiser, 1970). The primary
indicators have included attractiveness, facial expressions, posture, gestures,
and body movement among others (Argyle, 1975; Argyle, 1988b; Benford et al.,
1995; Berscheid 8 Walster, 1969; Bull 8 Rumsey, 1988; Ekman 8 Friesen, 1969;

Ferrari 8 Swinkles, 1996; Knapp, 1975; Mehrabian, 1972; Otta, Abrosio, 8

23

Hoshino, 1996). The process of categorizing others based on physical
appearance characteristics has appeared to be universal and automatic,
requiring little or no cognitive effort (Patterson, 1995). As Tagiuri (1958) argued,
“this evaluation of other persons, important as it is in our existence, is largely
automatic, one of the things we do without knowing very much about the
‘principles’ in terms of which we operate” (p. ix) Also, these automatic judgments
based on physical appearance and nonverbal behavior have been very accurate
(Ambady, Hallahan, 8 Rosenthal, 1995; Patterson 8 Stockbridge, 1998).

Bruner (1957) termed this process of categorization ”model building." In
this process, people have made an automatic assessment of how much
information they need about the person or object to fulfill their goals of the
interaction, compare it to the information they have, and then made every effort
to fill the gap between information they need and information they had. The
process of creating and evolving this model of others has been largely based, at
least initially, on categorization and stereotyping (Bruner, 1957; Fiske 8 Neuberg,
1990; Hamilton 8 Sherman, 1994; Lakoff, 1987). People evaluated the context of
a meeting, which included information about location and the goals of the
interaction as well as the perceivers available cognitive resources (Patterson,

1 994).

People required limited information for this process of model building

(Hamilton 8 Sherman, 1994; lchheiser, 1970; Reeves 8 Nass, 1996). They have

been able to distinguish what is the same and what is different about objects and

24

at the same time, understand which categories provided meaningful information
and which did not (Hamilton and Sherman, 1994). For example, sex category"
has been perceived to provide meaningful information about another, while eye
color has generally not been considered useful information. In this way,
categorizing has allowed people to successfully navigate and cope with a
complex environment with a limited cognitive processing system in that it has
prevented them from having to process every object in our environment in detail
(Bruner, 1957; Hamilton 8 Sherman, 1994).

Once a particular categorization has been made, it was then associated
with stereotypes, or judgments associated with a particular categorization that
subsequently influenced the mental model of others (Bodenhausen 8 Macrae,
1998). These predictions have been seen as expectancies and interpretations
for the behavior of others (Bodenhausen 8 Macrae, 1998; Hinton, 1993). In this
way, previous interactions with others lead perceivers to make predictions of
what people they encounter in the future will "be like" (Fiske 8 Neuberg, 1990;
Wyer 8 Carlston, 1994).

SOCIAL JUDGMENT FROM MENTAL MODELS

Now that the process of categorization and forming mental models of
others in the physical world have been introduced, this section examines some of

the evaluations people have made of their interaction partners when forming

 

" Sex category is achieved through the application of the sex criteria, which is categorized
through socially required displays that identify one‘s membership in one of the two categories.

25

mental models of them. The section will specifically explore a few categories and
judgments people are believed to make automatically during interactions with
intelligent others in the natural world and then turns to consider the implications
of these processes for SVEs.

SOCIAL ATTRACTION

 

For at least the past two decades, theorists and researchers in interpersonal communication have
centered much of their attention on interpersonal attraction. Not only has interpersonal attraction
been found to be a facilitator of interpersonal communication across a wide range of cultures, but
also much interpersonal communication exists for the primary purpose of enhancing interpersonal
attraction (McCroskey & McCain, I974, p. 261).

Social attraction has been shown to have a number of levels and to be
very important and salient to perceivers as they form mental models of others. In
the past, some have called this notion interpersonal attraction. However, given
the focus of this project, the notion of attraction to social entities more clearly
addressed the questions. This section considers two of the three dimensions of
social attraction recommended by McCroskey and McCain (1974) and Tardy
(1988). These are social attraction or liking and task attraction. These are
separated because they have been shown to be multidimensional (McCroskey 8
McCain, 1974). Also, the third dimension of attractiveness, task attraction, was
not used in this section, but was instead included in the measurement of partner
satisfaction because it has been shown to be a composite of a distinctive theme

from other measures of attractiveness (Burgoon 8 Hale, 1987).

 

However, “sex” is a determination of socially agreed upon biological criteria. West and
Zimmerman (1991)

26

Social attraction, or liking, has been considered an important part of
understanding social judgments of others. This judgment is a determinant of
whether or not people avoid or seek out the company of others (Horwitz, 1958).
Within this notion is the general concept of interpersonal attraction, which also
contains evaluative sentiments that are central components of relationships with
others (Tardy, 1988).

PARTNER SATISFACTION

 

Partner satisfaction includes interpersonal evaluations of credibility
(McCroskey, 1971; McCroskey, Hamilton, 8 Weiner, 1974) and trust (McCroskey
8 McCain, 1974; Tardy, 1988). Receivers have been shown to be more open to
communication from sources they perceive to be credible and/or homophilous
than to others (McCroskey et al., 1974). Credibility refers to the judgments made
by the perceiver about the believability of the communicator (O’Keefe, 1990). It
was arguably the central variable in all communication and, without it, the
communication would not be as effective (McCroskey, 1971 ). The two major
components of credibility have been expertise (competence) and trustworthiness
(McCroskey, 1966).

It has generally been assumed that as a communicator’s credibility
increases their effectiveness would as well. McCroskey (1971) argued that the
research in the area of credibility has been the clearest of any in the field of
communication. If a source has credibility they would influence the audience

and, if not, their impact would be minimal. Partner satisfaction and credibility

27

have been used as indicators of one person's perception of another person or
persons (McCroskey et al., 1974)

CONFIDENCE OF JUDGMENT

 

As discussed above, people engage in interactions with others to reduce
their uncertainty about how others are likely to behave in the future and to help
them understand current behavior in interactions (Berger 8 Calabrese, 1975).
People utilize categories and social judgments that experience has shown them
to be meaningful for perceiving others in interactions (Hamilton 8 Sherman,
1994; Lakoff, 1987; Wyer 8 Carlston, 1994). Although people are constantly in
the process of placing others into categories, they people do not like to think they
have judged another without sufficient information or that they have stereotyped
a person. Most would prefer to believe that their impressions of others are based
on individuating information”, even when the individuating information is related
to their category membership.

In fact, the mere presence of categorical information alone has been
shown to produce stereotypical answers (Bargh 8 Pietromonaco, 1982; Yzerbyt,
Schadron, Leyens, 8 Rocher, 1994). Even when people receive meta-information
about a person that was not diagnostic or related to the trait variable being
considered, people have been shown to be more confident about their perception
(Gill, Swann, 8 Silvera, 1998; Swann 8 Gill, 1997). Even information technically

irrelevant to the interaction, or not necessary for the judgment, would increase

 

‘2 It is possible that this is unique to Americans or other individualistic cultures.

28

information about the other person that a perceiver has, which has been shown
to increase the person’s confidence in their perception (Gill et al., 1998).

THE UTOPIAN PROMISE AND REALITY OF INTERACTIONS IN SVES

The original predictions about interactions in cyberspace represented a
type of utopia. Researchers predicted a democratization of computer mediated
interactions; that status, gender, race, and other categories accessed during the
process of perceiving human “others" discussed above would not be obvious and
thus not relevant (Hert, 1997; Lea 8 Spears, 1992; Rice 8 Love, 1987; Siegel et
al., 1986). The hope was that computer mediated communication would remove
these issues from the mental models of others, meaning that each voice would
truly be heard and equal, achieving true Jeffersonian democracy. In other words,
researchers were hopeful that more people would participate in the decision
making process than in traditional face to face interactions and that all voices
would carry equal weight.

With few small exceptions (Adrianson 8 Hjelmquist, 1991), the first reports
from cyberspace were not consistent with these utopian predictions, and
evidence of the entrenchment of traditional categorization suggested by
embodiment has surfaced in even the newest of media (Clark, 1995; Yates,
1997). Uncertainty can be very high in a virtual environment, so it is likely that
humans have based attributions of others on categories that have been familiar
and useful in traditional interactions in their attempt to “ground or stabilize that

which is new” (Clark, 1995, p. 114). This means that people would want to

29

perceive virtual others using the same indicators and categories that they have
used in the natural world in order to bring the familiar into an unfamiliar
environment.

The human experience in the natural world has led to a reliance on
physical cues for information about categorization. In virtual environments, the
physical cues have been less distinct and certainly less permanent if they existed
at all. This suggests that a reliance on physical distinctions for categorization
may lead to confusion in assigning others to category membership. In turn, this
may influence people’s ability to interact and navigate the environment, as well
as their ability to process or use other information. The impact of this on people’s
ability to interact with others and with the environment is as yet undefined and
may result in distortions in people’s perceptions of virtual others and their ability
to interact in the highly entropic worlds technology makes available.

Research in social cognition implied that categorization of others in
unfamiliar environments would be based on whatever features were salient within
the stimulus context (Hamilton 8 Sherman, 1994). Salience in these terms
means both a property of the stimulus (i.e. cues that are obvious and easy to
discern) and a property of our mental processes; what was accessible13 and
frequently or recently accessed (Bargh 8 Pietromonaco, 1982; Bodenhausen 8
Macrae, 1998; Gilbert 8 Hixon, 1991). The next section considers what features

are likely to be salient in social virtual environments.

 

‘3 Accessible categories are more likely to be used to process relevant information. In this way,
accessibility partly determines how social information is interpreted.

30

CATEGORIZATION PROCESS IN SVES

 

This section discusses the way the categorization process of the natural
world introduced above is likely to port into the worlds of social virtual
environments. Early research in cyberspace has indicated that many of the
processes involved in social cognition are robust and unchanging, even when
mediated by technology. Thus, it is likely that many of these processes would
remain unchanged, regardless of context, and that people would continue to
categorize others in an attempt to efficiently navigate the world with limited
cognitive resources (Bruner, 1957; Hamilton 8 Sherman, 1994). The human
ability to utilize whatever information is at their disposal (Hamilton 8 Sherman,
1994), and the human tendency to be cognitive misers14 may mean that humans
will continue the process of categorizing others in any environment, whether
virtual or not.

When traditional cues or indicators are not available to place another in a
category, people would be likely to use whatever cues are present in the
environment to make category assignment of others. Regardless of where they
get information about people and how to categorize them, people will likely
continue to rely on categories of individuals that have proven important in face to
face interactions. Further, when natural cues are available, people will use them
to categorize others in the traditional manner. Natural embodiment cues are the

most salient and perceived to be the most useful for the categorization process.

31

When traditional embodiment cues are not available, people will need to turn to
other indicators for information required to place others in categories. These may
include the virtual image or behavior such as language use or ability to navigate
the virtual environment and other potential category membership indicators.

Although many of the processes involved in categorization are likely to
remain stable, some changes are likely to occur given that mediating the
connections between people changes the perception of the source of that
information, as well as how the information is processed, remembered and used
(as predicted by McLuhan, 1964; Salomon, 1990). For example, it may mean that
people will evolve new categories that would be meaningful in virtual
environments as well as new indicators of membership in old categories.

These changes in the process of categorizing others during mediated
interactions may influence human-human interactions in the natural world.
Salomon (1990) explained that technology can alter our cognitive functions and
this altered function will influence interactions even in the absence of technology.
He described this notion as ”cognitive residue" or changing the cognitive function
so that it is observed outside the context in which it was internalized. Given this
notion, it is likely that interactions through and with media will alter interactions
outside the medium and will influence the way people perceive and interact with

others.

 

“ Gilbert and Hixon (1991) argued that the tendency of humans to reserve cognitive capacity by
making the best decision with the least cognitive resources is automatic.

32

Consider how people’s experience in the natural world influenced this
process. Human categories have been encoded almost instinctively upon an
individual entering a person’s range of senses and for all of our experience in the
natural world this has been meaningful. Further, these cues were automatically
encoded and usually easy to ascertain. This knowledge proved helpful in
categorizing and predicting others in the natural world, so people would be likely
to continue to encode it. As discussed above, people have utilized categories to
address the human need for “coherence, simplicity, and predictability in the face
of an inherently complex social environment” (Bodenhausen 8 Wyer, 1985, p.
267). People automatically categorize others based on any information provided.
This has included context cues as well as nonverbal information. All information
around a person would be incorporated into the mental model of them to varying
degrees whether that information is diagnostic or not (Gill et al., 1998; Swann 8
Gill, 1997).

This section predicted that people will continue to try to make sense of the
world and categorize in SVEs, and that they will continue to use many of the
familiar and salient indicators as they interact in the unfamiliar and new virtual
environments. It is possible that in the future, they may begin to rely on new
features not currently considered. If this ports (or translates) directly into SVEs,
then people with more information, even meta-information, will feel more
confident about their attributions of others in the virtual world. In this case, it

means that people who can perceive a virtual image representing the other are

33

both more likely to feel more confident about their attributions of others and to let
the virtual image influence those attributions. The next section presents a brief
description of the types of the categories that are likely to be meaningful in both
traditional interactions and in cyberspace and those that will be unique to each.

THE VIRTUAL IMAGE INFLUENCES PERCEPTION

 

Researchers observing interactions in cyberspace have provided support
for the prediction that the categories reserved for physically embodied humans
maintain salience and meaning in cyberspace. Major support comes from
participant observation research, which has revealed the first questions in social
interactions in cyberspace were related to gender and status (Spender, 1996;
Turkle, 1995; Waskul 8 Douglass, 1997). This section addresses the question of
what indicators people are likely to use to assign others to category membership.

The first reports from cyberspace have indicated that people continue to
rely on any available information pertaining to physical characteristics and that an
image has continued to be part of the mental model, even when no visual image
is provided to the perceiver. This has included whatever was available, from
interactions providing no image at all to the text-based description to the physical
characteristics of the virtual image. Further, perceivers have continued to apply
the same social judgments of virtual bodies that they would to natural bodies
(Reeves 8 Nass, 1996; Suler, 1997; Takeuchi 8 Naito, 1995). As Lipton (1996)
argued, "When participants represent their virtual body, the means of oneself

changes from an unconscious process to a conscious action. In reality, my body

is represented by its physical characteristics, and all people will unconsciously
read my body as a human body” (p. 340-341 ).

The expressive body has been considered to be the basis of human action
and identity in the world (Waskul 8 Douglass, 1997). It has been said that people
exist only because of their bodies (Hampshier, 1991). A human‘s physical
appearance is the basis of how others would identify them during interactions
(Eisenberg, 1990; lchheiser, 1970; Mehrabian, 1972). In other words, to some
extent the natural, unmediated embodiment (the human body) has appeared to
do more than represent a person, it has been perceived to "be: the person. As
human bodies increasingly enter cyberspace, it is likely that individuality will
increasingly be recognized by ”the multimedia, full-body electronic surrogates
that we may soon control" (Oravec, 1996, p. 48). If the body is the basis of
identity in the natural world, then this process is likely to port directly into SVEs,
meaning that the virtual body will become the basis of identity in the virtual world.
Further the use of embodiment has allowed for a form of personalized expression
that has made individuals distinguishable and more attractive as communication
partners (Henne, Mark, 8 Voss, 1998).

This reliance on the body for indications of category membership may
bring about distortions in people’s mental models of intelligent others in SVEs. In
the virtual world, it is possible for a person to swiftly inhabit one virtual body and
just as swiftly switch to inhabit a completely different virtual body (Oravec, 1996).

Advanced embodied virtual environments present the user with a continuum of

35

bodies. Users may alter the shape at any time during the interaction and others
may choose the same shape as those already in the environment (Oravec,
1996). In this way, body morphology becomes fluid in virtual environments and
not at all as consistent as would be expected from people’s experience with the
natural world. This may influence the ability to form meaningful relationships
given the inability to easily recognize and locate others with whom people have
previously interacted.

This section explained how the human practice of categorizing another
person based on what have been considered permanent features would cause
problems with perceiving others in SVEs. People have continued their reliance
on physical cues, but in virtual environments people have used the visual cues
provided by the virtual image as their primary indicator of whether one is human
or not and to assign others to membership in other categories accessed when
perceiving intelligent others. Given that the same image may be one entity today
and another tomorrow, people may need to re-define how they understand and
recognize others in these environments. The next section discusses the potential
influence of the features of the environment on perception of others and on the
nature of the interaction.

THE VIRTUAL ENVIRONMENT AND CONTEXT FOR INTERACTION

The setting and context of the environment influence the outcome and
nature of any interaction. This has also applied to interactions in social virtual

environments. As discussed above, people have continued to categorize others

36

in social virtual environments. This section considers the potential influence of
the features of the medium and environment on people’s ability to form mental
models of others. This section explores the features available in a variety of
environments, from text only to fluidly moving 3-D environments and predicts
what information people are likely to use in the process of categorizing and
creating a mental model of others and how they will use it. Refer to Law for a
summary of this exploration.

People have formed mental models of others based on whatever
information is salient, perceivable and accessible within the context (Bargh 8
Pietromonaco, 1982; Bodenhausen 8 Macrae, 1998; Bruner, 1957; Hamilton 8
Sherman, 1994). The indicators used to assign others to their membership in
categories have varied based on a number of things including context and
available stimulus materials as well as individual preferences and experience.

When the environment provided a fluidly moving 3-D virtual image, many
of the features people have used in the natural world for this categorization
process are available, though presented in a different manner which may
influence processing and memory for these features. It is likely that they would
be used in similar ways to categorize people in the relevant and perceivable
“human” categories. People could use proxemics, body type (which will include
quality of the image and other skill indicators), and the sex of the image, gait
information including latency, and/or the ability to navigate and interact in the

environment. Race and traditional status indicators may be more difficult to

37

obtain; although race could be assumed from use of language or from any
information provided by the virtual image. Status is likely to be associated with
everything from size of the virtual image to gait information as well as quality of
the virtual image and ability to navigate, and familiarity with, the environment and
culture.

When the environment provides only a 2-D virtual image, people could still
use proxemics, body type, and gender of the image and knowledge of the
environment as described above. When there is no visible image in the
environment, it may become more challenging to categorize the other on these
primary features, but people have continued make these categorizations (see
Schiano, 1999). People have created a mental model of the other and this model
has included an image of the other based on the categories to which they were
assigned. In other words, just because no visible image was provided in the
interaction did not mean people’s mental model did not include an image.

In some text-based environments the user may provide a physical
description of themselves. This description could include sex, perhaps race,
interests and even profession. Although this is not necessarily an accurate
description, people have used these indicators in the process of assessing the
person and assigning them to categories.

When there is no visible image and no description of the other, uncertainty
would be the highest of any other context. This means that creating a mental

model of another would require the most effort. Given no physical indicators, the

38

individual must treat every word typed and typing speed and style as indicators of
“human” categories.

Categorization would occur whether an image is visible or not and much of
this categorization has likely been based on cues that were salient and
meaningful in nonmediated interactions (Clark, 1995; Oravec, 1996). However, it
is likely that over time, what is meaningful and salient will continue to evolve. The
question of which cues will be meaningful in virtual environments is as yet
unanswered, perhaps even undefined. In Cyberspace, not only have people
continued to utilize the same cues and categories they have used in traditional
interactions, but also new categories are evolving and the meaning of these cues
are beginning to take on different meanings.

This indicates that the quality of the virtual image and familiarity with the
environment and culture could be the defining characteristics, while language
and the content of the interaction also play a large role in the perception process
as well. “So, although natural physical appearance, dress, and other status cues
recede, educational competencies and linguistic skills increase in importance.
Computer-communication media are not neutral with regard to culture, education,
and socio—economic class. And electronic persons are not more 'equal’ than
proximate individuals, we just use different criteria to rate them" (Schmitz 8 Fulk,
1991,p.85)

But when there is no image to be identified, or when a virtual image is

easily manipulated and transformed at will, the codes that structure the meanings

39

people make of bodies may take on new functions. The characteristics of
humans that have defined peoples’ positions in society have evolved in different
contexts with different environments and social values and this is likely to
continue in SVEs.

As more people enter cyberspace, the differences of access to and
experience with technology promises to increase in significance. As hinted at in
Stephen’s science fiction novel, Snow Crash, beautifully designed virtual bodies
with fluid movement and no discernible lags will contrast with the store bought
virtual bodies represented by people dialing in from public terminals, and the
“wrong side of the tracks“ will take on a new meaning (Clark, 1995; Stephenson,
1993).

RESPONDING TO ANTHROPOMORPHIC OTHERS

People have responded to anything that looks human as human.
Evolutionary psychology argued that it is simply automatic and efficient to
interpret stimuli that appear human as animated or driven by a human
intelligence (Reeves 8 Nass, 1996). This would be consistent with people’s
experience in the natural world. As discussed in Chapter 1, when people meet
others, their primary goal would be to be able to better predict and or control their
behavior and to reduce uncertainty (Berger 8 Calabrese, 1975). As people begin
the process of uncertainty reduction, they assess the goals of the interaction and

decide what they need to know to fulfill these goals. People then use whatever

40

information is available or salient in the environment to fill in the blanks (Bruner,
1957)

People have spent their lives interacting with others. During these
interactions, people form mental models of people and predict their motives
during interactions. People have less experience interacting with and considering
motives of agents and other non-human intelligent beings. Thus, it is likely that
people will feel more confident in making predictions about humans than agents
and that a more anthropomorphic other will be more likely to be classified as
human, which will influence the mental model of the other.

In interactions with no visible natural bodies a number of issues arise that
complicate the perceptual processes. First, the lack of embodiment would detract
from the ability to recognize one another and the ability to take turns during an
interaction (Benford et al., 1995). This would also make it difficult to identify the
speaker, which is important for credibility and trustworthiness issues (Short et al.,
1976). Interface designers have responded to this concern and there are
currently a number of different systems that allow varying degrees and types of
interactant recognition (Benford et al., 1995). The remainder of this section
explores the influence of the virtual image, especially when anthropomorphic, on
the process of forming mental models of others.

As indicated above, people have made rapid initial categorizations of
others based on some attribute (usually physical characteristics) that allowed

them to form impressions and mental models of the people they encountered

41

(Bruner, 1957; Fiske 8 Neuberg, 1990; Fiske 8 Pavelchak, 1986; Hamilton 8
Sherman, 1994; Lakoff, 1987). But how does the presence of the virtual image
influence this process? How does it influence the categorization? Which features
most influence a person’s judgment (i.e., the color, shape, size or
attractiveness)?

