A General Theory on Presence

Authors: Stefan Thie, Jacoliene van Wijk
Presenting author: Stefan Thie
Date: 13 May 98
=======================================================
Stefan Thie
KPN Research
PO Box 15000
9700 CD Groningen
The Netherlands
tel.: +31 50 5821053                            +31 50 5261318
fax: +31 50 3122415
email: S.Thie@Research.kpn.com
================================================================

Experimental Evaluation of Social Virtual Presence in a Decision Making Task

Abstract

Many points of departure have been chosen in research on presence. In this paper a new, experimental testable overall theory on presence is proposed and a first attempt to test this theory is made. Also a comparison is made with earlier presence research done regarding presence measurements.
In the overall theory, the degree of (virtual) presence is related to the task people accomplish in a Shared Virtual Environment. In the case of a decision making task the main hypothesis is that presence will be higher if social presence cues will be maximized.
This hypothesis was tested in an experimental setting were 48 subjects, in groups of three, were asked to complete decision making tasks in a Shared Virtual Environment (3D). Social Virtual Presence has been made operational by manipulating Social Presence Cues.
Four ways to measure virtual presence were used, namely: a virtual presence questionnaire, a social virtual presence questionnaire, the extremity of the decision and the comeback rate.
The main conclusion of this research is that there is significant evidence that the relationship proposed in this theory between virtual presence and social virtual presence exists. Furthermore, the presence theory can explain the non significant findings e.g. the influence of the world models on virtual presence due to technical complications. This research should have a follow up in a Collaborative Virtual Environment with other in- and output devices, such as a CAVE. 
 

Introduction

We started our research with the idea that when people chat on the internet in a three dimensional Shared Virtual Environment (SVE), people perform a so called decision making task (McGrath, 1984). In order for a person to accomplish this task, certain needs have to be filled. These needs can be filled by the Shared Virtual Environment (Hollan & Stornetta, 1992). Specifying the task in this way makes it also possible to compare presence research in a meaningful way.
This decision making task in the experiment has a meassure of correctness. A method for measuring task performance is useful in order to test the hypothesis that task performance increases when people are experiencing a stronger amount of Presence.

Theory on Presence

We asked ourselves what kind of interface an SVE should have in order to maximize performance on a decision task. We hypothesized that if people felt as if they 'were in the virtual environment' that performance would be best. According to Steurer (1992) presence means: "The feeling of 'being in an environment'." This feeling can vary between detachment and immersion. In the case of detachment, presence is weak and in the case of immersion, presence is strong. In our research the environment is one that is made possible by a computer, it is a: 'Computer generated graphical three dimensional (3D) environment, in which one can navigate and communicate with other people.' The presence that a person feels in this virtual environment we call: virtual presence.
Loomis (1992) considers presence to be a feeling resulting from a learning process. This learning takes place in the human brain. It is based on the in- and output signals a person receives respectively generates. This learning process results in mental models. The shape of these mental models sculpture the feeling of presence as well as virtual presence. So the interface should maximize the feeling of virtual presence in order to maximize decision task performance.

Susceptibility for Virtual Presence

A person receives and sends signals from and to the virtual environment. In this interaction is it thought that the way in which a person perceives (virtual) presence depends on the susceptibility of that person for virtual presence. The susceptibility for virtual presence contains of two main aspects:
The first aspect is the conscious ‘will’ to accept the signals from the virtual environment.
The second aspect is the subconscious ‘ability’ of the persons brain to register signals from the virtual environment and rule out other signals, a biological based trait. This relates to an extensive amount of research that takes place on signal processing.
It is thought that there is an interaction between these two personal characteristics, an interaction between the conscious ‘will’ and the subconscious signal processing. In research on virtual presence this relationship should be taken into account. Improving these personal traits, if possible, or the presumed relationship, should increase virtual presence.

Mental models

When a person interacts with SVE two mental models will be activated and shaped: The model of the Real World (RW), responsible for presence and the model of the Virtual World (VW), responsible for virtual presence. Based on several results from