Before answering these questions, this section considers the processes
likely to influence mental models. People automatically make judgments of others
based on their bodily appearance (Argyle, 1975; Chesebro 8 Bonsall, 1989;
Fiske 8 Neuberg, 1990; Goffman, 1963; Hinton, 1993; lchheiser, 1970). The
automaticity of these processes (Patterson, 1995) makes it unlikely to be
questioned or altered when individuals are in social virtual environments.
Although in SVEs the virtual image would not necessarily have a relationship to
the natural image, it is predicted that perception of the other will be largely based
on any available cues, such as the visible characteristics of the virtual image.
The judgments of the person are thus more likely to be influenced by
stereotypical cues associated with the virtual image than to information
presented in other forms during the interaction.

The categories reserved for humans are likely to be most salient in
perceiving anthropomorphic others, or those that look and act like humans.
Further, virtual others will be categorized in a manner consistent with
stereotypical notions related to the image presented. In other words, people will

use the features of the moving body, whether virtual or natural, to categorize their

42

interaction partner. People treat computers as social actors (Reeves 8 Nass,
1996), it is likely that the categorization process will be more complete and the
social response will be more exaggerated when the interface appears
anthropomorphic.

In considering the influence of anthropomorphism, it is important also to
consider the implications of this for designing and using interfaces. One
important and controversial area of discussion with regard to virtual humans is
the push to create a type of embodiment for humans that is clearly discernable
not only from agents or other objects, but also, that allows for easy differentiation
between and among individuals (Benford et al., 1995). On the other hand, a
group of human computer interaction researchers are attempting to blur the line
between human-human interaction and human-computer interaction by making
the interface so intuitive and intelligent that the difference is not easy to discern
(for a discussion see Brent 8 Thompson, 1999). Further, designers are
intentionally creating agents with personalities to further encourage people to
respond socially to computers (Dryer, 1999). Before the question of which
direction is most advisable can be addressed, one must first address how these
differences influence perception of intelligent others and it is to these issues this
section considers.

As discussed above, perceiving others in SVEs has required the
consideration of new types of images as well as different issues and a new range

of intelligent others. These others have not necessarily been physically

43

discernable from human others, as both have virtual images, which has
complicated the perception process. People should not assume that
anthropomorphic bodies represent humans because this may not be the case,
but this does not mean people will no longer make these assumptions. Also,
objects that are inanimate in the natural world can be animated and
anthropomorphized in a variety of ways in virtual environments and people may
choose to be embodied by them.

Researchers at Stanford University undertook a program of study in this
area that asked how people responded to computers in a variety of settings.
They concluded that people responded socially to computers, or in the same way
as they responded to other humans (Reeves 8 Nass, 1996). For example, they
found that people were polite to computers and seemed to worry about hurting
the computers feelings. They gave more positive evaluations of a computer-
based tutorial when they evaluated the tutorial on the same computer as
compared to the evaluation of the same tutorial when evaluated on a different
computer. Also, participants were more likely to feel affiliation with a computer
that had been assigned as a teammate. Participants also were shown to apply
gender stereotypes to synthesized voices on computer tutorials even though
participants reported that it would be unreasonable to do so.

Others have looked at the influence of computerized anthropomorphism in
two ways; language use (Wexelblat, 1997), and embodiment (Dryer, 1999; Koda,

1996). Wexelblat (1997) and Koda (1996) found no change in perceived

44

intelligence, as long as the virtual image was more complex than a simple line
drawing. This work indicated that, “even when dealing with an extremely realistic
human interface representation, people are not fooled into thinking the system is
more intelligent or more capable” (Wexelblat, 1997). However, Koda (1996)
found other differences between impressions based on the appearance of
characters and impressions in a playing poker manipulation. This included the
finding that a realistic face was more likely to be rated as more likable and
engaging than either a caricature face or smiley face. Further, she found that the
existence of a human face (as opposed to no face or a dog’s face) increased
ratings of likeability, engagement and comfort. People like to talk to attractive or
interesting looking people (Berscheid 8 Walster, 1969). This has continued to be
true in cyberspace, where attractive avatars are more popular and liked more
(Suler, 1996).

Similarly, Wexelblat (1997) reported that participants found the interaction
more enjoyable when the agents were more anthropomorphic. Essentially, this
means that people respond to computers and other objects as social actors (or
as human) whether these objects are anthropomorphized or not (Reeves 8 Nass,
1996). However, Dryer (1999) found that the artistic style of the representation as
well as the kind of thing (human or not) that was represented determined how the
agents were classified in terms of personality attribution.

Virtual bodies, especially more anthropomorphic bodies, have been shown

to be more engaging, interesting and attractive (Koda, 1996; Wexelblat, 1997).

45

More anthropomorphic virtual images have been shown to be more influential
than less anthropomorphic virtual images. Thus, a more anthropomorphic virtual
image is likely to have more extreme social reactions, to be treated as ”more
human." A less anthropomorphic image will be treated as more human than an
entity with no image, but not as human as a more anthropomorphic image. For
example, participants seemed to feel more confident of their prediction of the
other’s behavior when the other looked more like a person because people feel
more confident making attributions of human motives (as found by Parise,
Kiesler, Sproull, 8 Waters, 1999).

Although an interface that was more engaging and interesting has
provided more satisfactory interactions, the engagement with the image takes
processing resources that were then unavailable for processing other features of
the interaction. First, consider the influence of the natural body on perception in
an interaction. In any interpersonal interaction, people have been very sensitive
to communication cues provided by body morphology and movement (Argyle,
1988b; Burgoon et al., 1996; Ekman 8 Friesen, 1969; Mehrabian, 1972). These
have included attractiveness, facial expression, posture, gestures and body
movement among others (Argyle, 1988a; Berscheid 8 Walster, 1969; Bull, 1983;
Bull 8 Rumsey, 1988; Dion, Berscheid, 8 Walster, 1972; lchheiser, 1970). These
cues have been so powerful and influential that when these cues provided by the
body or nonverbal "signals" have contradicted information communicated

verbally, people have had more faith in the veracity of the cues provided by the

46

body (Burgoon et al., 1996). Psychologically, the body, virtual or otherwise, has
been seen as the more authentic expression of the other’s consciousness. It is
important to note that the sensitivity to bodily cues have been automatically
applied by individuals to draw inferences about the “intentions” and “mental
states” of computers when facial or body displays (even simplified simulations of
these cues) were provided (Reeves 8 Nass, 1996; Sproull, Subramani, Kiesler,
Walker, 8 Waters, 1996; Takeuchi 8 Naito, 1995).

This means that when a virtual image moved and acted in a computer-
generated environment it had a powerful psychological effect on the interaction
and on people’s perceptions of one another. The embodied humanoid form
triggered mental models of human intelligence in observers and influenced the
observers' interaction even when the observer is fully conscious that the “other
intelligence” is not human (Reeves 8 Nass, 1996; Sproull et al., 1996; Takeuchi
8 Naito, 1995). For example, people have tried to interpret facial expressions of
even simulated or iconic human faces at the expense of their task, or other
information in the environment (Koda, 1996; Nowak 8 Anderson, 1999; Takeuchi
8 Naito, 1995). Sproull, Subramani, Kiesler, Walker and Waters (1996) found
that providing a synthetic talking face on a computer monitor caused people to
perceive the interaction with the computer as more social and caused them to
modify their behavior as if they were interacting with another human. Nowak and

Anderson (1999) found that people tended to be more focused on the existence

47

of smiley faces in the comer of a computer than on important information
presented via text.

Although people were more involved with the interface when it was an
expressive face, this involvement was at the expense of attending to and
processing other information in the environment (Koda, 1996; Takeuchi 8 Naito,
1995). Users have tended to interpret all computer-based intelligence in
anthropomorphic terms. Giving computer interfaces humanoid embodiments has
enhanced this tendency of users to interact with agents and non-human entities
as if they were controlled by human intelligence or by humans.

So the decision of whether or not to use virtual images should depend on
the purpose of the interaction. If the goal is to transmit complex information, it
may be best to do so without images or facial expressions that may distract the
people involved. This is especially in persuasive communication, given that
systematic persuasion is more robust than heuristic based persuasion (Chaiken,
1980; Pallak, 1983).

The remainder of section examines how the various cues generated by
the virtual image influence the mental model of intelligent others. Advanced
embodied virtual environments can present the user with a continuum of bodies.
They can allow various shapes to be controlled by human and/or artificial
intelligences. If it is true that users have a tendency to anthropomorphize all
intelligent others (Reeves 8 Nass, 1996), especially if they take humanoid form,

then it is quite possible that we will see the rise of various types of new and

48

unique mediated relationships. All of these may not be interpersonal, because
the level of copresence may vary tremendously from system to system and the
relationship might be with a non-human mind, or some form of artificial
intelligence rather than a human intelligence. But, from the users viewpoint, they
may all possess some “interpersonal” quality of copresence with another, a
sense of access to another mind, whether that ”mind” is humanoid or computer
generated and human-like. It is also entirely feasible for the nature of the
intelligence, or the entity behind the embodiment, to change during the course of
an interaction (Oravec, 1996).

It follows that in human-computer interaction, interaction cues from
humanoid virtual bodies would be a powerful means to enhance the meaning and
apparent co-presence of a human-artificial intelligence interaction. This would be
especially true in any mediated interaction involving highly interactive virtual
environments and expressive virtual bodies. Research has found that “a number
of social conventions are carried over from our interaction in the physical world,
were conveyed through the embodiment and positioning, and play a strong role
in virtual interaction” (Henne, Mark and Voss, 1998, p. 1).

Contrary to experience in the natural world, there would not necessarily be
any physical distinctions between and among humans and computers in SVEs.
This further complicates the processes involved in perceiving anthropomorphic
intelligent others. This in itself is likely to result in distortions in people’s ability to

form mental models of intelligent others in these environments. Recent advances

49

in the design and anthropomorphic motion and embodiment of virtual humans
(Badler et al., 1993; Capin et al., 1998; Magnenat-Thalmann et al., 1998) have
given salience to the potential influence of anthropomorphism in social virtual
environments. Objects vary in their level of anthropomorphism”. A thing could be
more anthropomorphic in a variety of ways. The three most likely to influence the
mental model of the other are whether or not the other is perceived to be human,
the degree to which one appears human and the degree to which one can mimic
human behaviors.

The first level of anthropomorphism is really an ontological question. The
“other” is either human or it is not human. Although whether the other is
perceived to be human is actually more important when considering how this will
influence the development of a mental model of the other. Tagiuri (1958)
discussed the importance of the perceiver feeling the object had the potential of
representation and intentionality or they would not be perceived as human. It is
predicted that there are varying degrees of certainty of the perception of
humanity and that this perception would influence the mental model of the other.
The importance of this distinction will vary with the goals of the interaction.

The other two levels of anthropomorphism, appearance and behavior
were considered to have varying degrees. One either looked more or less like a
human (3 more or less anthropomorphic image) or acted more or less like a

human (displaying behavior like those humans display, such as intelligence).

 

'5 In this meaning, something that is more human is more anthropomorphic

50

The basic processes of categorization were not predicted to change in
SVEs. If this were true, then people would have different mental models for
humans than for objects. Further, people would have required different types of
information about bots than about humans. For example, people may not
consider intentionality of a bot in the same way they would a human. The
differences may be subtle, but the accessible categories will influence the
processing of information.

Whether in the natural world or the virtual world the initial categorization of
another was predicted to be the living non-living distinction. If living, then people
would ascertain whether or not the object is human. Once people established
category memberships in the above distinctions, they would continue the process
of creating a mental model of that person or object. This process of creating a
mental model of another was predicted to include their perception of the other
and their membership in each of the primitive categories such as gender, race,
age and social class (Fiske 8 Neuberg, 1990; Hamilton 8 Sherman, 1994).

The next section considers the theoretical predictions to which the above
literature leads and proposes a way to test them. It explores the potential
implications of the fact that there are no physical distinctions between and among
humans and computer intelligences in SVEs as well as the influence of

anthropomorphism on perceptions of intelligent others in SVEs.

51

 

cybe
lheo
unce
unce

and;

peep
infifie
Derce
The n
conce
WEGK
Date;
dECUS
anifioj

“NCEpj

Chapter 3

Model of AnthrOpomorphism of Mediated Intelligent Others on

Mental Models of Social lnteractants

As discussed previously, the anthropomorphism of virtual entities in
cyberspace has occurred to varying degrees and on several levels. The
theoretical model proposed in this section was based on the premises of
uncertainty reduction theory including that people interact with others to reduce
uncertainty about both the future behavior of others as well as to explain past
and present behavior (Berger 8 Calabrese, 1975).

The theory was the basis for the theoretical model that assumed that
people interacting in SVES would continue to strive to reduce uncertainty during
interactions. The model began with a consideration of the degree to which the
perceived intelligent other was anthropomorphized and perceived to be human.
The model presented this as a very important consideration in a person’s
concept of an intelligent other in social virtual environments. The model
predicted that the two most prevalent levels of anthropomorphism influencing
perception would be physical and behavioral (whether verbal or nonverbal). As
discussed below, the model predicted the influence of these two levels of
anthropomorphism on copresence and presence in the environment, as well as

perceptions of social attraction and partner satisfaction.

52

 

cf
hc

pm

St

or f
Pier
this

are

Illlloj

”lode

Please refer to the graphic outline of the theoretical model in Figure 1.

MENTAL MODELS OF OTHERS

As people navigate ever-changing environments, they process each
object they encounter to varying degrees. People have been able to perceive the
characteristics of a new or unfamiliar object and place it into categories, noting
how the new object is similar to and distinct from other objects that have been
previously encountered and processed (Fiske 8 Pavelchak, 1986; Hamilton 8
Sherman, 1994; Lakoff, 1987; Wyer 8 Carlston, 1994). This process of
categorizing the features of new objects in terms of its similarity to and distinction
from familiar objects has allowed humans to navigate a complex environment
with limited cognitive resources (Bruner, 1957; Fiske 8 Neuberg, 1990; Hamilton
8 Sherman, 1994). The categories used in this process vary based on a number
of factors, including what was salient in the environment (Hamilton 8 Sherman,
1994), what was mentally accessible and what has been shown to be meaningful
or frequently utilized in previous encounters with similar objects (Bargh 8
Pietromonaco, 1982; Bodenhausen 8 Macrae, 1998; Gilbert 8 Hixon, 1991). In
this way, the processes involved in perceiving living things and inanimate objects
are the same or very similar (Bruner, 1957; Fiske 8 Neuberg, 1990; Srull et al.,
1985).

Now that the processes of forming mental models of others have been
introduced, the next section discusses the theoretical model and explains the

model's predicted influence of whether or not one has been perceived to be

53

human, whether or not the other is represented by a virtual body and the
influence of an anthropomorphic virtual body on these processes and the
resulting social judgments of others.

EFFECT OF HUMANITY ON SOCIAL JUDGMENT OF INTERACTANT

This section explores the potential implications of the perception of
whether the other is human or not for social judgment. People create mental
models of both people and objects. The processes have been shown to be the
same, but different categories would become relevant and accessible when the
perceived object was believed to be human as compared to any other entity or
object encountered in the environment (Heider, 1958; Konner, 1991; Ritvo, 1991;
Sheehan, 1991a; Sheehan, 1991b; Tagiuri, 1958). People have almost
instinctively made this distinction from an early age (Keil, 1994). In this way, the
categories that would be accessed and utilized would be different when
perceiving a living or human as compared to not human or inanimate objects.

Therefore, the model predicted that the perception of whether or not the
other is human would influence the types of judgments included in the mental
model of the other. It further predicted that the information contained in the
mental model of the perceived other would be different when the other was
introduced as an agent compared to an avatar because the category models for
humans are different from that of objects (Heider, 1958; Keil, 1994). This would

only be true if an agent is classified as more similar to an object than a human.

54

In the natural world, all living things contain certain common properties
that distinguish them from inanimate objects (Asch, 1958; Hastorf et al., 1970;
Heider, 1958). So, the model predicted that on the first level of
anthropomorphism, there was the ontological question of whether or not the
other is human. Then, especially in the virtual world, the model predicted there
would be varying levels of “humanness” or qualities that brought up the
categories and perceptions used for humans.

Most living things have had more physical characteristics in common with
other living things than with inanimate objects (Keil, 1994). For example, in the
natural world, only living things have been able to autonomously move and
interact in the environment and only living things have been able to breathe and
eat. In the natural world, things that are not living have not moved or spoken.
However, in virtual environments 3 computer (an object) has been able to display
features and characteristics that are traditionally reserved for living things or
specifically humans, such as intelligence. The model predicted that this would
result in distortions in people’s perceptions of others.

It predicted that these distortions would eventually result in the evolution of
new categories that merge parts of the categorization accessed for humans and
part of that accessed for objects. For example, motivations, and the decision to
form an ongoing relationship with another are usually only relevant during
interactions with another human. Intelligent others interacting in SVEs may well

be a new category for intelligent others in the social world of virtual

55

environments. The model predicted that an anthropomorphic character and/or
anthropomorphic behavioral characteristics would make more of the categories
traditionally reserved for humans salient and accessible, whether the other was
simply an anthropomorphic agent or truly a human (avatar). The mental
distinction between them was predicted to be important because the distinction
defines or frames the categories that previous experiences have shown to be
important or useful (Hamilton 8 Sherman, 1994; Lakoff, 1987).

The theoretical model predicted how the mental model and corresponding
social judgment of an interaction partner that was perceived to be an avatar
would differ from the mental model of another perceived to be an agent.
Specifically, it predicted that the process would be similar and the perception of
the same virtual image would consistently result in similar categorization whether
the intelligent other is an agent or an avatar. This would mean that although the
mental model of an agent was different from that of an avatar, the salient
categories for any intelligent other would be similar, as is the case for salient
categories of any animate object.

The model predicted that in some ways perceiving an agent would not be
so different from perceiving an avatar because it would appear animate and
could have many of the characteristics normally reserved for living things and
even characteristics traditionally reserved for human others. If this were true,
then many of the same categories would become salient. For example,

anthropomorphic non-living things could raise categories traditionally reserved for

56

humans such as attractiveness, gender, social status and race. The model
predicted that people would interact with anything displaying intelligence or
anthropomorphic cues in similar ways (Reeves 8 Nass, 1996). Although some
have argued that this will result in the same mental models for agents as for
avatars the theoretical model predicted that there would still be discernable
differences because of fundamental information from basic categories
(human/non-human, living/non-living).

Hypothesis I: People will feel more social attraction for avatars than for agents.

Hypothesis II: People will feel more partner satisfaction with an avatar for an
interaction partner than with an agent.

Given the human tendency to prefer things that are familiar and similar to
them, the model predicted that people would prefer interactions with other
entities that were human and, if not human, they would prefer those that either
looked and/or acted human. If the model were correct, then people would feel
more attracted to and satisfaction with their choice of partners when they
interacted with an avatar as compared to an agent because they would feel more
confident in their ability to perceive intentions and motives (Dryer, 1999).

Credibility and intelligence have been generally considered part of the
notion of partner satisfaction. The model predicted that there would be some
individual differences in perceptions of the credibility of an agent as compared to
an avatar. If the model was correct, there would be differences between social
responses to a human other when compared with social responses to non-

human "others”.

57

Hypothesis Ill: People will feel more copresent with an avatar than with an agent.

People have experienced copresence with other humans in both the
physical and the psychological sense. Two humans with physical bodies could be
copresent and physically present in the same location with other humans in the
natural world. However, the idea that one can be copresent with a virtual other,
that one can be in the same physical location as an entity with no true existence
outside the computer may be difficult to perceive. Therefore, the model predicted
that feeling copresent with another human would seem more real than the feeling
of copresence or involvement and intimacy with a computerized agent. If the
model was correct, this would influence people’s perception of their level of
connectedness with an intelligent other as well as the perception that their
interaction partner was copresent in the interaction.

Hypothesis IV: People will feel less uncertainty in social judgments of avatars
than of agents.

When people encounter an intelligent other, they have been shown to
automatically respond to that entity in social ways (Reeves 8 Nass, 1996). This
has represented part of the script humans have for intelligent others. However,
the model predicted that people would not feel as certain in using this script for
agents. If the model was correct, people would have felt more uncertainty in their
interactions with agents because they did not have a script to perceive or
evaluate computers or bots (Dryer, 1999). Therefore, the model predicted that
they would substitute the only script they have for intelligent others-that which is
used to evaluate humans but that they would feel less certain in using this script.

Hypothesis V: Judgments of avatars will be more extreme than judgments of

58

 

be

DE

Illl

EX

Da

agents
The model predicted that the intensity of the social responses and

complexity of the model formed about the other would be more extreme when the
perception was of another human as compared to the perception of a computer.
If the model was correct, people would attend more to cues and other indicators
during interactions with other humans as compared to interactions with agents.
In this way, the model used extremity as an indication of involvement in the
interaction in the process of making social judgments. The model predicted that
the closer the other was perceived to be human the more people would attend to
the creation of a mental model of the other, and the stronger the sense of the
other will be; when positive it would be more positive and when negative it would
be more negative. Categories that were not predicted to be relevant when
perceiving a bot that were predicted to be relevant when perceiving a human
include social attraction, partner satisfaction and copresence. If the model is
correct, people’s ratings of agents on these measures would have been less
extreme or less applicable as compared to the ratings of avatars on copresence,
partner satisfaction and social attraction.

THE RELATIONSHIP OF PRESENCE AND COPRESENCE

Hypothesis VI: People that feel more physical presence in the environment will
feel more copresence with their interaction partner.

The model predicted that people who felt more copresent with their
interaction partner, would report feeling more physical presence and more

engaged in the environment. If the model was correct, then the feeling of

59

copresence with another would make the environment more “real” and it would
make the surroundings more interesting, engaging and compelling. The feeling
that one has connected with another in the environment would therefore
influence the perception of the environment itself, or the perception of the
environment may be influencing the feeling of connection with the other. If the
model is correct, people that felt more copresent with their interaction partner
would feel more physical presence in the environment.

EFFECT OF EMBODIMENT ON SOCIAL JUDGMENT OF INTERACTANT

This section focuses on the question how the virtual image is likely to
influence the social judgment of agents and avatars in SVEs. The theoretical
model predicted that when people enter an environment, whether virtual or
natural, they would make mental models of the things and people they
encountered. As discussed above, if the model was correct, people would
continue to place objects and people into categories that related to people and
objects they have encountered in previous environments. People would work to
reduce uncertainty in the environment and to function in a complex world with
limited cognitive resources (Bruner, 1957; Hamilton 8 Sherman, 1994). People
have used visual information to assign things and people to categories (Argyle,
1975; Chesebro 8 Bonsall, 1989; Fiske 8 Neuberg, 1990; Goffman, 1963;
Hinton, 1993; lchheiser, 1970). In the natural world, people have relied on the
relatively stable and permanent physical features of the face and body to

categorize other humans (Argyle, 1975; Argyle, 19883; Hinton, 1993; lchheiser,

60

1970). The natural, physical body has provided information that people have
used to categorize others in meaningful ways.

People have continued to categorize other people in social virtual
environments and the same categories have remained salient (Lipton, 1996;
Reeves 8 Nass, 1996; Spender, 1996; Suler, 1996; Takeuchi 8 Naito, 1995;
Waskul 8 Douglass, 1997). The theoretical model predicted that people's
experience in the natural world would lead them to rely in similar ways upon the
features of the virtual image to provide information for categorization in virtual
environments. If the model is correct about this, then just the presence of an
embodied other, whether truly human or not, would trigger mental models of
human intelligence in observers and would subsequently influence the observers
interaction even when the observer was fully conscious that the “other
intelligence” was not human (Reeves 8 Nass, 1996; Sproull et al., 1996;
Takeuchi 8 Naito, 1995).

Hypothesis VII: When a virtual image is visible, people will feel copresence with
their partner than when no virtual image is visible.

The image that represents a person communicates not only information
about the person, but also information about another person's physical presence
(Benford et al., 1995; Biocca, 1997; Biocca 8 Nowak, 1999a). The theoretical
model predicted that when there was a visible image, virtual or otherwise, people
would feel they have shared an experience or connected with their interaction
partner. If the model was correct, participants would feel more copresence when

the other was represented by a virtual image.

61

Hypothesis VIII: When a virtual image is visible, people will feel more social
attraction for their partner than when no virtual image is visible.

Hypothesis IX: When a virtual image is visible, people will feel partner satisfaction
than when no virtual image is visible

The model predicted that people would feel that their interaction partner
was closer to human when their partner was represented by a virtual image. This
was predicted to make more of the categories traditionally reserved for human
others salient. If this were true, even an agent represented by an image would be
perceived to be more human than one not represented by a human. Further, it
would make humans represented by virtual images seem more real and more
attractive than those not represented by virtual images. If the model was correct,
people would have felt more social attraction and partner satisfaction for others
represented by a virtual image than for those not represented by a virtual image.

Hypothesis X: When a virtual image is visible, people will feel less uncertainty in
judgments of others than when no virtual image is visible.

The theoretical model predicted that even when information was not
technically diagnostic of the judgment being rendered (such as the appearance
of a highly anthropomorphic virtual image on attributions of partner satisfaction
and social attraction), more information about another would increase confidence
of attribution and would increase the perception that the other was familiar (Gill et
al., 1998; Swann 8 Gill, 1997). If the model was correct, the presence of a virtual
image would make people feel more connected to and able to identify with the
intelligent other and they would feel as if they have more information about the
other. This would specifically result in less uncertainty regarding their interaction

partner. Although the model predicted that people would continue to create

62

mental models of others who were not represented by bodies, it also predicted
that the appearance of an image, whether natural or virtual, would lead to more
confident social judgment of the other.

Hypothesis XI: Judgments of others will be more extreme when a virtual image is
perceivable.

The theoretical model predicted that the appearance of an image, whether
virtual or natural, would cause categories that had been traditionally reserved for
human others to be salient. Thus, social judgments traditionally reserved for
human others, such as likeability and credibility as discussed above would be
more salient and meaningful. If the model was correct, then people would feel
more strongly about the social judgments of others when they were represented
by a virtual image than when they were not represented by a virtual image. If
people felt more strongly about their social judgments of others, then their
judgments would be more extreme and less neutral than when people did not
feel strongly about the social judgment.

MORE ANTHROPOMORPHIC VIRTUAL BODIES GENERATE MORE

POSITIVE SOCIAL JUDGMENTS OF HUMAN AND NON-HUMAN
INTERACTANTS

This section considers the influence of anthropomorphism on people’s
sense of presence in the physical environment and their sense of copresence
with virtual entities as well as the influence on social judgment. Anthropomorphic
bodies have been shown to engage the user more and provide a more
interesting interaction for participants (Koda, 1996; Takeuchi 8 Naito, 1995). The

theoretical model predicted that the influence of a virtual image will be most

63

extreme when the virtual image was a more anthropomorphic image. Thus, the
appearance of the image was predicted to influence the degree to which the
intelligent other was perceived to be human. This in turn was predicted to
influence the people’s sense of copresence, physical presence, partner
satisfaction and social attraction. The influence of a more anthropomorphic
image was predicted to be very influential in the type of mental model people
formed of another entity (Koda, 1996; Takeuchi 8 Naito, 1995).

Hypothesis XII: People will perceive intelligent others with highly
anthropomorphic images as more socially attractive than those with low-
anthropomorphic images.

Hypothesis XIII: People will be more satisfied with partners represented by highly
anthropomorphic virtual images than those with low-anthropomorphic
images.

The theoretical model predicted that people would feel that the more
anthropomorphic character was more human and thus were more likely to rate
the highly anthropomorphic avatar or agent as higher in attractiveness and
satisfaction with partner than the low anthropomorphic or no image person. The
model predicted that people who were able to interact with a partner who was
represented by a highly anthropomorphic virtual image would be more satisfied
with their potential interaction partner. If the model were correct, then in virtual
environments, individuals represented by highly anthropomorphic avatars would
be rated as more socially attractive than people represented by less

anthropomorphic avatars or people with no visual representation at all.

Hypothesis XIV: People will feel more copresent with a partner represented by a
highly anthropomorphic image than those with low-anthropomorphic
images.

The theoretical model predicted that people would treat any intelligent
others as social actors, especially when they were highly anthropomorphic, but
that there would be differences in the type of information processed and how
thoroughly it is processed. The model also predicted that because people have
experience feeling copresent with other humans, a more anthropomorphic image
would remind a person of other humans they have felt copresence with and

increase the feeling that they were connected with their interaction partner.

Insert Table II Here

65

Chapter 4

Method

DESIGN

This study used a between subjects experimental design with two factors.
The first factor, degree of anthropomorphism of virtual image had three levels,
high, low and no virtual image. The second factor, anthropomorphism of the
intelligent other, had two levels; whether participants were told they were

interacting with a human (avatar) or a bot (agent).

Insert Table III about here.

The dependent variables included copresence, physical presence, social
attraction, partner satisfaction and attractiveness evaluations. Also, this included
a measure of the degree of certainty of the attribution of personality variables and
sex category of the perceived person and an exploratory measure”.

PARTICIPANTS

134 undergraduates at a large midwestern university took part in this

experiment for class credit. Participants were stratified by sex and randomly

 

’5 social presence

66

assigned to condition. There were a total of 94 males and 40 females. The
average age of participants was 21, and ages ranged from 19 to 33.

STIMULUS MATERIALS

Virtual Environment. This interaction took place in a 3-dimensional
environment created in the MIND Lab using Director, V-Realm and 3D Studio
Max, all are computer software programs used to provide the illusion that
participants were interacting in a real time virtual environment. This environment
appeared on a computer screen and resembled a meeting room. In all
conditions, there was a sign indicating that participants were in the scavenger
hunt meeting place. In the no image condition, only the environment was
perceivable (See Figure 2). In the other conditions, participants first went to the
character selection screen (See Figure 3), where they selected the character
they wanted to represent them during the interaction. When participants
“entered” the virtual environment, the virtual face representing the virtual
confederate (agent or avatar) was already in the room, ready to begin the
interaction. The difference was in whether the virtual face is highly
anthropomorphic (See Figure 4) or less anthropomorphic (See Figure 5). This
environment allowed participants to speak into a microphone and hear their
interaction partner via headphones. lt resembled an interaction between

individuals across networked computers.

67

Avatars/Agents. The avatars and agents appeared the same and were
represented by one of two images; a female virtual face (high-anthropomorphic
condition, see Figure 4) or a very abstract face with only a mouth and eyes (Iow-
anthropomorphic condition, see Figure 5). All participants in each condition saw
the same virtual face, which was associated with the same female voice that
read the same script. The two images contained the same number of cues
including faces with eyes, mouths and noses. The only difference was that one is
more iconic (low anthropomorphism) and one appeared more human (high
anthropomorphism). In the control condition, there was no virtual image, only the
virtual room and the voice reading the script.

Audio Scripts. A female read the script. All participants in all conditions
heard the same voice reading the same script (See Appendix A. The content of
the script contained a short biography, listing their skills relevant to a web
scavenger hunt task. The script began with the virtual confederate (agent or
avatar) introducing themselves and then waited for the participant to do the
same. Following the introduction, the participant pressed a button indicating they
had introduced themselves and then the virtual confederate (agent or avatar)
discussed their Internet skills. The script included information about the virtual
confederate’s (agent or avatar) self-proclaimed web experience. Following
participant’s “turn” to indicate their skills, the virtual confederate said goodbye

and indicated a wish to continue working with the participant. Participants were

68

then given a chance to say goodbye before pressing the button to end the
interaction.

MEASUREMENT INSTRUMENTS

The pre-test questionnaire was done on paper and pencil (See Appendix
B). The post-test questionnaire (see Appendix C) was presented to the
participants using Surveysaid, a computerized questionnaire software (Masters,
2000). This Software presents each question one at a time on the computer
screen. Participants select their response by clicking the appropriate box with a
mouse and then click a button marked “next question” to continue through the
questionnaire.17

Presence. Presence is a measure of the feeling a person has that they
are “inside” a virtual environment, a sense of “being there.” This measure comes
from a development of a measure for presence with the subheading of presence
as immersion (Lombard 8 Ditton, 1999). Eight Iikert-type items with a 7-point
metric were used to form a scale (see Appendix C, items 28-35). This scale
included indicators such as how intense the experience in the environment was
and the extent to which the experience was involving and immersive.

Copresence. Copresence as discussed above is related to the feeling of
connection between two people. Given its dual nature, this was measured by two

separate scales, one asked about the participant’s perception of their partner’s

69

involvement in the interaction and the other asked the participant about their own
involvement in the interaction.

The scale measuring the perceived copresence of the virtual confederate
(agent or avatar) included 15 indicators taken from three of the dimensions of
immediacy (see Appendix C, items 2-16). This included immediacy/affection,
similarity/depth and receptivity/trust. This scale was derived from a combination
of the indicators for intimacy, involvement and immediacy from Burgoon and Hale
(1987). This included whether the other was perceived to be involved, interested
or emotional about the conversation. It also included whether or not the
interaction partner made the conversation seem superficial or created a sense of
distance between the interaction partners. These were likert-type items with a
five-point metric.

The second scale included 11 indicators similar to those above, but they
were revised to ask how involved the participant was in the interaction (see
Appendix C, items 17-27). These items measured the extent to which the
participant self-reported being copresent in the interaction and included
questions about whether they were interested in a deeper relationship or more
intimate conversation with their interaction partner. These were also likert-type

items with a five-point metric.

 

'7 Thus, the instructions appear above every question instead of just above a cluster of questions
in the printed version of the text.

70

Social Attraction. A 9-item likert-type scale with a 7-point metric
measuring social attraction was modified for the purposes of this project (see
Appendix C, items 36-44). This scale came from McCroskey and McCain (1974)
and included questions about the extent to which they feel the other person could
be a friend, was pleasant or offensive and whether or not the participant desired
a future interaction.

Partner Satisfaction. Partner Satisfaction included two separate scales.
One was a Iikert-type scale consisting of ten indicators of partner satisfaction
from McCroskey and Wright (1971), including whether the confederate was
professional, cooperative and knowledgeable (see Appendix C, items 59-68).
The second scale was a semantic differential scale consisting of 6 items
including whether the participant considered their partner reliable, informed or
intelligent (see Appendix C, items 69-74).

Certainty. Confidence was based on Clatterbuck's (1979) 7-item
Attributional Confidence Scale (see Appendix C, items 45-51 ). This included
items asking whether people were completely confident or unable to answer how
their partner would behave in the future and whether they felt able to predict their
attitudes, values and feelings.

Demographics. Demographic questions included previous experience
with computers, chat rooms and computer games, as well as their sex, ethnicity,

age and year in school (see Appendix B).

71

PROCEDURE

After signing a consent form indicating their voluntary participation in this
experiment, participants filled out their pretest using paper and pencil in a
conference room. The pretest asked for demographic information, including
biological sex, age, year in school and computer use and chat room experience.
Participants then read a piece of paper containing their assignment and the
scenario. They were told that their goal was to get to know their partner who may
work with them in the future on a scavenger hunt on the World Wide Web. There
were two versions of the instruction sheet with only one difference. In the avatar
condition, participants were told they were to interact with a student from another
university (See Appendix D). In the agent condition, participants were told they
were interacting with a “bot” or computer programmed to be interactive (see
Appendix E). The instruction sheet also told participants that if their team were
selected they would be asked to participate in an online scavenger hunt where
they would search for a variety of computer software and technologies. Further,
they were led to believe that the team that finished the scavenger hunt and found
the best prices in the least amount of time would win $100. The participant was
then taken to a computer in a computer laboratory and the instructions were
explained again in more detail. The participant was verbally reminded that their
partner was either an agent or an avatar and any questions participant had were

answered at this time.

72

In the no image condition, participants entered their participant
identification numbers and began the interaction. In either of the other
conditions, participants entered their participant id numbers, selected a virtual
face to represent them and then began the interaction. Although the participants
were told they were interacting with their interaction partner in real time, they
were exposed to pre-recorded scripts”. All of them heard the same script
associated with the same virtual face. Their interaction partners are virtual
confederates whether in the agent condition or in the avatar condition. In the
embodied conditions the virtual confederates were represented by a character,
either highly anthropomorphic or less anthropomorphic. In the no image control
condition, the virtual confederates were not represented by a virtual image at all.

After the interaction, participants completed the post-test questionnaire.
This included measures of presence, copresence, social attraction, partner
satisfaction, and certainty. Finally, participants responded to the exploratory
measures of social presence and filled out open-ended questionnaires aimed at
exploring participants' perceptions of (the virtual images and the interaction as
well as exposing participant suspicion or difficulty with the manipulation or the

experience. After participants completed the questionnaires, they were debriefed.

 

’° During open-ended questions at the end of the interaction, 28 participants expressed suspicion
that their interaction partner was pre-recorded. The implications of this for the agent/avatar
manipulation are discussed below.

73

Chapter 5

Results

Hypotheses

An alpha level of .05 was used for all statistical tests.

SCALE CONSTRUCTION

Standardized item alpha is included for all scales. The dimensionality of
each scale was determined in two ways. First, confirmatory factor analysis tests
of internal consistency were applied to each instrument. All retained items met
the criteria for internal consistency: (a) face validity and (b) a primary factor
loading of at least .5. Further, items were removed when they had greater errors
with other items than what would be expected by sampling error. Items were
dropped from their respective scales when item correlations failed tests of
internal consistency. Second, all scales were evaluated together and all items
loaded highest on their primary factor. Any item that did not meet all tests was
removed from the scale. The final number of items in each scale is detailed when

the scale is first used in analysis.

74

ON THE INFLUENCE OF AGENCY

A one-way ANOVA for Agency was conducted for all analyses in this

section.

(Insert Figure 6 Here)
Hypothesis I: People will feel more social attraction for avatars than for agents.

This hypothesis was not supported.

Of the 9 Iikert-type items originally included in the scale of social attraction
(Standard alpha = .83), six items remained after the tests of internal consistency
and reliability (See Appendix C, items 41, 42 and 43 were removed), and
responses on this scale ranged from 6 to 42 (See Table IV for items remaining in
this scale).

The effect of agency on social attraction is not statistically significant, E_(1,
132) = .07, Q = .80. Means (with standard deviations in parentheses) for those in
the agent condition and avatar condition were 22.39 (6.36) and 22.10 (6.86),
respectively.

In an open-ended questionnaire, 28 participants (10 in the agent condition
and 18 in the avatar condition) expressed doubt that they were interacting with
another entity in real time, or doubt that they were interacting with another human
(in the avatar condition). As a control, in the analysis below, those participants
were removed from the analysis. Eliminating those participants did not influence

this result. The effect of agency on social attraction is still not significant, E(1,

75

104) = .10, p = .75. Means (with standard deviations in parentheses) for those in
the agent condition and avatar condition were 22.27 (6.33) and 21.86 (6.90),

respectively

Hypothesis II: People will feel more partner satisfaction with a human interaction
partner than with a bot.

This hypothesis was not supported on either scale.

All ten likert-type indicators of partner satisfaction remained after the tests
of internal consistency and reliability (Standard alpha = .89) and responses on
this scale ranged from 10 to 41 (See Table V). Also, all six of the semantic
differential partner satisfaction items remained after tests of internal consistency
and reliability (Standard alpha = .91) and responses on this scale ranged from 6
to 28 (See the bottom of Table V).

The effect of agency on partner satisfaction is not significant, E(1, 132) =
.79, g = .38. Means (with standard deviations in parentheses) for those in the

agent condition and avatar condition were 18.37 (6.75) and 17.42 (5.65),
respectively. Eliminating those participants who expressed doubt about whether
they were interacting with another in real time did not influence this result. The
effect of agency on partner satisfaction is still not significant, E(1, 104) = 1.43, p =
.23). Means (with standard deviations in parentheses) for those in the agent
condition and avatar condition were 18.16 (6.26) and 16.83 (4.99), respectively.

The effect of agency on partner satisfaction as measured with the

semantic differential scale is also not significant, 5(1, 132) = .27, p = .61. Means

76

(with standard deviations in parentheses) for those in the agent condition and
avatar condition 13.58 (4.49) and 13.18 (4.42), respectively. Eliminating those
participants who expressed doubt about whether they were interacting with
another in real time does not influence this result. The effect of agency on partner
satisfaction is still not significant, E_(1, 104) = .18, p_ = .68. Means (with standard
deviations in parentheses) for those in the agent condition and avatar condition

were 13.40 (4.13) and 13.06 (4.19), respectively.

Hypothesis III: People will feel more copresent with a human interaction partner
than with a bot.

This hypothesis was not supported on either dimension.

The scale measuring the perceived copresence of the virtual confederate
(agent or avatar) maintained 13 of the original 15 indicators after the tests of
internal consistency and reliability (standard alpha = .9) and responses on this
scale ranged from 15 to 60 (See Appendix C, items 4 and 5 were removed, see
Table VI for retained items). Of eleven items originally included in the scale of
self-reported copresence, five items remained after the tests of internal
consistency and reliability (Standardized alpha= .78) and responses on this scale
ranged from 6 to 30 (See Appendix C, items 17, 18, 19, 21, 25 and 26 were

removed, see Table VII for retained items). The items that were dropped

appeared to be bad items in this manipulation”.

 

’9 All of these included a measure of how involved, detached or intimate this particular
conversation was. They were not related directly to their perception of their interaction partner.
Another manipulation may be a better test of this dimension of the construct.

77

The effect of agency on the perception that the virtual confederate (agent
or avatar) was copresent is not significant, _E(1, 132) = .27, p = .60. Means (with
standard deviations in parentheses) for those in the agent condition and avatar
condition 36.48 (9.76) and 37.37 (9.96), respectively. Eliminating those
participants who expressed doubt about whether they were interacting with
another in real time did not influence this result. The effect of agency on
perceived copresence of the virtual confederate (agent or avatar) is still not
significant, E(1, 104) = .13, p = .72. Means (with standard deviations in
parentheses) for those in the agent condition and avatar condition were 36.46
(9.99) and 35.77 (9.80), respectively.

The effect of agency on the participant’s self-reported copresence in the
interaction is not significant, E(1,132) = .37, p = .54. Means (with standard
deviations in parentheses) for those in the agent condition and avatar condition
were 17.39 (4.13) and 16.92 (4.77), respectively. Eliminating those participants
who expressed doubt about whether they were interacting with another in real
time does not influence this result. The effect of agency on the participant’s self-
reported copresence in the interaction is still not significant, E(1, 104) = .96, p =
.33. Means (with standard deviations in parentheses) for those in the agent
condition and avatar condition were 17.44 (4.38) and 16.54 (5.10), respectively.

Hypothesis IV: People will feel less uncertainty in judgments of avatars than
agents

This hypothesis was not supported.

78

All seven indicators from the Attributional Confidence Scale remained after
tests of internal consistency and reliability (Standard alpha = .84) and responses
on this scale ranged from 7 to 32 (See Table VIII).

The effect of agency on uncertainty is not significant, _E(1, 132) = 2.90, p =
.09. Means (with standard deviations in parentheses) for those in the agent
condition and avatar condition were 15.88 (6.51) and 14.17 (5.00), respectively.
Eliminating those participants who expressed doubt about whether they were
interacting with another in real time does not influence this result. The effect of
agency on uncertainty is still not significant, E(1, 104) = .96, p = .33). Means (with
standard deviations in parentheses) for those in the agent condition and avatar
condition were 15.79 (6.46) and 14.28 (5.00), respectively.

Hypothesis V: Judgments of avatars will be more extreme than judgments of
agents

A folded scale technique was used to test this hypothesis. The scales are
measured from 1 (not at all) to 7 (very much) with 4 resulting in a neutral
judgment of the other or from 1 to 5 with 3 resulting in a neutral judgment. The
folded scale technique recoded the variables to uncover the absolute values from
neutral of the attribution of the other. In this way, a 4 (or 3) rating is neutral and a
1 or a 7 (or 5) are on the absolute extremes.

This hypothesis was not supported on any measure.

On both of the measures of copresence, a main effect for Agency is not

significant.

79

The effect of agency on the extremity of perceived copresence of the
virtual confederate (agent or avatar) is not significant, _E(1, 132) = 1.21, 9 = .27.
Means (with standard deviations in parentheses) for those in the agent condition
and avatar condition were 24.74 (5.25) and 25.71 (4.92), respectively.

The effect of agency on the degree of copresence self-reported by
participants is not significant, E(1, 132) = .21, p = .65. Means (with standard
deviations in parentheses) for those in the agent condition and avatar condition
were 11.34 (2.37) and 11.54 (2.60), respectively.

The effect of agency on social attraction is not significant, E(1, 132) = .08,
g = .78. Means (with standard deviations in parentheses) for those in the agent
condition and avatar condition were 13.03 (3.32) and 13.21 (4.22), respectively.

On both measures of partner satisfaction a main effect for agency is not
significant. The effect of agency on partner satisfaction is not significant, _E(1,132)
= .22, p = .64. Means (with standard deviations in parentheses) for those in the
agent condition and avatar condition were 23.44 (4.68) and 23.81 (4.57),
respectively.

The effect of agency on the semantic differential scale of partner
satisfaction is not significant, E(1,132) = .04, p = .84. Means (with standard

deviations in parentheses) for those in the agent condition and avatar condition

were 1 6.84 (3.76) and 16.98 (4.11), respectively.

80

Hypothesis VI: People that feel more physical presence in the environment will
feel more copresence with their interaction partner.

This hypothesis was supported.

For the scale of presence (standard alpha= .88), six of the 8 original items
were retained after tests of internal consistency and reliability (See Appendix C,
items 33, 34 and 35 were removed, see Table IX for retained items)

There is a significant positive correlation between copresence and
physical presence. There is a positive correlation between presence and the
degree to which participants felt their partners were copresent (r = .46, p = .00),
as well as a significant positive relationship between the degree to which people
felt themselves to be copresent in the interaction and the degree to which they
felt present in the environment (5 = .54, p = .00). The correlation between the
degree to which people felt copresent and the perceived copresence of the
virtual confederate (agent or avatar) was also significant, but smaller, ([ = .33, Q =
.00). People that felt more copresence with their interaction partner on either
dimension also felt more physical presence in the environment.

INFLUENCE OF A VIRTUAL IMAGE

A one-way ANOVA with a priori contrasts (-2 no image, +1 high-anthropomorphic
image, +1 low-anthropomorphic image) for the effect of Agency was conducted

for all analyses in this section.

Hypothesis VII: When a virtual image is visible, people will feel copresence with
their partner than when no virtual image is visible.

This hypothesis was not supported.

81

This hypothesis was tested for both the perceived copresence of the
virtual confederate (agent or avatar) and the self-reported copresence of the
participant.

The effect of a visual virtual image on people’s perception that the virtual
confederate (agent or avatar) is copresent is not significant, 1(2, 131) = -1.25, p
= .22. Means (with standard deviations in parentheses) for those in the no image
condition, low-anthropomorphic condition and high-anthropomorphic condition
and were 38.36 (9.52), 38.83 (9.46), and 33.47 (9.83), respectively and
responses on this scale ranged from 15 to 60.

The effect of the existence of a visual virtual image on people’s self-
reported copresence is not significant, 1(2, 131) = -.57, p = .57). Means (with
standard deviations in parentheses) for those in the no image condition, high-
anthropomorphic condition and low-anthropomorphic condition were 17.45

(4.17), 17.96 (4.69), and 16.01 (4.29), respectively.

Hypothesis VIII: When a virtual image is visible, people will feel more social
attraction for their partner than when no virtual image is visible.

This hypothesis was not supported.

The effect of a visual virtual image on social attraction is not significant,
1'(2, 131) = -.58, p = .56. The means are not in the predicted direction. Means
(with standard deviations in parentheses) for those in the no image condition,

high-anthropomorphic condition and low-anthropomorphic condition were 22.67

(6-05). 24.00 (6.47), and 19.95 (6.70), respectively.

82

 

Hypothesis IX: When a virtual image is visible, people will feel partner satisfaction
than when no virtual image is visible.

This hypothesis was not supported.

The effect of a visual virtual image on partner satisfaction is not significant
_'l_'(2, 131) = .47, p = .64). Means (with standard deviations in parentheses) for
those in the no image condition, high-anthropomorphic condition and low-
anthropomorphic condition were 17.51 (5.84), 19.35 (7.07) and 16.75 (5.42),
respectively.

The effect a visual virtual image on partner satisfaction with the semantic
differential scale is not significant, I( 2, 131) = .77, Q = .44. Means (with standard
deviations in parentheses) for those in the no image condition, high-
anthropomorphic condition and low-anthropomorphic condition were 12.93
(4.03), 14.51 (5.00) and 12.61 (4.02) respectively and responses on this scale

ranged from 6 to 28.

Hypothesis X: When a virtual image is visible, people will feel less uncertainty in
judgments of others than when no virtual image is visible.

This hypothesis was not supported.

The effect of a visual virtual image on uncertainty is not significant, 1(2,
131) = .58, p = .56. Means (with standard deviations in parentheses) for those in

the no image condition, high-anthropomorphic condition and low-

anthropomorphic condition were 14.63 (5.42), 14.49 (5.88) and 16.02 (6.23),

respectively and responses on this scale ranged from 7 to 32.

83

Hypothesis XI: When a virtual image is visible, people will make more extreme
judgments of others than when no virtual image is visible.

This hypothesis was not supported.

Tests of this hypothesis used the folded scale technique discussed above
and a one-way ANOVA with a priori contrasts (-2 no image, +1 high-
anthropomorphic image, +1 low-anthropomorphic image) on all dependent
variables.

The effect of a visual virtual image on extremity of perceived copresence
of the virtual confederate (agent or avatar) is not significant, 1(2, 131) = .31, p =
.76). Means (with standard deviations in parentheses) for those in the no image
condition, high-anthropomorphic condition and low-anthropomorphic condition
were 25.03 (4.81 ), 24.97 (4.94) and 25.67 (5.58), respectively.

The effect of a visual virtual image on extremity of self-reported
copresence is not significant, 1(2, 131) = 1.09, Q = .28. Means (with standard
deviations in parentheses) for those in the no image condition, high-
anthropomorphic condition and low-anthropomorphic condition were 11.10
(2.23), 11.70 (2.61) and 11.49 (2.56), respectively.

The effect of a visual virtual image on extremity of social attraction is not
significant, 1(2, 131) = .71, p = .48. Means (with standard deviations in
parentheses) for those in the no image condition, high-anthropomorphic condition

and low-anthropomorphic condition were 12.80 (3.71), 12.82 (3.86) and 13.76,

(3.75), respectively.

 

The effect of a visual virtual image on extremity of partner satisfaction is
not significant, 1(2, 131) = -.62, p = .54. Means (with standard deviations in
parentheses) for those in the no image condition, high-anthropomorphic condition
and low-anthropomorphic condition were 24.00 (4.33), 22.72 (4.87), and 24.22
(4.56), respectively.

The effect of a visual virtual image on extremity of uncertainty is not
significant, 1(2, 131) = -.29, g = .77). Means (with standard deviations in
parentheses) for those in the no image condition, high-anthropomorphic condition
and low-anthropomorphic condition were 16.53 (2.66), 16.75 (2.90) and 16.02
(2.86), respectively.

INFLUENCE OF AN ANTHROPOMORPHIC VIRTUAL IMAGE

(Insert Figure 7 Here)

Hypothesis XIII: People will perceive intelligent others with highly
anthropomorphic images as more socially attractive than those with low-
anthropomorphic images.

This hypothesis was supported.

This was tested with a one-way ANOVA with a priori contrasts (0 no
image, +1 high-anthropomorphic image, -1 low-anthropomorphic image). The
effect of a highly anthropomorphic virtual body on social attraction is significant,
1(2, 131) = 3.01, g = .00). This was confirmed by a one-way ANOVA. The effect
of a highly anthropomorphic body on social attraction is significant, _F_(2, 131) =

4.66, Q = .01 ). Means (with standard deviations in parentheses) for those in the

85

no image condition, high-anthropomorphic condition and low-anthropomorphic
condition were 22.67 (6.05), 24.00 (6.47), and 19.95 (6.70), respectively.
Participants who interacted with high-anthropomorphic virtual images felt
significantly more social attraction than those that interacted with low-
anthropomorphic virtual images. Those interacting with high-anthropomorphic
virtual images felt the highest social attraction and those interacting with low-
anthropomorphic virtual images felt the lowest social attraction.

This was also tested using a one-way ANOVA with contrasts (1 no image,
1 high-anthropomorphic image, -2 low-anthropomorphic image). The effect of a
high-anthropomorphic virtual body on social attraction is significant, 1(2, 131) =
2.86, g = .01. People who interacted with a low-anthropomorphic image felt less
social attraction than either those interacting with no image or with a high-

anthropomorphic image.

Hypothesis XIV: People will be more satisfied with partners represented by highly
anthropomorphic virtual image than those with low-anthropomorphic
image.

This hypothesis was supported with both scales.

The contrast test (0 no image, +1 high-anthropomorphic image, -1 low-
anthropomorphic image) showed that the effect on partner satisfaction of a low-
anthropomorphic image compared to a high-anthropomorphic image is
significant, 1(2, 131) = 2.0, Q < .05. Participants who interacted with a high-
anthropomorphic image felt more partner satisfaction than those that interacted

with low-anthropomorphic virtual images. Means (with standard deviations in

86

parentheses) for those in the no image condition, high-anthropomorphic condition
and low-anthropomorphic condition were17.51 (5.84), 19.35 (7.07) and 16.75
(5.42), respectively.

This was also tested using a one-way ANOVA with contrasts (-1 no image,
+2 high-anthropomorphic image, -1 low-anthropomorphic image). The effect of a
high-anthropomorphic virtual body on partner satisfaction is significant, 1(2, 131)
= -1.98, p < .05. People who interacted with a low-anthropomorphic image felt
less partner satisfaction than either those interacting with no image or with a
high-anthropomorphic image.

The same tests were used with the semantic differential scale measuring
partner satisfaction. The contrast test (0 no image, +1 high—anthropomorphic
image, -1 low-anthropomorphic image) showed that the effect on partner
satisfaction of a low-anthropomorphic image compared to a high-
anthropomorphic image is significant, 1(2, 131) = 2.07, p_ = .04. Means (with
standard deviations in parentheses) for those in the no image condition, high-
anthropomorphic condition and low-anthropomorphic condition were 12.93
(4.03), 14.51 (5.00) and 12.61 (4.02), respectively. Participants interacting with a
high-anthropomorphic image felt more partner satisfaction than either those
interacting with no image or the low-anthropomorphic image.

This was also tested using a one-way ANOVA with contrasts (0 no image,

+1 high-anthropomorphic image, -1 low-anthropomorphic image). The effect of a

87

high-anthropomorphic virtual body on partner satisfaction is significant, 1(131) =
-2.19, p = .03. People who interacted with a less anthropomorphic avatar felt less
partner satisfaction than either those interacting with no image or with a high-
anthropomorphic image. People that interacted with high—anthropomorphic
virtual images felt more partner satisfaction than those in the no image condition
or than those who interacted with less anthropomorphic images.

Hypothesis XV: People will feel more copresent with a partner represented by a
highly anthropomorphic image than those with low-anthropomorphic
images.

This hypothesis was supported on both dimensions.

The contrast test (0 no image, +1 high-anthropomorphic image, -1 low-
anthropomorphic image) showed that the effect of an anthropomorphic image on
the perceived copresence of the virtual confederate (agent or avatar) is
significant, 1(2, 131) = 2.66, p = .01. Means (with standard deviations in
parentheses) for those in the no image condition, high-anthropomorphic condition
and low-anthropomorphic condition were 38.36 (9.52), 38.83 (9.46) and 33.47
(9.83), respectively. This was confirmed by a one-way ANOVA. The effect of an
anthropomorphic body on perceived copresence of the virtual confederate (agent
or avatar) is significant, E(2, 131) = 4.26, p = .02. Participants interacting with a
high-anthropomorphic image felt more partner satisfaction than either those
interacting with no image or the low-anthropomorphic image.

This was also tested using a one-way ANOVA with contrasts (+1 no

image, +1 high-anthropomorphic image, -2 low-anthropomorphic image). The

88

effect of an anthropomorphic virtual body on perceived copresence of the virtual
confederate (agent or avatar) is significant, 'l_'(2, 131) = 2.90, g = .00. People
interacting with the less anthropomorphic virtual image reported their partners to
be less copresent than those who interacted with either a high-anthropomorphic
virtual image OR with no image.

The same trend in the results was found with regard the participants' self-
reported copresence in the interaction. People that interact with highly
anthropomorphic virtual images felt more copresent than those that interact with
low-anthropomorphic virtual images.

This was tested with a one-way ANOVA with a priori contrasts (0 no
image, +1 high-anthropomorphic image, -1 low-anthropomorphic image). The
effect of an anthropomorphic virtual body on self-reported copresence is
significant, I(2, 131) = 2.11, g = .04. Means (with standard deviations in
parentheses) for those in the no image condition, high-anthropomorphic condition
and low-anthropomorphic condition were 17.45 (4.17), 17.96 (4.69) and 16.01
(4.29), respectively. People that interacted with highly anthropomorphic virtual
images felt more copresent than those that interacted with low-anthropomorphic
virtual images.

This was also tested with a one-way ANOVA with contrasts (+1 no image,
+1 high-anthropomorphic image, -2 low-anthropomorphic image). The effect of a

low-anthropomorphic image on people’s self-reported copresence is significant,

89

1(2, 131) = 2.09, p: .04). People interacting with a less anthropomorphic virtual
image felt less copresence with their partner than those interacting with either no

image or with a highly anthropomorphic virtual image.

Chapter 6

Discussion

The results of this project have shed light on how people perceive and
respond to mediated others in social virtual environments. They specifically have
provided further insight into the influence of an anthropomorphic virtual image on
the social judgment of a mediated other. This chapter explores the findings in
relationship to the theoretical model (Figure 1), with specific attention to the
implications of the results on the distinction between the perception of agents
and avatars, and of the existence of anthropomorphic virtual images on the
feelings of social attraction, partner satisfaction, uncertainty, presence and
copresence and suggests a revised theoretical model in light of the results. The
final section of this chapter explores the potential limitations of the manipulation
and findings.

ON THE DIFFERENCE BETWEEN HUMAN AND NON-HUMAN
INTERACTANTS

The model predicted that people would be less involved in interactions
with agents and that they would be seen as less human and less deserving of
social responses than would avatars. This led to the theoretical prediction seen in
Figure 1 that is represented by the Agent/Avatar circle just to the left of the
Mental Model of Interactant 8 Environment square. This circle represents the

classification of whether one is perceived to be an agent or an avatar.

91

The model predicted that the mere suggestion that one was not interacting
with a human would influence social judgment. This was specifically tested by
hypotheses I through V (See Table II). Tests of these hypotheses examined the
influence of the perception of avatars as compared to the perception of agents on
judgments of social attraction, uncertainty, partner satisfaction and copresence
as well as the extremity of social judgment. The results did not support this
prediction on any measure and no evidence of a distinction was found between
the perception of avatars and the perception of agents. This led to the removal
of the agent/avatar distinction from the far left side of the model in the revised
model (See Figure 8), but one cannot conclude that there is no influence on
social judgment and mental models of others of the distinction between agents
and avatars, only that this manipulation and these measurements did not reveal
any distinctions.

The fact that participants did not respond in discernibly different ways to
agents as compared to avatars strengthened the conclusion that people respond
socially to all intelligent others, whether human or not. It supported the findings
of other researchers showing that people respond to intelligent non-human
others in the same ways they respond to human others (Reeves 8 Nass, 1996)
and the script read by the virtual confederate (agent or avatar) was sufficient to
qualify the partner as an intelligent other. There are other possible explanations
for these results and interpreting null results is always questionable, but the fact
that it was consistent with previous research makes this conclusion easier to

support. As can be seen from the revised model (figure 8), the agent/avatar

92

distinction may have influenced the mental model in some way, but directionality
and the judgments it influences are as yet undetermined.

ON THE EXISTENCE OF A VIRTUAL IMAGE

The theoretical model predicted that any visible virtual image would lead
to increased social judgment and involvement in the interaction. This led to the
theoretical prediction seen in Figure 1 that is represented by the Embodiment
Visible/Not and Body Morphology Looks Human/Not circles on the far left side of
the model. These circles represented the classifications of whether or not there is
a visible image present and whether or not that image was anthropomorphic.

In essence, this prediction was based on the notion that when an
interaction partner was represented by any virtual image, they would be more
likely to be classified as human or more human-like than when there was no
virtual image. This notion was tested by hypotheses VII through Xl (See Table II).
Tests of these hypotheses specifically examined whether or not participants felt
more copresence, social attraction, partner satisfaction, uncertainty or extremity
of social judgment when participants interacted with a partner represented by a
virtual image as compared to those who interacted with a partner not represented
by a virtual image. Contrary to what was predicted above, the simple existence of
a virtual image did not lead to either the perception that one was human or to
increased social judgment or response.

NOT ALL VIRTUAL IMAGES ARE EQUAL

People’s experience in the natural world has led them to rely on the

physical, relatively unchanging features, especially those related to the physical

93

body for social judgment and to place others into meaningful social categories
(Argyle, 1975; Chesebro 8 Bonsall, 1989; Fiske 8 Neuberg, 1990; Goffman,
1963; Hinton, 1993; lchheiser, 1970). The process of categorizing others based
on physical appearance has been shown to be universal and automatic, not even
requiring conscious attention (Patterson, 1995; Tagiuri, 1958). The theoretical
model predicted that even in a highly structured task-oriented interaction people
' would continue to rely on the characteristics of the virtual image for category
assignment, whether they perceived their partner to be an agent or an avatar.
This was specifically tested in hypotheses XII, XIII and XIV (See Table II), which
examined the influence of an anthropomorphic virtual image on perceptions of
copresence, uncertainty, partner satisfaction and social attraction.

In examining the implications of the results of these hypotheses on the
theoretical model (Figure 1), note that the arrows indicated the prediction that the
existence and/or shape of the virtual image would influence the perception of
another as avatar or agent. The mere existence of the virtual image did not
influence perception in the direction that was predicted by the model. In other
words, the existence of a virtual image did not result in an increased perception
of humanness as predicted. This does not imply that the presence of a virtual
image did not influence perception; it only means that the influence of a virtual
image was not always in the direction of a more social, or human judgment. The
model did not predict that the influence of the virtual body would be this much

Stronger than the influence of the perception of whether the interaction partner

Was an agent or avatar. The perception of agents did not differ from the

perception of avatars when the virtual image was the same on any of the
dependent variables represented in Figure 1. Essentially, the results indicated
that when the image of an interaction partner was more anthropomorphic the
judgments were more social than when the image was less anthropomorphic.
Across the measures, the most positive perception was of the virtual
confederate (agent or avatar) represented by a highly anthropomorphic image as
the model predicted. Participants in the high-anthropomorphic image condition
perceived that the virtual confederate (agent or avatar) was more copresent in
the interaction than the virtual confederate represented by a low
anthropomorphic virtual image. Further, participants in the high anthropomorphic
condition reported feeling more copresence with the confederate, were more
socially attracted to their partner, and they reported more partner satisfaction.

THE FEELINGS OF COPRESENCE AND PRESENCE ARE
INTERCONNECTED

As predicted by the model, both dimensions of copresence were highly
and significantly correlated with the concept of presence. This was predicted in
hypothesis VI (See Table II) and tested with a correlation that showed a
significant relationship between the variables. This significant association
indicated one of two things: either the connection between people influenced
presence in the interface, or the presence in the interface influenced attention to
their interaction partner, the virtual confederate. It is difficult to ascertain the

direction of this influence and whether presence increases copresence or

Whether copresence increases presence. Future research should explore the

directionality of this relationship.

95

Even without making claims regarding causality, the relationship between
these variables represented an important consideration for both designers and
users of telecommunication technology. When people did not feel present in the
environment, they felt less satisfied with, and less of a connection with, their
interaction partner. Interestingly, presence was also negatively correlated with
uncertainty (r = -.45, p = .00). This indicated that the less uncertainty one feels
about their ability to predict their interaction partner’s behavior the more presence
they felt in the environment, or the more presence one felt in the environment the
less uncertainty they felt about their interaction partner.

As researchers continue their exploration of the influence of presence on
people’s mental processes they should also consider the importance of this
sensation on the perception of copresence with other people and objects in the
environment. Designers and users of interfaces for interpersonal interactions
ought to take careful note of this information. This could allow them to design or
choose media that best facilitates the needs of their interaction goals, promoting
the necessary levels of both presence and copresence.

ON REVISING THE MODEL AND UNDERSTANDING THE INFLUENCE OF
THE VIRTUAL IMAGE

The results supported the notion that people have continued to use the
visible features of the body, whether virtual or natural, to make social judgments
of others. The body influenced participants even when the interaction,
experience and body were mediated or virtual. Even a virtual image and
mediated interaction influenced participants’ social judgment of their interaction

partner. This was consistent with the notion that people in virtual environments

96

have used the characteristics of the body or image that were salient and familiar
based on their experience in the natural world for social judgment. More research
is needed to clarify the strength of the virtual body as compared to other
indicators and information in the interaction and environment.

It is important to look at the implications for the findings from hypotheses l-
V and VIl-Xl together (See Table II). This provided unique insight into the
influence of a visible virtual image on the perception of an intelligent other. It was
these groups of hypotheses together that contribute the most to understanding
the implications of the results on changes to the theoretical model. The
theoretical model (Figure 1) predicted first that people would respond more
socially to others represented by a virtual image - any virtual image. Further, the
model predicted that people would use the features of the image to form their
mental model of them. This part of the prediction indicated that a less
anthropomorphic virtual image would invoke less social response and less social
judgment as compared to those interacting with another represented by a more
anthropomorphic image. The model was incorrect in predicting that any virtual
image would increase the tendency of people to respond socially. However, the
model was correct to predict that the appearance of the virtual image would be
influential and that people represented by more anthropomorphic virtual images

would be responded to more socially.
The fact that the responses of participants that interacted with a virtual
body and those who did not were not significantly different should not be

interpreted to mean that the virtual body is not influential. Having a non-

97

anthropomorphic virtual image provided less social response than either the high
anthropomorphic virtual image or no image. In other words, people responded
more socially to those represented by more anthropomorphic images or by no
image at all than to those represented by a low anthropomorphic image. The
model does not accurately reflect the direction or extremity of the influence of a
low-anthropomorphic virtual image.

It is likely that the default mental model of an intelligent other included an
anthropomorphic image and, when the image was more anthropomorphic, the
image created by the mental model was not distinguishably different. On the
other hand, when the virtual image was less anthropomorphic, the difference
between the mental model and the image were very different, which caused a
negative, less social perception of the virtual confederate. Specifically, when the
virtual face did not appear human, the virtual confederate (agent or avatar) was
perceived to be a less intelligent other than when there was either no image or
when the virtual face was more anthropomorphic.

The model should have two different circles on the far left side (Figure 1).
These are a highly anthropomorphic virtual body and a low anthropomorphic
body. The results here did not show a difference between the existence of a
virtual body and the lack of one, there was no distinguishable difference between
those in the high anthropomorphic virtual image condition and those in the no
image condition. Having only two circles would more clearly show the difference

between a high anthropomorphic virtual image and a low anthropomorphic virtual

image (See revised model on Figure 8).

98

When virtual image was less anthropomorphic, the virtual confederate
represented by it was responded to in less social ways or as less human. This
supported the prediction that if it does not look human, it was not perceived to be
human, as long as it demonstrated some form of intelligence. In other words, just
because there was a virtual image did not mean the other was more likely to be
perceived as human, but the characteristics of the virtual image were the keys to
this distinction. It further supported the notion that people relied on their
experience in the natural world, where all living things have had certain common
properties that distinguished them from inanimate objects (Asch, 1958; Hastorf et
al., 1970; Heider, 1958). This means that, when both intelligence and the
physical properties traditionally ascribed to humans were seen in agents,
participants still perceived them to be human, or at least human-like enough to
form the same or an indistinguishable mental model of them.

In the theoretical model, having no image was predicted to lead to less
copresence and partner satisfaction than having any image. This was not
supported. The results supported the conclusion that having no image for a task-
oriented interaction was better for attributions of credibility and copresence and
partner satisfaction than having a less-anthropomorphic image.

This supported previous research showing that virtual images, especially
more anthropomorphic images, were more engaging, interesting and attractive

(Koda, 1996; Wexelblat, 1997) and that they influenced people’s perception of
their interaction partner. It appeared that when visual cues were presented

people used them, even if the indicators provided by the virtual image

99

contradicted other information provided during or before the interaction (Koda,
1996; Nowak 8 Anderson, 1999; Takeuchi 8 Naito, 1995). Further, the
judgments of the virtual confederate (agent or avatar) were related to judgments
that could logically be associated with the virtual image. The script spoken by the
virtual confederate (agent or avatar) was written to represent an individual very
qualified to be a good partner during the Internet Scavenger Hunt. People who
saw no image thought their partner was credible and Iikeable as did those who
saw the more human-like face.

The difference between the influence of the more anthropomorphic virtual
image and the low-anthropomorphic virtual image was hypothesized, but the
original theoretical model did not accurately reflect the definitive significance of
this influence. In this case, the appearance of the virtual image significantly

influenced all measures except for uncertainty.

POTENTIAL LIMITATIONS OF THIS MANIPULATION

In this section, specific issues are discussed that may limit or influence the
interpretation of the results. These include issues related to the interaction and
manipulation as well as unique properties of the virtual confederate (agent or
avatar) and the interface. First, whenever research is done in a laboratory there
are concerns regarding external validity. This project took place in a computer
laboratory where participants knew their behavior was being monitored, so they

may have behaved differently than they would have in a more natural setting.

100

FEATURES OF THE INTERFACE

The virtual environment used was a new interface and many participants
expressed discomfort interacting with another through a microphone attached to
a computer. Not all participants had interacted in SVEs before and was possible
that experienced users would have perceived this differently. If more time were
allowed to get used to the system, there may have been an influence on
uncertainty, but all participants were highly uncertain across conditions. It is
possible that it was the uncertainty in the interaction that was driving participant’s
responses and especially the lack of distinction between the perception of agents
and avatars. It is important to note that participants were from computer science
and telecommunication and were thus likely to react differently to this interface

than would less technologically oriented students.

ISSUES RELATED TO THE INTERACTION

This was a very limited interaction. A number of participants expressed
frustration at getting only 3 “turns” in the interaction. The interaction was highly
controlled and scripted. In the future, researchers should consider having people
interact in natural interactions but tell some they are interacting with an agent.
This would remove the control over script and content but may allow more
variation in the sense of copresence.

It was possible that the results would have been different if the interaction
had been more social or less controlled. Specifically, it is likely that this, in

Combination with participant’s uncertainty, influenced the results with regard to

101

measuring the copresence in this interaction. It is likely that the stiff nature of the
interaction kept people from feeling copresent in the interaction and this measure
should be tested in a more natural interaction to see how the items hold together
in different contexts. In other words, people did not feel compelled or immersed
in this interaction. It was short and controlled and participants repeatedly
mentioned that this did not allow them to get engaged in the interaction.

FEATURES OF THE VIRTUAL CONFEDERATE (AGENT OR AVATAR)

THE UNIQUENESS OF THE VIRTUAL IMAGES

The low-anthropomorphic virtual face was very iconic and odd.
Participants were unlikely to have experienced anything like it in the past. This
image was selected because it was very clearly not human, although it had a
face, it still did not resemble anything human, living or even a familiar object. It is
possible that an image could be less anthropomorphic like an animal, every day
object or other image that would not cause the negative perception to be as
extreme and unsocial as that seen by participants in this project. However, it is
the extremity and level of uniqueness of the low-anthropomorphic character that
made it less anthropomorphic. People anthropomorphize all familiar objects, so
the less familiar an object is the less anthropomorphic it would seem. It seems
that this iconic image was far from the default mental image people created when
no image was provided. These results may be tapping into some type of

expectancy violation or social cost related to the unexpected low-

102

anthropomorphic virtual face. Future research should further explore this
possibility.

The more anthropomorphic virtual image was also unique. The image
represented a woman with brown hair and eyes. The face was not very lifelike
and movements were halted and the contours of the face were not as smooth as
a natural human face would be. It may be that a face that is closer to a natural
human face would have made the results more extreme, but it is unlikely that this
altered the direction of the results. If there is a sufficiency threshold for
anthropomorphism, it appeared that this face reached it.

The images selected for this project were on the extremes; one highly
anthropomorphic and one very iconic and less anthropomorphic. This left open a
few questions including what would have happened if the virtual image was less
anthropomorphic and non-iconic or what would happen if the more
anthropomorphic image was slightly less anthropomorphic, more masculine or
less attractive. Future research should explore the influence of these factors.

THE VOICE AND SCRIPT

The script was written to represent an intelligent and highly qualified
partner. Also, the woman who read the script was chosen specifically because of
her ability to articulate well and sound credible. In the open-ended questions,
some participants wondered about the veracity of one person having all these
skills. It is possible, but unlikely, that the voice or the script influenced the
direction of the results. In other words, it was the script that made the partner

seem credible and Iikeable and by itself made participants perceive the virtual

103

confederate as credible and Iikeable. This may have partially influenced the
results but would not explain the difference between those in the high
anthropomorphic and those in the low anthropomorphic condition.

ISSUES RELATED TO THE MANIPULATION AND LAB SETUP

The manipulation was very task-oriented and restricted. One of the things
that distinguish the perception of humans from the perception of other living
things is the consideration of intentionality and emotion (Asch, 1958; Heider,
1958; Ritvo, 1991; Sheehan, 1991b; Turkle, 1991; Williams, 1991). It was
possible that the task—oriented nature of this interaction gave people the
perception of intentionality in both conditions and removed this consideration
from the perception of the other. In other words, the manipulation provided and
set up the intention of the perceived other, which neutralized this consideration
regardless of whether the virtual confederate was agent or avatar.

As considered in Chapter 6, there were a number of possible reasons for
the lack of influence of agency on people’s perception of their virtual partner.
Chapter 2 considered the degree to which the perceived intelligent other was
perceived to be human. It may be that no differences were detected in this
project because the virtual confederate (agent or avatar) was considered human
enough for the task at hand across the board. Bruner (1957) argued that people
assess how much information about another person they need to fulfill their goals
and use the resources and information available to fill the gap between what they
need to know and what they already know. It was possible that the task-oriented

nature of this interaction reduced the amount of information participants needed

104

to make these social judgments, given the context. Perhaps there was a
sufficiency threshold where an intelligent other begins to be perceived intelligent
or human enough for the purposes of the interaction. Finding this sufficiency
threshold, if it exists, may be an important area of future research. The results
from this project indicate that the sufficiency threshold included some indication
of intelligence, such as the script. The presence of the low anthropomorphic
virtual image seemed to have influenced this perception in negative ways, but it
was difficult to conclude why this happened without further research.

If there was indeed a sufficiency threshold for 'humanness' or a point at
which the other was human enough for the purposes of the interaction, then it
seems that the standard for social interactions would be higher than the standard
for task-based interactions. It may be that in this type of interaction the level of
humanness did not matter. It is possible that a more personal interaction would
uncover a distinction between human others and non-human others. Even if the
interaction was less controlled, it is unlikely that this would alter the fact that
agency did not influence perception or social judgment because it confirmed
similar findings in previous research using love tutorials (Reeves 8 Nass, 1996)
and other more personal situations which have also revealed a lack of distinction
between agents and avatars. Future research may manipulate intention as well
as amount of information required to fulfill interaction goals as variables to

examine the extent to which this influences perception.

105

THE AGENT/AVATAR MANIPULATION

There was the possibility that the manipulation did not work and that
people did not perceive they were interacting with a human in the avatar
condition, interacting with a bot in the agent condition, or even interacting with a
person or thing in real time in either condition, even though they were told in their
instruction sheet and verbally by the experimenter. At the end of the
questionnaire were open-ended questions designed to serve as a manipulation
check. At this point, 28 people expressed suspicion that they were interacting
with a human (in the avatar condition) or that they were interacting in real time
with a bot (in the agent condition).

Although eliminating the 28 people who expressed doubt that they were
interacting in real time did not change the direction of results, it was possible that
none of the participants believed they were interacting with a person in real time.
However unlikely, this could explain the lack of variance between conditions.
Researchers should carefully consider this possibility as they consider to ask
questions about the influence of agency on social judgment.

However, past research is consistent with the finding that people respond
to machines as they do humans (Reeves 8 Nass, 1996). Also, an open-ended
question at the end of the experiment directly asked participants what they
thought was really going on. If anything, this pointed question would lead to more
people expressing suspicion than actually felt it during the interaction. Future

research should explore these possibilities.

106

ENVIRONMENTAL ISSUES: LAB SETUP

Finally, participants were in a lab with eight computers and at times more
than one participant was in the lab at one time. Some participants were speaking
to their interaction partner while others were on other computers filling out post-
tests. A few participants mentioned that they felt self-conscious speaking into a
microphone to their partner when the rest of the room was relatively quiet. This
may have had an influence on perceived presence and copresence in the
interaction.

INTERACTING WITH INTELLIGENT OTHERS IN SVES; IMPLICATIONS OF
FINDINGS FOR INTERFACE DESIGNERS AND USERS

The features of the virtual image influenced social judgment in the virtual
world. A human looking other triggered more social judgments of human
intelligence in participants, even when they were aware that their interaction
partner was not human. This may have been related to people's experience in
the natural world where things that look similar are similar. Further, a lack of
experience with intelligent others who are not human and a lack of experience
with human others who do not look human (as discussed by Reeves and Nass,
1996) may explain why no distinction was made between an intelligent agent and
an avatar. Reeves and Nass (1996) predicted that this is due to a hard-wired
tendency to in human's categorization process that make the perception of
anthropomorphic images and intelligence to automatically register humanness.
While this is one possible explanation, there are other possible explanations as

well.

107

It may well be that the human experience in the natural world has only
provided interaction with human intelligent others and no intelligent others who
were not human. It is possible that the human brain has not adapted to dealing
with entities that look human but are not or entities that are human but do not
look it. Because of this, people may not make a meaningful distinction between
human and non-human intelligences. If this explanation is true, then this will
change over time as experience in social virtual environments becomes more
common. Perhaps further experience with non-human intelligent others will
promote the evolution of a more meaningful categorical distinction between
agents and avatars with new meanings and interpretations.

As predicted, on the level of visual anthropomorphism influenced social
judgment, which may have reflected experience in the natural world. The only
entities in the natural world that look human actually are humans, so using
anthropomorphism as a way to identify other humans has been useful and
meaningful. The results indicate that people interacting in SVEs continued to use
categories that have been useful in the natural world. Participants used a
combination of visual anthropomorphism and the script of an intelligent other;
future research should examine the extent to which other categories continue to
be salient and useful.

Participants who saw the less natural and iconic face rated their partner as
less capable than those who saw either the high-anthropomorphic image or no
image. It could be said that the low-anthropomorphic image led individuals to

make a less accurate judgment of their interaction partner because it led them to

108

believe the other was not human and therefore less intelligent, credible and
Iikeable. Future researchers should examine whether or not the reverse is true or
whether a script written to be less credible would be perceived as more credible
when read by a more anthropomorphic virtual image as compared to a less
anthropomorphic virtual image. If this is true, then the script was influential to
participants in that it made the virtual confederate seem intelligent and credible. If
it was the script that made the virtual confederate seem intelligent, then the
negative perception of participants in the low-anthropomorphic condition may
well be explained by an expectancy violation. In other words, an intelligent other
represented by a low anthropomorphic image sent mixed signals to perceivers
who then rated them as less social.

Although the results answered some questions about the influence of the
virtual image on the perception of the other, there are a number of unanswered
questions remaining. Future researchers should examine the influence of
changing the virtual image during the interaction. For example, what is the
influence of beginning with a highly anthropomorphic image and switching at the
end to a low-anthropomorphic virtual image at the end of the interaction? What if
the order were reverse and the interaction began with the low-anthropomorphic
virtual image? Which of the images would most influence the mental model of the

other?

109

Chapter 7

Conclusion

The results of this project support what may, on the surface, appear to be
a contradiction. People have responded socially to all mediated others displaying
any level of intelligence, whether they are agents or avatars, unless they are
represented by low-anthropomorphic images. Whether one is responded to as
human is not based on whether one is or is not perceived to be human, but
instead on whether one does or does not appear human-like. Even a rough
computer animation resembling a female face was anthropomorphic enough to
cause people to respond to their partner socially.

In fact, the perception of humans (avatars) and computers (agents) was
indistinguishable when they were represented by the same virtual image. This
means that cyberspace has not become a utopia where all voices are equal, nor
are all entities responded to equally. The results indicated that the appearance
of the virtual image is used in SVEs as the natural body is used in the natural
world. Hence, the visual representation was very influential on the mental model
of the other. However, the influence of the highly credible script may have also
played a role in this perception in that it may have influenced the perception that
the virtual confederate (agent or avatar) was intelligent, whether human or not.

Understanding when an agent can replace an avatar without negative

response or when the absence of an image is preferable will be increasingly

110

important as more and more of our business and personal interactions move
online. It is possible that the influence of the non-anthropomorphic image would
not be as significant or as influential in a more personal or less task-oriented
interaction. Further, it is possible that the distinction between agents and avatars
would be more influential in a more personal or social interaction. It is important
that researchers consider the context of their interaction. For example, many
researchers in social situations report that knowing the gender of your interaction
partner is the first question asked (Spender, 1996; Turkle, 1995; Waskul 8
Douglass, 1997). This might not matter to the same extent in a task-oriented or
business interaction. As designers and users of SVEs, it is important to
understand the implications of selecting a low-anthropomorphic image during a
task-based interaction. When credibility is an issue, a person will be better
served by a high-anthropomorphic representation in cyberspace than a low-
anthropomorphic image.

Providing no virtual image does not imply that people do not have an
image in their mental model of the other. The default image people create is
likely to be anthropomorphic. If this is true, this would explain why the results
showed that the perception of the virtual confederate (agent or avatar) with no
virtual image was similar to the virtual confederate with a more anthropomorphic
image and not at all like that of the virtual confederate represented by a low-
anthropomorphic image.

The virtual image influenced the perception of the virtual confederate more

than any other variable examined in this project. This is an important finding and

111

corresponds to previous research that visual cues will be processed at the
expense of other information in task-oriented interactions (Koda, 1996; Nowak 8
Anderson, 1999; Takeuchi 8 Naito, 1995). This has served to reinforce the
conclusion that cue lean media are best for task-oriented interactions or for
interactions requiring the transmission of complex or linear information. In other
words, more cues are not always better and in some cases can distract people
from important information presented elsewhere in the interface.

People in SVEs continued to use the visible image to perceive intelligent
others and the image influenced social judgment. Further, participants responded
socially in mediated interactions, but this does not mean that other factors were
not influencing their social judgment.) The results extended the understanding of
this phenomenon. The results of tests of the hypotheses added one qualification
to the conclusion that people respond socially to all mediated others; people did
not respond to intelligent others as social actors when they had a face that is not
anthropomorphic. The results support the drive to create a type of embodiment
that is clearly discernable from agents and other objects (as discussed by
Benford et al., 1995). Designers are working to create agents that blur the line
between humans and computers to encourage people to respond socially to
computers (Brent 8 Thompson, 1999; Dryer, 1999). An anthropomorphic face in

the interface would greatly support this effort.

112

 

TABLES

113

Table I

Types of Virtual Images and Indicators Available

 

No Visible Image

Visible Image

 

 

 

 

 

 

 

 

 

 

 

Text with No Text with 20 Virtual Image 30 Virtual Image Natural Image
Description of Description of
Image Image
Text Text Text Text Structure/use
structure/use of structure/ use structure/use of stmcture/use of of language
language, of language, language, language, or accents
spelling, typing spelling, spelling, typing spelling, typing
speed, literacy typing speed, speed, literacy speed, literacy
literacy
Context of Context of Context of Context of Context of
interaction interaction interaction interaction interaction
Content of Content of Content of Content of Content of
Interaction Interaction Interaction Interaction Interaction
Body Type- Body Type- Body
Avatar Avatar type/height
quality/selection/ quality/selection!
appropriateness appropriateness
Gender of Image Gender of Image Gender
/Biological
sex
Proxemics Proxemics Proxemics
Knowledge ofl Knowledge of Clothing
Familiarity with /Familiarity with Choice
Environment Environment
Gait information/ Gait
latency/ ability to information

 

 

 

navigate
environment

 

 

114

 

Table II

Summary table of Hypotheses

 

Effect of Agency on Social Judgment of lnteractant

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

image is visible.

 

Hypothesis I: People will feel more social attraction for avatars than Not
for agents. Supported
Hypothesis 11: People will feel more partner satisfaction with an avatar Not
for an interaction partner than with an agent. Supported
Hypothesis Ill: People will feel more copresent with an avatar than Not
with an agent. Supported
Hypothesis lV People will feel less uncertainty in judgments of avatars Not
than agents Supported
Hypothesis V Judgments of avatars will be more extreme than Not
judgments of agents Supported
Relationship of Presence and Co-presence
Hypothesis Vl: People that feel more physical presence in the Supported
environment will feel more copresence with
their interaction partner.
Effect of Embodiment on Social Judgment of lnteractant
Hypothesis Vll: When a virtual image is visible, people will feel more Not
copresence with their partner than when no virtual Supported
j image is visible
I Hypothesis VIII: When a virtual image is visible, people will feel more Not
social attraction for their partner than when no Supported
virtual image is visible.
Hypothesis IX: When a virtual image is visible, people will feel partner Not
satisfaction than when no virtual image is visible Supported
Hypothesis X: When a virtual image is visible, people will feel less Not
uncertainty in judgments of others than when no Supported
virtual image is visible.
Hypothesis XI: When a virtual image is visible, people will make more Not
extreme judgments of others than when no virtual Supported

 

 

Effect of More Anthropomorphic Embodiment on Social Judgment of

 

 

 

 

 

 

represented by a highly anthropomorphic image
than those with low-anthropomorphic images.

 

lnteractant

Hypothesis Xll: People will perceive intelligent others with highly Supported
anthropomorphic images as more socially
attractive than those with low-anthropomorphic
images.

Hypothesis XIII: People will be more satisfied with partners represented Supported
by highly anthropomorphic virtual images than

_\ those with low-anthropomorphic images.
HYimthesis XIV People will feel more copresent with a partner Supported

 

115

 

Table III

Design of Experiment and Number of Participants in Each

 

 

 

 

 

Condition
Agent Avatar

Anthropomorphic 23 24
(avatar 1)

Not

Anthropomorphic 22 22
(Avatar 2)

NO avatar 23 20
TOTAL 68 66

 

 

 

 

 

116

Table IV

Six Items retained in Social Attraction Scale. These indicators were measured on
a 7-point metric. 1 strongly agree to 7 strongly disagree.

 

I think he (she) could be a friend of mine
I would like to have a friendly chat with him (her)

 

 

We could never establish a personal friendship with each other

 

She (he) just wouldn't fit into my circle of friends

 

She (he) would be pleasant to be with

 

 

 

I don't care if I ever get to interact with him (her) again

 

117

Table V

Ten Items retained in partner satisfaction scale. These indicators were measured
on a 5-point metric. 1 strongly agree to 5 strongly disagree.

 

My interaction partner is NOT of very high intelligence,

 

My interaction partner is a reliable source of information on the topic

 

My interaction partner lacks information on the subject

 

I would consider my interaction partner to be an expert on lntemet searches

 

I believe my interaction partner is quite intelligent

 

This interaction partner is an unreliable source of information on Internet
Searches

 

This speaker has had very little experience with lntemet Searches

 

This speaker has considerable knowledge of the factors involved in lntemet
searches

 

This speaker has very little knowledge of the factors involved in Internet
searches

 

 

 

My partner has had substantial experience with Internet Searches

 

Six items retained in the semantic differential partner satisfaction scale. These
indicators were measured on a 7-point metric, participants were asked whether
they considered their partner to the far left (1 ), or the far right (7) on each of the
followi_ng pairs of words.

 

Reliable/Unreliable

 

 

lnformed/Uninformed
Qualified/Unqualified
Intelligent/Unintelligent
ValuableNVorthless
ExperUlnexpert

 

 

 

118

 

Table VI

Thirteen Items retained in perceived copresence of the virtual confederate scale.
These indicators were measured on a 5-point metric. 1 strongly agree to 5
strongly disagree.

 

My interaction partner was willing to listen to me

 

My interaction partner was intensely involved in our interaction

 

My interaction partner seemed to find our interaction stimulating

 

My interaction partner communicated coldness rather than
warmth

 

My interaction partner created a sense of distance between us

 

My interaction partner seemed detached during our interaction

 

My interaction partner was unwilling to share personal
information/feelings with me

 

My interaction partner made our conversation seem intimate

 

My interaction partner created a sense of distance between us

 

My interaction partner created a sense of closeness between us

 

My interaction partner acted bored by our conversation

 

My interaction partner was interested in talking to me

 

 

 

My interaction partner showed enthusiasm while talking to me

 

119

 

Table VII

Five Items retained in self-reported copresence scale. These indicators were
measured on a 5-point metric. 1 strongly agree to 5 strongly disagree.

 

I did not want a deeper relationship with my interaction partner

 

I wanted to maintain a sense of distance between us

 

l was unwilling to share personal information/feelings with my
interaction partner

 

I wanted to make the conversation more intimate

 

 

 

I was interested in talking to my interaction partner

 

120

Table VIII

Seven Items retained in Uncertainty scale. These indicators were measured on a
5-point metric. 1 unable to answer to 5 Completely Confident.

 

Following your interaction, how confident are you of your general
ability to predict how he/she will behave?

 

Following your interaction. How certain are you that he/she likes
you?

 

Following your interaction. How accurate are you at predicting the
values he/she holds

 

Following your interaction. How accurate are you at predicting
his/her attitudes?

 

Following your interaction. How well can you predict his/her
feelings and emotions?

 

Following your interaction. How much can you empathize with
(share) the way he/she feels about himself/herself?

 

 

 

Following your interaction. How well do you know him/her?

 

121

 

Table IX

Five Items retained in Presence scale. These indicators were measured on a 7-
point metric. 1 Not at All to 7 Very Much.

 

How involving was the experience?

 

How intense was the experience?

 

To what extent did you feel like you were inside the environment
you saw/heard?

 

To what extent did you feel immersed in the environment you
saw/heard?

 

To what extent did you feel surrounded by the environment you
saw/heard?

 

 

 

122

FIGURES

123

 

FIGURE I. THIS MODEL REPRESENTS THE THEORIZED INFLUENCE OF
LEVELS OF ANTHROPOMORPHISM ON THE MENTAL MODEL OF ANOTHER
AND THE RESPONDING SOCIAL JUDGMENT OF A MEDIATED INTERACTANT.

 

Categories ‘15: “r “' ‘1‘
Living/inanimate Jyggm?t ow lfid
lactan

 

 

Human/nonhuman
i ’4

 

V
Mental
Model of
lnteractant
8
Environment

 

 

 

 

 

 

 

Physical Presence
In the Environment

 

 

 

 

 

|:IMeasured O Manipulated. """""""""" ETheorized not manipulated.

 

124

 

FIGURE 2. IN THE NO IMAGE CONDITION. WHERE THE VIRTUAL

CONFEDERA TE WAS NOT REPRESENTED BY AN IMAGE, PARTICIPANTS VIEWED
THIS SCENE OF THE VIRTUAL ENVIRONMENT.

r, .

Treasure unt
arrlin Uriel.

 

125

FIGURE 3. IN EITHER THE ANTHROPOMORPHIC OR LOW-ANTHROPOMORPHIC
IMAGE CONDITIONS. PARTICIPANTS WERE ALLOWED TO SELECT ONE OF

THESE CHARACTERS ON THE VIRTUAL REPRESENTA TION SELECTION SCREEN
TO REPRESENT THEM.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

126

FIGURE 4. IN THE HIGH ANTHROPOMORPHIC IMAGE C ONDI TION.
PARTICIPANTS VIEWED THIS SCENE OF THE VIRTUAL ENVIRONMENT AND
THIS WAS THE HIGH ANTHROPOMORPHIC IMAGE.

Treasure un‘r
51‘4th Lind.

 

127

FIGURE 5. IN THE LOW-ANTHROPOMORPHIC IMAGE CONDITION.
PARTICIPANTS VIEWED THIS SCENE OF THE VIRTUAL ENVIRONMENTAND
THIS WAS THE LOW-ANTHROPOMORPHIC IMAGE.

 

128

FIGURE 6. MEANS OF PERCEPTION OF VIRTUAL CONFEDERATE ON EACH
DEPENDENT VARIABLE: AGENT CONDITION VS. AVATAR CONDITION

Agent vs. Avatar Perception

 

 

El agent
I avatar

 

 

 

Copres1 CopresZ SocAtt PartSat1 Partsat2 certainty

Measurement Instruments
Relative to number of Indicators in Scale, but not relative to
scale metric.

 

Measurement Instruments are measures of the dependent variables.
Copres1= perceived copresence of the virtual confederate (1-5 metric).
Copre52=Self-reported copresence with the virtual Confederate (1-5
metric).

SocAtt= Social Attraction (1-7 metric).

PartSat1=Partner Satisfaction measure (1-5 metric)

PartSat2= Partner Satisfaction Semantic Differential (1-7 metric).
Uncertainty (1-5 metric)

 

 

 

129

FIGURE 7: MEANS OF PERCEPTION OF VIRTUAL CONFEDERATE ON EACH
DEPENDENT VARIABLE: NO IMAGE CONDITION VS. ANTHROPOMORPHIC
IMAGE CONDITION VS. LOW-ANTHROPOMORPHIC CONDITION

Influence of Image on Perception

 

 

El No Image
I High Anthropomorphic
El Low-anthropomorphic

 

 

 

Copres1 CopresZ SocAtt PartSat1 Partsat2 uncertainty

Relative Measurement Instrument Means:
Adjusted to for number of lindicators, but not relative to
scale metric

 

Measurement Instruments are measures of the dependent variables.
Copres1= perceived copresence of the virtual confederate (1-5 metric).
CopresZ=Self-reported copresence with the virtual Confederate (1-5
metric).

SocAtt= Social Attraction (1-7 metric).

PartSat1=Partner Satisfaction measure (1-5 metric)

PartSat2= Partner Satisfaction Semantic Differential (1-7 metric).
Uncertainty (1-5 metric)

 

 

 

130

FIGURE 8. REVISED MODEL: INFLUENCE OF AN THROPOMORPHISM ON

MENTAL MODELS OFAGENTS AND A VA TARS IN SVES

This model represents the influence of levels of anthropomorphism on the mental
model of another and the responding social judgment of a mediated interactant
as indicated by the data reported in this project.

Levels of

Anthropomorphism

Body Morphology

High-Anthropomorphic

 

Body Morphology

Low-Anthropomorphic

 

 

 

Categories
Living/inanimate
Intelligent/Not

 

 

V
Mental Model
of lnteractant

&
Environment

4

9

Judgment of
Mediated
lnteractant

 

 

Partner
Satisfaction

 

 

 

Social Attraction

 

 

 

 

Copresence

 

 

 

 

 

Perceived to be
Agent or Avatar

 

Physical Presence
In the Environment

 

 

 

131

 

 

APPENDICES

132

APPENDIX A

AUDIO SCRIPT

133

Hi this is Kim.

Ok, my name is Kim and I use my lntemet connection almost constantly. I have
some experience with searching for a product and finding the lowest price online.
I worked for a company where I made some online purchases of hardware and
software as well as electronics. I also know how to find good deals, coupons and
rebates for online shops. I like search engines that give me a lot of room to
maneuver»- Alta vista, Lycos. I really like web rings because you can get really
specific, you don't have to go around looking at a lot of things you don't need. I
don t like Go network or hotbot or other search engines that are more limited.
Anyway, it usually takes me a while to find things at times, but I have to say that l
have a pretty decent vocabulary so I can usually find what I am looking for.

 

 

Great, thanks. I hope we can do the Scavenger Hunt together.

134

APPENDIX B

PRE-EXERIMENT QUESTIONNAIRE

135

 

rescue
.EEcE 6:23 553:. 2.. 3...... :3: sex 3:: wco. >5:

3:... BE: :0 w

3:... Wm 2:98:53
8:... «1m .923. .:::.> new:
uE: _ no» 3:: 3:... E9: Bo:
.932
D D D D B

3:32 :25 3E..oEcm EoEom .632

...mEoo: .9... u:..:: now: :26 so» 3::

D n. D D D D D 9.: c. 9.2.: @580
U D D n. D D D 5:25.... wins:
D U _U D D D U 8.32.33835 .32.: w:.:..:»..
U .U .U D U D D E95: o. :o..:E._o..:. MEUSEA.
n. _U D D D D D 23.53 w:...:. ..w.:.

3.33:» MEEO

U D D D U U U 5.55250 \3 :25. :. wcaoov.

n. D D n. U D n. 3.5: .3 5.5. c. Page.

0 U D D D D D .65.... .3 :25. :. wcioov.

U D D .U U U D w:.:.:... 3:35:96

0 D D D D n. D 5.3555235.

U D .U D D D _U wcfiaocm

D D D D D D D 23.... £25 MES: ..wd.

8.8. .28.}.

D D U D D U D ..53 .ccfim

D D D D D D D 3:2:33: :2.
.293 3.3.3
bu> __< a 52

inow wEBo:o.. 2...: :2... 5.. .23...

£32.53 cc... 3:» .323 o. 33.6 2.. 9.86:. 3:97.

B D D n. D .U D D
n o m 9. m m . c
.c.. < :35. 52
55.33 328:3.
A23“... _::.._.> 39. .29.: >55. :9. o... :92: >5:

 

U n. n. n. D n. n. 3:25: 2:5
n. n. D _u n. n. n. 8.5.2:: acreage:
D n. U n. n. n. n. are: .855:

n. n. D n. a U 0 3.35.3233 :2
D D n. D U D 05.30.5883 .283:

D _U .U n. n. D D 33.5: o:..-:O
D D D n. D _U D 8:80.». :o..:E._o..:.
D D D n. D D D moE:O
D D D D D D B 33:3:23
B D D D D .U 0 3.3389903
» . o :2 a B
:_ =5» 5......

©3382. w:..5c..c..

o... 5.. 8.39:3 .33» um: 39» on >35. .0 3.53.5... .30...

3.35.: 3:39..
...ocsom.....53 .: 5.39:3 : 3:: 9:2. so» .2 £225.: :9.
cc: :39. ..c: 325:: 3:97.. a...oo:om\..53 .: 9:23:50

23 as» o: .8: : on... :02: 3o: .ow:..o>: 2.. :0

8.35:. 3.32.
A 2:9. .: 5.39:3 : 3:: ..:c.u no» ... £235.:
:9. US: :52. ..o: .533: 8:2... :25... .: 3:22:55

3: so» 0: .3: : up... :92: 3c: .owfig: a... :O

95:92). w:.::3::< D .535 .35 .03.. D

9:27. :90 U EOmdU D 9:382 .8... D
..owE. D :28... DO U on:o.:.z D

>P 93:0 0 5.5 o....u.:m .U SE:oE:U D
mm; D 06:”. D >b D

5%.: :2. .E. 3.8: ..: :. .323 3:9...
...oEo: .: as» o. u.::..:>: 2: 2:2. w:.3o..c.. 2....0 5...?

 

DDDDUDDDDDD

o._oE:oc.ownoanm.o
2.56 5:: 2... :. macaw co...»

.5 3:8: .92: so» 2: 3:: ES: 32. x_o.:E.on:<

{ii}i}§§§{il{iii}*iifillfiillllfiii§lii§*§iii}

 

5:5 .o

::u.._oE< 3.32 .m

3.5.3. 56:92.0 .w

25%.: .m

::..»< .m

::u..uE< ::u.....< ..
3:39.86. arise»

A25... >355... .mcE as» c: n:..c..w..o:: 3:...» .25 5.?

2:34:35 .m
5.5m .: 5.5... .m
PEELQOm .m 53:53:“. .
3:: 9.8.... 3:65.; 3:3 So» m. .23
...Eon so» 953 EPA .33
0.25:. .m 2:5. ..
$35». :5» m. :33
iiii¥i¥¢il#CliiiCtillifitiitfffiiifiii¥iiiiiii
.sumeeuhgeoo 53.5: 26:552....
m5. 0:9: 3:: :3 5926:: 83:6 0:: 3.232... 3...
so» 22. o... $32.0. .::. 86:552.... 2.. w:......:Eou
.3 .5... so» .23 >55. :3 .u. 83.: as 3 Eureka...
be: m. 3.3.05.8. :62: .39.: as??? $3..

 

__ 2... oz... _—

 

Q. .::a.o.::n.

J

3.39.. 9.: 39.. .2: .8:9£.:9&.m 2.35.. 9.39:
8993.3 3.23: .39.. 3\n«§§

 

.9:.C

9E:w : w:.z:.a ._ .

3.9.0:: 95.5. .2

8:3 :93 .8582. f: :o w:....oB .o

..2..ww39:

.c .9... w::.:E\t9m.m: : Es... 9w:mm9E : w:.::8
95339:

.9 .m: w::.::._...>.9.r..m: : o. 9w:mm9E : 3:3:

.9: 9:. wcwtam

:58 .::9 : w::.:.>

:99}: .9. “99.53 . :o::E.o..:. .0. w:.:8:9m
_so:9m...:c.. ..E .9. @939: . :o:::_.o..:. .c. w:.:8:9m
..:E-9 w:.::9m

.589 b: wc...99:9

.>-.A...< ...<Ib ....< 9.8.0 ...9Eo: .:

.9525 9:. .99.... 2:. 9E: 3:. 9:. 9.9 as» n... mam .::>>

06

.aaewés

 

 

.950 .m.
25% : 33:... .. _
3.9.0:: 95.5. .o.
9?... :93 .::om.2. .3. :o w:....o3 .0
9.9333:
.9 .9... w::.::.\>.9m.m.. : Es... 9w:mm9E : w::.:9. .w
:28w39:
.c .9... w:...:E\>.9m.¢.: : o. 9339:. : 3:8:
.9: 9:. MES:
Eco. .::9 : w:...:.>
..9me .o.. 3.5.3 _ :c::E.o..:. .o.. w:.:8:9m
30:93.9... .9: .o. 3:99: . :o..:E.c..:. .c.. w:.:9.:9m
..:E-9 w:.::9m
..:E-9 b: w:.:99:9
.>..ZO :20 9.9.... ...cho 893 :9. w:.:. :.:E
9:. .33 .::3 9:5: .: .9:.9.:. 9:. :92. :9.» 98.. 5:. 9: H

Arid-Q3666

 

.950 .5
...9me .. .9889 :5. E938: 9:. So 83...... .o
.l ...9w< .32. 9.8.9. 9.:E9... .o 2:2.
98 29: 9.2:; 89.8.: ..m:...:.. .m
.I ...9w< :95 9.8.9. ...9_:E9... .9 2:2.
.2: 22. o. 98.... a .3239 .v
.Il ...9w< ..9:o 9.9..9. ...9.:E9... .o 9.22.
.9E ...9: o. .9::.9E :.c:9v.:c:\»..E:.. : :2 .m
.9.:. :.:w: b. :5. 5.2.5.8 9:.».0 E3. .m
5:3: Em: :.:w: .0. :5: .82.:59 9:..co E: ._

.>..ZC :20 9.8.0 ...o: 3.2.9: :9. o:
.::3 8959.59 9E0: So» was: 3.930.: 9>:: so» :9: 3

£35.: .882 .m 3293.98“. ..V 35.95% .m
:28 .o: .3: i:::o.m:990 .m .9>9: .m:E.< ._
29.3.59 8:9:
.30» w:.m= 9.5938: 9.2:: as» o: 29...: >5: ._:L9:9w :.

 

5: .3
E>.:¢:N.§>sz
0000000
3:885 3:888
32.93:”. :2.
0000000
8.58.: 8.3
5252......” .2398
0000000
.335 §3
banana: v:~.:u_abm
0000000
8.53... 8.2
:88... 9 a q a N .58..»
0000000
coauu::_ oa9u<
58.3... 9 a v a N .583.
0000000
8.53.. 8.3
:2§.9:.:~.:8§m
0000000
833 82
buaobmhcmv: azfigbm
DDDDDDD

29.2.63 .39..
:93 :9» 98.. 5:.
9:. on 9. 3.53
so» .::3 o: o.

.. .35 ESE... 30..

29.2.E09

.30» U99...

:0» 9:... 3:. 9:. a:
0. 38:3 .5» .::3
wEo: :. so» 893
3.39993: 32..

3.95....
m. .25.: 9:. 3...:

..9E9.:. 9:. 9.3

. 5:3 333:: .9w .
.9E9.:.

9:. w:.:: .32.:
:92: 335. ..E... .

E...
m. .9E9.:. 9:. 9...»:

2:: b9>9 $25:
.2525 9:. 9m: .

..9:.9.:. 9:. ”.5....
.: 3.2:: .09.» E: .

.99:9..9mx9-:::

 

m._...m .9E9.:. Ham» w:.:.:w9. $5.99.. So» 9.36:. 99:9...

 

.

:25 .o

:9.::U .m 980.. m.9>..:.9~. .n
9E9..m.::9..h. .c .250 >..::EEoU .m
:2.-. ..V .85: .m 9:5: .m .53 ._

...9.::E¢9 : :92. .86: 3.3:»... aim.
2.9.: 8:. :: 9.8. U ...9E9.:_ 9:. :92. so» 9>:: 89: 3

00000

As...

.09.;

...9E9.:. 9:. 9:: so» 3: .9w

an.

.9)9:

~59)? C
. . O .:
9...3 .33:

I37

APPENDIX C

POST EXPERIMENT QUESTIONNAIRE

138

l0

l2

SuveySaldbyMarltetiggMasters Page I

Instructions:
0 -selectONlYONEChoice D -se|ectAl.l.thatapply.

lnthefllstpartofthisouettionnaire, pieasetlinltofthe theyouinteraction haveiust
experienced. Take time to read the instructionsand andanswer the following questions.
Instructions: Using the following scale, for each question please indicate how confident
youareinyouabilitiestoanswer,“ predct, thecorrectretponsetothequettionbased
onthetomnuterhteractionyouimtwhad.

Heasehtdcatethedegreetowflchyouagreewiththeiolowingstatement: My
mteracdonpm-tnerwaswllngtolistentome

StronglyApee IO 20 30 4O 50 StronglyDisaaee
Heasemdcatedtedegreetowflchyouagreewiththefolowhgstatement:fly
interactionnartnerwashtenselyinvolvedmom mterattion

StronglyAgree IO 20 30 40 SO StronglyDisuee

Heasemtlcatethedegreetowllch agreewlththefolowhlgstatement:
interactionpartnertldnotwantadzzp’Ierrelationslfi My

StronglyAu-ee IO 20 30 40 SO Stratum

Heasehdatedtedeu'eetowflchyouagleewiththefolowhgstatemmmm
hateractionpm'tnerwasnot

StronglyAgreelO 20 30 4O 50 MM

Heasebtlcatethedegreetowlichyouagreewiththefolowingstatement: My
interactionpm'tnerseemedtolhdotn’ hteractionstimtlathg

StronglyAgree IO 20 30 4O ‘50 MM

Pleatebdcatethedegreetowflchyouaaeewiththefolowhgstatemencfly
interactionwtnercomnnnicatedcolbetsratherthawm'mth

StrongiyAgreeIO 20 30 40 SO StronglyDisaaee

Heasemtltatedledegreetowlichyouagreewidtdlefolowingstatement: My
interactionpartnercreatedasenseohlstance betweenus

StronglyAgree IO 20 30 40 SO StronglyDisauee
Pleasemdatethedegreetowllchyouagreewithdlefolowhgstatemelmfly
interaction seemeddetacheddu'ingotn'interaction

partner
StrongiyAgI-ee IO 20 30 4O 50 StronglyDisagree
Pleasehulcatethedegreetowlichyouagreewiththe folowing statement”!
hteractionpartnerwaslllwiflngtoshare sharepersonalhiormationlfeeiingswithm’e

StronglyAgreelO 20 30 4O 50 StronglyDisagree

Pleasehdcatethedegreetowllchyouagreewithdtefolowingstatement: My
interactionparmermadeourconversationseembtimate

StronglyAgree i0 20 30 40 SO StronglyDisagree

Pleaseintlcatedledegreetowilchyouureewidldiefolowingstatement: My
hteractionpartnercreatedasenseoftlstance betweenus

SaonglyAgree IO 20 30 4O 50 Stronginisagree

139

 

 

 

13

I4

IS

l6

1.7

I9

20

2|

22

23

24

25

Survey Said by MarketigLMasters Page 2

Heaseintlcatethedeaeetowlichyonagreewiththefolowingstatemmt:ny
interactionpartnercreatedasenseofdosenessbetweenus

StronglyAgree lo 20 30 4O 50 Mbisagree

Heasehtlcatedledegnetowflchyouagreewithdleiolowhgstatemencfly
hteractionpartneractedboredbyom'conversation

StronglyAgree IO 20 30 40 SO Mbisagree

Hemehdcatethedegreetowlidtyouagree‘owithdmiolowingstatement:fly
hteractionparmerwasinterettedintdithgto

StronglyAu'ee lo 20 30 40 SO StronglyDisagree

Pleasebtlcatethedegreetowlichyou withthefolowingstatementzfly
interactionpartnershowed showedtiltingentlmsiasmWEe

StronglyAu'ee IO 20 30 40 50 MM

Pleasehdatethedeueetowlichyonaueewiththefolowhustatementflwaswm
tolstentomyinteractionpartner

StronglyAgree IO 20 30 40 SO StroquDisagree

Heasehtlcatedmdegnetowliduyouagreewithdnfoflowhgstatement:lwas
detachedthn'mgthe conversation

StronglyAgree IO 20 30 40 SO Militant

Heaseindcatethedeueetowfldnyonaueewkhthefolowhvgstatmlwas
htenselyinvolvedmtlishteraction

StronglyAgree IO 20 3O 4O 50 StronglyDisuree

Heasehclcatethedegreetowhich agreewlththetolowhngstatementzlddnot
wmnadeeperrelationswwithmymracdonparmer

StronglyAm'ee IO 20 30 4O 50 Mbisagree

Phasehclatedmdeueetowflchyouaueewidltheiolowhgstatemmndfumddm
interactionstimtlathg

StrongiyAgreelO 20 30 40 SO SMDisagree

Heaseindaudndeaeemwlichyouagreewiththefolowhgnatementdwantedto
mittainasenseohlstmtebetweenus

StronglyAgree lo 20 30 4O 50 StronglyDisagree

Pleaseindicatethedeu‘eetowflch agreewiththefollowhngstatemerml
Imwflngtosharepersonalfl Hell-coring! ieelingswithmyhlteractionpmtner was

StrongiyAgree IO 20 30 40 SOStronglyDisagree

Phaehukatedndeaeetowflchyouaueewiththeioflowhgstatemenclmedm
maltetheconversationlnoreintimate

StronglyAgree Io 20 30 40 50 Mbisagree

Heasehdiatethedeueetowlichyouaueewkhdnfolowhgstmementdwmnedto
maltetheinteractionseemcaud

StronglyAgree lo 20 30 4O 50 Stronglkaagree

140

26

27

28

29

30

31

32

33

34

35

36

37

38

39

Survey Said by MarketinLHasters Page 3

thebtlcatedledeueetowfldlyonaoeewiththeiolowhgstatementflaiedm'
createasenseofclosenessbetweenus

woulyAa-ee IO 20 30 4o 50 mum

Heasehclcatedledegleetowllchyouagreewithdieioliowingstatement: lwas
hterettedhtallthgtomyinteractionparmer.

Strongly/Agree IO 20 3O 4O 50 StronglyDisagree
i-lowhwolvingwastheexperience?
NotAtAll IO 20 3O 40 SO 60 7O VeryMuch
l-lowintensewastheexperience?
NotAtAil lo 20 3O 40 SO 60 7O VeryMuch
Towhatextenttldyoufeellheyouwelehsidedtemwhomentyonsawlhem'd?
NotAtAil to 20 so to so so '10 VeryMuch
Towhatextenttldyoufeelimmersedhtheenvkonmentyousawlhemd?
NotAtAll IO 20 30 40 SO 60 7O Veryl'luch
Towhatextenttldyoufeelsm'romtdedbytheenvh'onmentyousawlhem'd?
NotAtAll IO 20 3O 40 SO 60 70 Veryl‘luch
llowoftenddyouwmttotonchsometlmlgyonsawlhem'd?
NotAtAli lo 20 30 40 SO 60 7O VeryMuch
Howoftendidyoutrytotonchsometlmlgyousawlhem’d.’
NotAtAll IO 20 30 40 SO 60 7O VeryMuch
l-lowoftendidyouwanttosmelsometlingyousawlheard?
'ilotAtAil IO 20 30 40 SO 60 7O VeryMuch

"enchantethedeueewwflchyouagreewkhthefolowingstatementdtflnkhe
(she)touldbeafriendofnine '

MM" StronglyDisagree
IO 20 3O 40 50 6O 7O

lesehdatedndeueetowlichyouagreewhhthefolowhlgstatemenclwmddlke
tohnveafriendlydiatwithlimmer).

StronglyAgree Strodybisagree
IO 20 3O 40 SO 60 7O

PhasehdicatethedeueemwflchyouaaeewiththefolowthWecodd
neverestablishapersonalfriendslipwitheachother

MAC"! StronglyDisau'ee
IO 20 30 40 SO 60 7O
Heaeindcatethe towhich agreewlththefollowlnnghe he t
wodb'tfltintomy olfriendsy.” ( )ius
StronglyAgree StronglyDisagree

IO 20 3O 40 50 6O 70

141

4|

42

43

45

46

47

48

49

50

SI

52

SmeySaidbyMarketthasters Page4

Heaehdatedmdegreemwflchyouagreewidndnfolowingstatemen:$he(he)
woddbepleasanttobewith

flmuane Smnmbuuu
no 20 3o 40 so 60 7o

Pleaebdmedndegreetowlichyouaaeewithdnfolowlngstuementdiedlknow
he'(lnm)pe'sonally

StronglyAgree StronnglyDisw-ee
IO 20 3O 4O 50 6O 7O

Heasemdatethedeueetowlldnyonagreewiththefolowhmstatementzfle(dm)k
pe'sondyoflensivetome

StronglyAgree Strondybisqee
IO 20 30 40 SO 60 7O

Heasehdcatethedeaeetowflchyouureewkhthefolowi'smtementddon'tcme
iilevergettointeractwithlmn(he’)ag&n.

“Nahum: , finnbflnmn
no 20 30 so So 60 7o

Phasemdatethedegreetowflchyouameewidnthelolowhstntemmclwkhlwee
morellteflm(her).

5mm MM
'0 20 30 4O 50 6O 7O

hteacdomhowoodldentm'eyouolyom'gene'alabltytopnenlcthow
helslne behave?

UnabletoAnswer IO 20 3O 40 SO CompletelyConlident
Followhyom‘interaction, Howtertinareyouthathelshelltesyou?
IlntletoAnswer IO 20 3O 40 SO ConmletelyConlident
Folowhyominteacdon,flowatanateaeyouatprenlctkdmvalneshelsheholds?
llnabletoAnswer IO 20 30 4O 50 ConmletelyConlldent
Foflowhmmhteacdomflowacanatem'eyouatprenlcthgflslheatdtudet?
UnabletoAnswer i0 20 30 40 SO ConnpletelyCodident
FMyomhteacdon,Howwelanyonpredalislhefeelngsandemodons?
UnabletoAnwer no 20 3O 40 SO ConletelyCodident

Folowmgyom mteractionJlowmuchcanyouenmatflzewithohmehhewayhelshe
ieelsabontflmsellllnesell?

llndnletoAnswer IO 20 30 4O 50 WW
Folowhngyom'innteractionJ-Iowweldoyouknowflmlher?
llnabletoAnswer IO 20 3O 4O 50 CompletelyConfident

Heasehndcuedmdegreetowflchyouagneewlthdwfouowingsutemencldinkthe
vin'tnalbodywasveryattractive

StronglyAuee StronglyDisagree
IO 20 30 4O 50 6O 10

142

$3

54

55

$6

$7

58

59

6|

62

63

64

Survey Said by MarketinLMasters Page 5

Pleasehnnlcatethedegreetowflchyouareewiththelolowingstatement: ltlinltthe
vh'tualbodywasve'ysexyioolthng

StronglyAgree mm
IO 20 3O 4O 50 6O 7O

Pleasehnlcatethedegreetowlichyouaueewiththelolowhustatement: llikethe

lookolthevirtualbody

amubNmn Smmmnuuu
IO 20 3O 4O 50 6O 7O

Pleasehnclcatethedeu'eetowlichyouagreewithdneiolowhngstatement: itlinltthe
virtudbodyissomewhatngiy

flnmhAuu 7flmuhflhflfl
no to 30 so so to

Pleasehdtatethedegreetowhichyonagreewiththefolowingstatement: ltllmtthe
vh'tualbodywasverybecodng

Smmmfimu Smmmomme
no to 30 so so to 70

Phaehnnlatethedeueetowflchyouureewhhtheiolowhgsmtemencldnnhdne
virtualbodywasnotvenygoodloohing

Whine StronglyDisagree
IO 10 30 40 50 6O 7O

Heasehnlatethedeueetowflchyonameewithdmiolowhgstatementzlhevm
bodywasrepdsivetome

StronglyAgn-ee A Stronglybisagree
IO 20 30 4O 50 60 7o

Pleasehnnlcatethen‘l‘efiee agreewiththelolowhstatement:
hnteractionpm-tne'is Tofveymhntm My

StrongiyAgreeIO 20 3O 4O 50 StronglyDisagree

l’leaseinnnlcatethedeu‘eetowldchyonn"t withthelolowhgstatement:Hy
interactionpartnerkareliablesonn'ceol ormationontlne tinetopic

StronglyAgreelO 20 3O 4O 50 StronglyDisagree

Pleasehulcatethedegreetowflchyonmeewidnthelolowhgstatement:fly
innteractionpm'tnerlacltsinformationonthesnlniect.

StronglyAgreelO' 20 30 4O 50 MM

Pleasehndcatethedegreetowllchyouagreewlthdneiolowhngstatement: lwonid
considermyhnteractionpartnertobemnexpertonlnternetsem'dnet

Somme: no 20 30 to so StronglyDisagree

Pleaseintlcatethedegreetowlidnyonagreewidnthefolowhngstatement: Ibelievemy
interactionpm'tnerisqnitehntellgent

StronglyAgree lo 20 3O 40 SO StronglyDisagree

Pleasehndatethedegreetowlichyouagreewiththeioflowhngstatement: This
innteractionpartneris annnnreliablesonnceofinformationoninternetSean-chet

StronglyAgreeIO 20 30 4O 50 StronglyDisagl’te

143

 

 

 

 

 

65

66

67

68

69

70

7!

72

73

74

75

76

77

78

79

Snnrvey Said by Marketing Masters Page 6

 

Pleasehntlcatethedegree towinchyou ureewiththefollowingstatementfl’hisspealter
hashadverylittleexperiencewithlnternetSeardnes

StronglyAgree IO 20 3O 4O 50 StronglyDisagree

Pleaseindicatednedegreetowiichyouagreewithdnefoflowingstatement: Tlisspeaker
hasconsideableltnowledgeofthefactorsinnvolvedinlntenetseardnes

StronglyAgree IO 20 30 40 SO StronglyDisagree

Pleaseinntltatethedegreetowlichyou agreewiththefollowhgstatementfl'lisspealte'
hasve'ylittieltnowledgeofthefactorsinvolvedinnlntenet searches

StrondyAgree IO 20 3O 4O 50 StronglyDisagree

Pleaseintlcatethe winch withthef statement:
hashad degreeto withyouagn'ee olowhng Nypartner

StronglyAgn-ee IO 20 30 4O 50 StronglyDisagree
Considethehnte'actionyoniustheard.Woddyouconsideryompmtner
Reliable IO 20 3O 4O 50 6O 7O Unreliable
Considetheinte’actionyouiustheard.Vlonldyonconsideyonnpttne
idormed IO 20 3O 4O 50 6O 70 Unmnformed
Considerthehnteractionyouhnstheard.Wonldyouconsideyonlpmtne
Qualified IO 20 30 4O 50 60 7O Unquflfied
Considethehneacdonyouhsthethoddyoutonsideyompartner
Intelhgent no 20 30 40 So 60 7o Ilnninteligent
Considethehnte'attionyouiustheard.Wonldyoutonsideryompetne
Valuable IO 20 3O 4O 50 6O 7O Worthless
Considertheinteractionyouhnsthem’d.Wotddyouconsideryom'pm'tner
Expert 10 20 3O 4O 50 6O 7O lnexpert
Wouldyousaythepersonnyoutlscussedyonn'rmnldngswithwas
lnnnersonal IO 20 Personal
Woddyousaythepersonyoutlscussedyotn'rankingswithwas
insensitive IO 20 Sensitive
Wouldyousaythepersonyoudiscussedyonn'ranltingswithwas
Cold IO 20 Warm
Wouldyousaythepersonyoutlsanssedyonn'rankingswithwas
linsociable IO 20 Soddnle

Towhatextentdidyoufeelyougotagoodenoughideaofhowpeopleattheotherend
arereactinng.

Verygoodidea IO 20 Notgoodatall

144

80

82

83

84

85

86

87

89

90

91

Survey Said by Marketing Maste's Page 7

Towhatextenttldyouieelyougota"feel'forthepersonattlneotherend?
abletogeta'feel" IO 20 notabletogeta'feel"

Towhatextenttldyoufeelyouwereabletofornnanimpressionofpersonalcontactwith
yonn'pm'tner?

abkmformaninmressionofpesonailmefléoformmnimpresionoipesondcomaa
O O
Towhatextentdidyoufedyouweeabletoassesyonnparmesreacdomtowhatyou

fletoassessreactiom IO 20 notabletoassessreactions
To what extent was this like a late-to-face meethng?

Alotlkefacetoface IO 20 notlikelatetofateatal
‘l’owhatextentwastlislikeyouwe'ehnthesameroomwithyonn'partne?
Alotlkebeinnginthesameroom IO 20 notlkebehnginthesanneroomatalil
Towhatextentdidyonnpm'tne'seem'real"?

veryred IO 20 notredatall

Howltdyishdnatyouwonlddmosetousetflssystemofmteacdonforameethgin
wlnichyonwanntedtopersuadeothe'sofsometilng?

verylkely IO 20 notlkelyatall

Towhatextenttldyoufeelyouconldgetoknowsomeonethuyoumetoniydwmngh
thissystem?

verywell IO 20 notatall
Pleasemnlcatewhatsexyouthouhtyonn’pm'tnerwas:
Ollnnlete'mhned OMale Ofemale

Youhavenowflnishedtheexpeimenalpauofyonnrtddpadon.Thefolowhngfew
questionsaredesi togiveusannideaofhowyou eltaboutthisexpe'ienteJ’leasebe
ashonnestandstr ghtforwardasyoucan.Yonn'commentswillbeusedtohelpusdesign
betteexpe'innentsinthefutnn'e.‘l'hankyoui

Considethevhtualbodyyouchoosetorepresentyou.Whytldyoudmosedisone
overothers?

 

 

 

 

Considethevhandbotledyszuhadmdmosehom.Weeyoumdsfledwidndnrmngeof
choices? Is there annytlinng you would have liked to see.

 

 

 

 

145

92

93

94

95

96

97

98

Survey Said by MarketingMasters Page 8

Considednevh'tualbotleszboyouhaveanyothecommentsabonnthemflsthee
mnythingyouwonndhavechanged?

 

 

 

 

Pleasetaketflstimetohntlcateanymmtionsyou haveabouttlisexpe'iment.Arethe'e
slingsyoudo notnnnderstand?Tlingsthatareuncleartoyou?

 

 

 

 

Doyonnmderstandtheoverallpurposeoftheexpe'iment?
OYes Ollo
Ifeverytfigisnotdear,pleaseexplfiwhatisnnndear.

 

 

 

 

 

 

 

 

Wmtheeanytlfingindmexpeimemthatyoufomndodd,coniushng,ortlstwbhng?
OYes ONo
mewumeflngtMmfoudoMmmwhawmh,mdwhyddyon

 

 

 

 

146

SurveySaidbyMarketfl Masters Page 9

99 Heuelemknowflyoufdtanytflngindmprocednneafleaedyowbehaviorduingdm
expe'Iment.Thiswillheipusinnprovethe procednn'einthe future.

 

 

 

 

100 Thankyouforyom'hnvolvemenLPhaseteldmexpeimentedutyouarefinflshed.

147

APPENDIX D

AVATAR PARTICIPANT INSTRUCTION SHEET

148

Instruction Sheet

Thank you for agreeing to participate in this project.

During this part of your participation, you will enter a computerized virtual
environment. During your time in the environment, you can speak into a
microphone to communicate with others.

Later this semester, some teams will be asked to return to the lab to participate in
a virtual scavenger hunt. This may include you and your partner. As you may
know, in a scavenger hunt, you are given a list of items and must retrieve all of
the items on the list in the quickest amount of time possible. In this scavenger
hunt, you will search the Internet for objects not only as quickly as possible, but
also those offered at the best price. The best team will win $100 and other teams
will win prizes just for participating. We may contact you later regarding this
opportunity.

Your assignment for today is to meet your partner who is a student at another
university. Please begin by introducing yourself and then, once your partner does
the same, summarize your skills and experience that will be relevant and/or
helpful in carrying out this task and they will do the same. First, tell your partner
who you are and they will do the same. Following this brief introduction, you
should discuss your skills and experience, relevant to online searches for
telecommunication technologies, or lntemet searches generally. Please tell them
all they need to know, including previous experience and favorite search engines
or tools. You will only get one turn to tell them about your skills due to time and
technology constraints, but you can speak for as long as you feel you need
during your turn. Please speak slowly and clearly so they can understand.

Note: When you have finished introducing yourself, please push the button on
the screen to indicate that your turn is over and you are ready to listen to your
partner. This interaction environment is the one you will use for the scavenger
hunt, so try to get a feel for it and a sense of your partner as well.

If you have any questions at this point, please ask the experimenter. Othenlvise,
you may use the back of this page to draft your introduction and think about how
you will introduce yourself and which skills will best help your team in the future
project.

Please let the experimenter know when you are ready.
Thank you.

149

Summary of Instructions

Press appropriate button

Briefly introduce yourself

Press button to indicate your turn is over

Listen to your partner's introduction

Describe your skills and experience relevant to an Internet scavenger hunt
Press button to indicate your turn is over

Listen to your partner's skills and experience

Let us know you are finished.

93939199.”?

To avoid repetitions or difficulty, we recommend that you draft your skills in the
space below. This will allow you to read your skills to your partner so you do not
forget anything and so you can be clear and concise when you speak.

150

APPENDIX E

AGENT PARTICIPANT INSTRUCTION SHEET

151

Instruction Sheet

Thank you for agreeing to participate in this project.

During this part of your participation, you will enter a computerized virtual
environment. During your time in the environment, you can speak into a
microphone to communicate with others.

Later this semester, some teams will be asked to return to the lab to participate in
a virtual scavenger hunt. This may include you and your partner. As you may
know, in a scavenger hunt, you are given a list of items and must retrieve all of
the items on the list in the quickest amount of time possible. In this scavenger
hunt, you will search the Internet for objects not only as quickly as possible, but
also those offered at the best price. The best team will win $100 and other teams
will win prizes just for participating. We may contact you later regarding this
opportunity.

Your assignment for today is to meet your partner who is a bot, programmed at
another university to interact with you. Please begin by introducing yourself and
then, once your partner does the same, summarize your skills and experience
that will be relevant and/or helpful in carrying out this task and they will do the
same. First, tell your partner who you are and they will do the same. Following
this brief introduction, you should discuss your skills and experience, relevant to
online searches for telecommunication technologies, or lntemet searches
generally. Please tell them all they need to know, including previous experience
and favorite search engines or tools. You will only get one turn to tell them about
your skills due to time and technology constraints, but you can speak for as long
as you feel you need during your turn. Please speak slowly and clearly so they
can understand.

Note: When you have finished introducing yourself, please push the button on
the screen to indicate that your turn is over and you are ready to listen to your
partner. This interaction environment is the one you will use for the scavenger
hunt, so try to get a feel for it and a sense of your partner as well.

If you have any questions at this point, please ask the experimenter. Otherwise,
you may use the back of this page to draft your introduction and think about how
you will introduce yourself and which skills will best help your team in the future
project.

Please let the experimenter know when you are ready.
Thank you.

152

Summary of Instructions

9. Partner will introduce themselves (red “wait” button will be on)

10. Briefly introduce yourself

11.Press green “done” button to indicate your turn is over

12. Listen to your partner's skills and experience

13. Describe your skills and experience relevant to an lntemet scavenger hunt
14. Press green “done” button to indicate your turn is over

15.Your partner will sign off

16.Sign off with your partner and press green “done” button

17. Let us know you are finished.

To avoid repetitions or difficulty, we recommend that you draft your skills in the
space below. This will allow you to read your skills to your partner so you do not
forget anything and so you can be clear and concise when you speak.

153

BIBLIOGRAPHY

Bibliography

Adrianson, L., 8 Hjelmquist, E. (1991). Group Processes in face-to-face and
computer-mediated communication. Behavior 8 Information Technology,
10(4), 281-296.

Ambady, N., Hallahan, M., 8 Rosenthal, R. (1995). On Judging and Being
Judged Accurately in Zero-Acquaintance Situations. Journal of Personality
and Social Psychology, 69(3), 518-529.

Argyle, M. (1975). The Syntaxes of Bodily Communication. In J. Benthal 8 T.
Polhemus (Eds), The Body as A Medium of Expression : E.P. Dutton 8
Co., Inc.

Argyle, M. (1988a). Clothes, Physique, and other aspects of appearance. In M.
Argyle (Ed.), Bodily Communication. 2nd Ed. (2 ed., pp. 233-254):
International Universities Press, Inc.

Argyle, M. (1988b). The Explanation of bodily communication. In M. Argyle
(Ed.), Bodily Communication. 2nd Ed. (2 ed., pp. 290-304): International
Universities Press, Inc.

Asch, S. (1958). The Metaphor: A Psychological Inquiry. In R. Tagiuri 8 L.
Petrullo (Eds), Person Perception and Interpersonal Behavior (pp. 86-94).
Stanford, CA: Stanford University Press.

Badler, M, Phillips, 0., 8 Webber, B. (1993). Simulating Humans: computer
graphics animation and control. New York: Oxford University Press.

Bargh, J., 8 Pietromonaco, P. (1982). Automatic Information Processing and
Social Perception: the Influence of Trait Information Presented Outside of
Conscious Awareness on Impression Formation. Journal of Personality
and Social Psychology, 43(3), 437-449.

Benford, S., Greenhalg, C., Bowers, J., Snowdon, D., 8 Fahlén, L. (1995, May
7-11 1995). User Embodiment in Collaborative Virtual Environments.
Paper presented at the CHI '95 Mosaic of Creativity, Denver, CO.

Beniger, J. (1987). Personalization of Mass Media and the Growth of Pseudo-
Community. Communication Research, 14(3), 352-371.

Berger, C., 8 Calabrese, R. (1975). Some Explorations in initial interaction and
beyond: Toward a developmental Theory of Interpersonal communication.
Human Communication Research, 1, 99-112.

Berscheid, E., 8 Walster, E. (1969). InterpersonaIAttraction: Addison-Wesley
Publishing Co.

Biocca, F. (1997). The Cyborg's Dilemma: Progressive Embodiment in Virtual
Environments. Paper presented at the Proceedings of the Second
lntemational Cognitive Technology Conference, Aizu, Japan.

155

Biocca, F., 8 Nowak, K. (19993). Communication and Progressive Embodiment
in Virtual Environments. Paper presented at the International
Communication Association, San Francisco, CA.

Biocca, F ., 8 Nowak, K. (1999b). I feel as if I’m here, inside the computer:
Toward a theory of presence in Advanced Virtual Environments. Paper
presented at the lntemational Communication Association, San Francisco,
CA.

Biocca, F., Nowak, K., 8 Lauria, R. (1997). Virtual Reality. In C. H. Sterling (Ed.),
Focal Encyclopedia of Electronic Media. Boston: Focal Press.

Bodenhausen, G., 8 Macrae, C. N. (1998). Stereotype Activation and Inhibition.
In J. Robert Wyer (Ed), Advances in Social Cognition (Vol. Xl, pp. 1-52).
Mahwah: Lawrence Erlbaum Associates.

Bodenhausen, G. V., 8 Wyer, R. S. (1985). Effects of Stereotypes on Decision
Making and lnfonnation-Processing Strategies. Joumai of Personality and
Social Psychology, 48(2), 262-282.

Bolter, J. D. (1984). Tun'ng's Man: Western Culture in the Computer Age.
Chapel Hill: The University of North Carolina Press.

Brent, E., 8 Thompson, G. A. (1999). Sociology: Modeling Social Interaction
With Autonomous Agents. Social Science Computer Review, 1 7(3), 313-
322.

Bruner, J. (1957). On Perceptual Readiness. Psychological Review, 64(2), 123-
151 .

Bull, P. (1983). Body Movement and Interpersonal Communication: John Wiley
8 Sons Ltd.

Bull, P., 8 Rumsey, N. (1988). The Social Psychology of Facial Appearance:
Springer-Verlag.

Burgoon, J. (1991). Relational Message Interpretations of Touch,
Conversational Distance, and Posture. Joumai of Nonverbal Behavior,
15(4), 233-259.

Burgoon, J., Buller, D., 8 Woodall, W. (1996). Nonverbal Communication; The
Unspoken Dialogue. (Second Edition ed.): McGraw-Hill Companies, Inc.

Burgoon, J., 8 Hale, J. (1987). Validation and Measurement of the Fundamental
Themes of Relational Communication. Communication Monographs, 54,
19-41.

Capin, T. K., Pandzic, I. S., Magnenat-Thalmann, N., 8 Thalmann, D. (1998).
Realistic Avatars and Autonomous Virtual Humans. Paper presented at
the VLNET Networked Virtual Environments Virtual Worlds in the Internet.

Cartwright, G. F. (1994). Virtual or Real? The Mind in Cyberspace. The Futurist,
March-April, 22-26.

156

Cassell, J., 8 Vilhjalmsson, H. (1999). Fully Embodied Conversational Avatars:
Making Communication Behaviors Autonomous. Autonomous Agents and
Multi-Agent Systems, 2(1).

Chaiken, S. (1980). Heuristic Versus Systematic Information Processing and the
Use of Source Versus Message Cues in Persuasion. Joumai of
Personality and Social Psychology, 39(5), 752-766.

Chaiken, S., 8 Eagly, A. (1983). Communication Modality as a Determinant of
Persuasion: The Role of Communicator Salience. Joumai of Personality
and Social Psychology, 45(2), 241-256.

Chesebro, J., 8 Bonsall, D. (1989). Computer-Mediated Communication; Human
Relationships in a Computerized World: The University of Alabama Press.

Ciolek, T. (1982). Zones of Co-presence in face-to-face interaction: some
observational data. Man-environment systems, 12(6), 223-242.

Clark, N. (1995). Rear-View Mirrorshades: The Recursive Generation of the
Cyberbody. Cyberspace/Cyberbodies/Cyberpunk; Cultures of
Technological Embodiment. Sage Publications.

Clatterbuck, G. (1979). Attributional Confidence and Uncertainty. Human
Communication Research, 5(2), 147-157.

Coker, D., 8 Burgoon, J. (1987). The Nature of Conversational Involvement and
Nonverbal Encoding Patterns. Human Communication Research. 13(4),
463-494.

Damer, B. (1997). Avatarsl: Exploring and Building Virtual Worlds on the
lntemet. Peachpit Press.

Davis, D. (1995). Illusions and Ambiguities in the Telemedia Environment: An
Exploration of the Transformation of Social Roles. Joumai of Broadcasting
8 Electronic Media, 39, 517-554.

Dion, K., Berscheid, E., 8 Walster, E. (1972). What is Beautiful is Good. Joumai
of Personality and Social Psychology, 24(3), 285-290.

Don, A. (1992). Anthropomorphism: from Eliza to Terminator 2. Paper presented
at the CHI '92 Conference Proceedings of the Human Factors in
Computing Design.

Drew, D. G., 8 Grimes, T. (1987). Audio-visual redundancy and TV news recall.
Communication Research, 14(4), 452-461.

Dryer, D. C. (1999). Getting Personal with Computers: How to Design
Personalities for Agents. Applied Artificial Intelligence, 13, 273-295.

Eisenberg, E. (1990). Jamming; Transcendence Through Organizing.
Communication Research, 17(2), 139-164.

Ekman, P., 8 Friesen, W. (1969). The Repertoire of Nonverbal Behavior:
Categories, Origins, Usage, and Coding. Semiotica. 49-98.

157

Ferrari, J., 8 Swinkles, A. (1996). Classic Cover-Ups and Misguided Messages:
Examining Face-Trait Associations in Stereotyped Perceptions of
Nonverbal Behavior. Joumai of Social Behavior and Personality, 11(1),
27-42.

Fisher, J. (1997). The Postmodern Paradiso; Dante, Cyberpunk, and the
Technosophy of cyberspace. In D. Porter (Ed.), lntemet Culture.

Fiske, S., 8 Neuberg, S. (1990). A Continuum of Impression Formation, from
Category-Based to lndividuating Processes: Influences of Information and
Motivation on Attention and Interpretation. In M. P. Zanna (Ed.), Advances
in Experimental Social Psychology (Vol. 23, pp. 1-74): Academic Press,
Inc.

Fiske, S. T., 8 Pavelchak, M. A. (1986). Category-Based Versus Piecemeal-
Based Affective Responses: Developments in Schema-Triggered Affect.
In R. M. Sorrentino 8 E. T. Higgins (Eds), Handbook of Motivation and
Cognition: Foundations of Social Behavior (pp. 167-203). New York: The
Guilford Press.

Franklin, S. (1997). Autonomous Agents as Embodied Al. Cybernetics and
Systems: An lntemational Joumai, 28, 499-520.

Gilbert, D. T., 8 Hixon, J. G. (1991). The Trouble of Thinking: Activation and
Application of Stereotypic Beliefs. Joumai of Personality and Social
Psychology, 60(4), 509-517.

Gill, M. J., Swann, W., 8 Silvera, D. H. (1998). On the Genesis of Confidence.
Joumai of Personality and social Psychology, 75(5), 1101 -1 1 14.

Goffman, E. (1963). Behavior in Public Places; Notes on the Social Organization
of Gatherings. New York: The Free Press.

Graber, D. (1990). Seeing is remembering: How visuals contribute to Ieaming
from television news. Joumai of Communication, 40, 134-155.

Hamilton, 0., 8 Sherman, J. (1994). Stereotypes. In J. Robert S. Wyer 8 T. K.
Srull (Eds.), Handbook of Social Cognition (Vol. 2, pp. 3-58). Hillsdale,
New Jersey: Lawrence Erlbaum Associates.

Hampshier, S. (1991). Biology. Machines, and Humanity. In J. Sheehan 8 M.
Sosna (Eds), The Boundaries of Humanity: Humans, Animals, and
Machines. (p. 253-256). Oxford, England: University of California Press.

Hastorf, A., Schneider, 0., 8 Polefka, J. (1970). Person Perception: Addison-
Wesley Company.

Hedberg, S. (1996). Agents for sale: first wave of intelligent agents go
commercial. IEEE Expert (special issue, December 1996), 16-23.

Heider, F. (1958). Perceiving the Other Person. In R. Tagiuri 8 L. Petrullo
(Eds.), Person Perception and Interpersonal Behavior (pp. 22-26).
Stanford, CA: Stanford University Press.

158

Henne, R, Mark, G., 8 Voss, A. (1998). Gestures for Social Communication in
Virtual Environments. Paper presented at the Workshop on Presence in
Shared Environments, Ipswich, Suffolk.

Hert, P. (1997). Social Dynamics of an On-Line Scholarly Debate. The
lnfonnation Society, 13, 329-360.

Hinton, P. R. (1993). The Psychology of Interpersonal Perception. New York:
Routeledge.

Horwitz, M. (1958). The Veridicality of Liking and Disliking. In R. Tagiuri 8 L.
Petrullo (Eds), Person Perception and Interpersonal Behavior (pp. 191-
209). Stanford, CA: Stanford University Press.

lchheiser, G. (1970). Appearances and Realities; Misunderstanding in Human
Relations: Jossey-Bass, Inc. Publishers.

Infante, D., Rancer, A., 8 Womack, D. (1997). Building Communication Theory.
(3 ed.). Prospect Heights, Illinois: Waveland Press, Inc.

Katz, E., Adoni, H., 8 Parness, P. (1977). Remembering the News: What the
Picture Adds to Recall. Journalism Quarterly, 54, 231-239.

Keil, F. (1994). The Birth and Nurturance of Concepts by Domains: The Origins
of concepts of Living Things. In L. Hirschfeld 8 S. Gelman (Eds), Mapping
the Mind: Domain Specificity in Cognition and Culture (pp. 234-254):
Cambridge University Press.

Knapp, M. (1975). The effects of the face and eyes on human communication.
In D. Porter (Ed.), Nonverbal Communication in Human Interaction. .

Koda, T. (1996). Agents with Faces: A Study on the Effects of Personification of
Software Agents. Unpublished MS Thesis, Massachusetts Institute of
Technology.

Konner, M. (1991). Human Nature and Culture: Biology and the Residue of
Uniqueness. In J. Sheehan 8 M. Sosna (Eds), The Boundaries of
Humanity: Humans, Animals, Machines (pp. 103-124). Oxford, England:
University of California Press.

Lakoff, G. (1987). Women, Fire, and Dangerous Things; What Categories
Reveal about the Mind. Chicago and London: The University of Chicago
Press.

Lang, A. (1995). Defining AudioNideo Redundancy from a limited- Capacity
Information Processing perspective. Communication Research, 22(1), 86-
1 15.

Laurel, B. (1990). Interface Agents: Metaphors with Character. In B. Laurel
(Ed.), The Art of Human-Computer Interface Design (pp. 235-366).
Reading, MA: Addison-Wesley.

159

Lea, M., 8 Spears, R. (1992). Paralanguage and Social Perception in Computer-
Mediated Communication. Joumai of Organizational Computing, 2(384),
321-341.

Lipton, M. (1996). Forgetting the Body: Cybersex and Identity. In L. Strate, R.
Jacobson, 8 S. Gibson (Eds), Communication and Cyberspace: Social
Interaction in an Electronic Environment. Cresskill, New Jersey: Hampton
Press, Inc.

Lombard, M., 8 Ditton, T. (1997). At the heart of it all: The concept of presence.
Journal of Computer-Mediated Communication, 3(2),
(http://www.ascusc.org/jcmclvol3/issue2l).

Lombard, M., 8 Ditton, T. (1999). Presence Measures (Unpublished
Manuscript): Temple University.

Magnenat-Thalmann, M., Kalra, P., 8 Escher, M. (1998, MAY 1998). Face to
Virtual Face. Paper presented at the Proceedings of the IEEE.

Masters, M. (2000). SurveySaid (Version 1) [Survey Respondent software].
DePere, Wisconsin.

McCroskey, J. (1966). Scales for the Measurement of Ethos. Speech
Monographs, 33(1), 65-72.

McCroskey, J. (1971). Ethos, Credibility, and Communication in the Real World.
North Carolina Joumai of Speech, 4, 24-31.

McCroskey, J., Hamilton, P., 8 Weiner, A. (1974). The Effect of Interaction
Behavior on Source Credibility, Homophily, and Interpersonal Attraction.
Human Communicafion Research, 1(1), 42-52.

McCroskey, J., 8 McCain, T. (1974). The Measurement of Interpersonal
Attraction. Speech Monographs, 41 , 261-266.

McLuhan, M. (1964). Understanding Media: McGraw Hill.

Mehrabian, A. (1967). Orientation and behaviors and nonverbal attitude
communication. Joumai of Communication, 17, 324-332.

Mehrabian, A. (1972). Nonverbal Communication. Aldine Atherton, Inc.

Moore, 0., 8 Cavanagh, P. (1998). Recovery of 30 volume from 2-tone images
of novel objects. Cognition, 67, 45-71.

 

Nowak, K., 8 Anderson, T. (1999, May 1999). Communicating Emotions in
CMC: In search of a Sufficiency Threshold. Paper presented at the
lntemational Communication Association, San Francisco, CA.

Oravec, J. (1996). Virtual Individuals, Virtual Groups; Human Dimensions of
Groupware and Computer Networking: Cambridge University Press.

Otta, E., Abrosio, F., 8 Hoshino, R. (1996). Reading a Smiling Face: Messages
Conveyed by Various forms of Smiling. Perceptual and Motor Skills. 82,
1 1 1 1-1 121.

160

Pallak, S. (1983). Salience of a Communicators Physical Attractiveness and
Persuasion: A Heuristic Versus Systematic Processing Interpretation.
Social Cognition. 2(2), 158-170.

Palmer, M. (1995). Interpersonal Communication and Virtual Reality: Mediating
Interpersonal Relationships. In F. Biocca 8 M. Levy (Eds),
Communication in the Age of Virtual Reality. Lawrence Erlbaum
Associates.

Parise, S., Kiesler, S., Sproull, L., 8 Waters, K. (1999). Cooperating with life-like
interface agents. Computers in Human Behavior, 15, 123-142.

Parks, M., 8 Adelman, M. (1983). Communication Networks and the
Development of Romantic Relationships; An Expansion of Uncertainty
Reduction Theory. Human Communication Research, 10(1), 55-79.

Parks, M., 8 Floyd, K. (1996). Making Friends in Cyberspace. Joumai of
Communication, 46(1), 80-97.

Parks, M., 8 Roberts, L. (1997). Making MOOsic: the Development of personal
relah’onships on-Iine and a comparison to their off-line counterparts. Paper
presented at the Presented at the Western Speech Communication
Association. February 1997. Monterrey, California.

Patterson, M. (1994). Interaction Behavior and Person Perception; An
Integrative Approach. Small Group Research. 25(2), 172-188.

Patterson, M. (1995). Invited Article: A Parallel Process Model of Nonverbal
Communication. Joumai of Nonverbal Behavior, 19(1), 3-29.

Patterson, M. L., 8 Stockbridge, E. (1998). Effects of Cognitive Demand and
Judgment Strategy on Person Perception Accuracy. Joumai of Nonverbal
Behavior, 22(4), 253-263.

Paulos, E., 8 Canny, J. (1997). Ubiquitous tele-embodiment: applications and
implications. lntemational Joumai of Human-Computer Studies, 46, 861 -
877.

Petrie, C. (1996). Agent-Based Engineering, the Web, and Intelligence. IEEE
Expert, 35400, 24-29.

Reeves, 8., Lang, A., Thorson, E., 8 Rothschild, M. (1989). Emotional
Television Scenes and Hemispheric Specialization. Human
Communication Research, 15(4), 439-508.

Reeves, 8., 8 Nass, C. (1996). The media equation: How people treat
computers, television, and new media like real people and places.
Stanford, CA: CSLI Publications.

Rice, R. (1993). Media Appropriateness; Using social presence theory to
compare traditional and new organizational media. Human
Communication Research, 19(4), 451-484.

161

 

Rice, R., 8 Love, G. (1987). Electronic Emotion; Socioemotional Content in a
Computer-Mediated Communication Network. Communication Research,
14(1), 85-108.

Rice, R., 8 Tyler, J. (1995). Individual and organizational Influences on Voice
Mail use and Evaluation. Behavior and Information Technology, 14(6),
329-341.

Ritvo, H. (1991). The Animal Connection. In J. Sheehan 8 M. Sosna (Eds), The
Boundaries of Humanity: Humans, Animals, Machines (pp. 68-84). Oxford,
England: University of California Press.

Rock, l., Schreiber, C., 8 R0, T. (1994). The Dependence of 2-Dimensional
shape perception on orientation. Perception, 23(12), 1409-1426.

Salomon, G. (1990). Cognitive Effects With and Of Computer Technology.
Communication Research, 17(1), 26-44.

Schiano, D. J. (1999). Lessons from LambdaMOO: A social, text-based virtual
environment. Presence: teleoperators and virtual environments, 8(2), 127-
139.

Schmitz, J., 8 Fulk, F. (1991). Organizational Colleagues, Media Richness, and
Electronic Mail; A test of the Social Influence Model of Technology Use.
Communication Research, 18(4), 487-523.

Sheehan, J. (19913). Introduction: Humans and Animals. In J. Sheehan 8 M.
Sosna (Eds), The Boundaries of Humanity: Humans, Animals, Machines
(pp. 27-35). Oxford, England: University of California Press.

Sheehan, J. (1991b). Introduction: Humans and Machines. In J. Sheehan 8 M.
Sosna (Eds), The Boundaries of Humanity: Humans, Animals, Machines
(pp. 135-141). Oxford, England: University of California Press.

Short, J., Williams, E., 8 Christie, B. (1976). The Social Psychology of
Telecommunications. London. John Wiley 8 Sons, Ltd.

Siegel, J., Dubrovsky, V., Kiesler, 8., 8 McGuier, T. (1986). Group Processes in
Computer-Mediated Communication. Organizational Behavior and Human
Decision Processes, 37, 157-187.

Spender, D. (1996). Nattering on #16 Net Women, Power and Cyberspace.
Australia. Spinifex Press.

Sproull, L., Subramani, M., Kiesler, S., Walker, J., 8 Waters, K. (1996). When
the Interface is a Face. Human Computer Interaction. 11(2), 97-124.

Srull, T., Lichtenstein, M., 8 Rothbart, M. (1985). Associative Storage and
Retrieval Processes in Person Memory. Joumai of Experimental
Psychology: Learning, Memory and Cognition, 11(2), 316-345.

Steinfield, C. (1986). Computer-Mediated Communication in an Organizational
Setting: Explaining Task-Related and Socioemotional Uses. In M.
McLaughlin (Ed.), Sage Publications. .

162

Stephenson, N. (1993). Snow Crash. New York: Bantam Books.

Steuer, J. (1994). Defining Virtual Reality: Dimensions Determining
Telepresence. In F. Biocca 8 M. Levy (Eds), Communication in the Age of
Virtual Reality. Hillsdale, NJ: Lawrence Erlbaum.

Suler, J. (1996). Life at the Palace; A cyberpsychology case study. online
www1.rider.edu/~suler/psycyber/psycyber.html.

Suler, J. (1997). Communicative Subtlety in Multimedia Chat; How many ways
can you say "hi" at the Palace? online
www1.rider.edu/~suIer/psycyber/psycyber.html.

Swann, W. B., 8 Gill, M. J. (1997). Confidence and Accuracy in Person
Perception: Do we Know What We Think We Know About Our
Relationship Partners? Joumai of Personality and Social Psychology,
73(4), 747-757.

Tagiuri, R. (1958). Introduction. In R. Tagiuri 8 L. Petrullo (Eds), Person
Perception and Interpersonal Behavior (pp. ix-xvii). Stanford, CA: Stanford
University Press.

Takeuchi, A., 8 Naito, T. (1995, May 7-11, 1995). Situated Facial Displays:
Towards Social Interaction. Paper presented at the "Chi '95 Mosaic of
Creativity, ACM".

Tardy, C. (1988). Interpersonal Evaluations: Measuring Attraction and Trust. In
C. Tardy (Ed.), A handbook for the study of human communication.

Turkle, S. (1991). Romantic Reactions: Paradoxical Responses to the Computer
Presence. In J. Sheehan 8 M. Sosna (Eds), The Boundaries of Humanity:
Humans, Animals, Machines (pp. 224-252). Oxford, England: University of
California Press.

Turkle, S. (1995). Life on the screen: Identity in the age of the lntemet. New
York: Simon 8 Schuster.

Ullman, S. (1998). Three-dimensional object recognition based on the
combination of views. COGNITION, 67(1-2), 21-44.

Vilhjalmsson, H. (1996). Avatar Interaction.
http://hannes. www. media. mit. edu/people/hannes/project/index. html.

Walther, J. B. (1996). Computer-mediated communication: Impersonal,
interpersonal, and hyperpersonal interaction. Communication Research,
23, 3-43.

Waskul, D., 8 Douglass, M. (1997). Cyberself: The Emergence of Self in On-
Line Chat. The Information Society, 13(4), 375-397.

Watson, N. (1997). Why We Argue about Virtual Community: A Case Study of
the Phish.Net Fan Community. In S. Jones (Ed.), Virtual Culture: Identity
and Communication in Cybersociety. London: Sage Publications.

163

West, C., 8 Zimmerman, D. (1991). Doing Gender. In J. Lorber 8 S. A. Farrell
(Eds), The Social Construction of Gender (pp. 13-37). Newbury Park, CA:
Sage Publications.

Wexelblat, A. (1997). Don 't Make That Face: a report on anthropomorphizing an
interface, [Web Page]. Available:
http://wex.www.media.mit.edu/people/wex/anthro-expt-paper/Anthro-r.htm
[1999, November].

Williams, B. (1991). Prologue: Making Sense of Humanity. In J. Sheehan 8 M.
Sosna (Eds), The Boundaries of Humanity: Humans, Animals, Machines
(pp. 13-25). Oxford, England: University of California Press.

Wyer, R. S., 8 Carlston, D. E. (1994). The Cognitive Representation of Persons
and Events. In J. Robert S. Wyer 8 T. K. Srull (Eds), Handbook of Social
Cognition (2 ed., Vol. 1: Basic Processes, pp. 41-98). Hillsdale, New
Jersey: Lawrence Erlbaum Associates.

Yates, S. (1997). Gender, identity and CMC. Joumai of Computer Assisted
Learning. 13, 281-290.

Yzerbyt, V., Schadron, G., Leyens, JP. 8 Rocher, S. (1994). Social
Judgeability: The Impact of Meta-Informational Cues on the Use of
Stereotypes. Joumai of Personality and Social Psychology, 66(1), 48-55.

164

IIIIIIIIIIIIIIIIIIIIIIIIIIIIII

'lll‘llllllllllllllllllllllllllllllllllll'lllllllHI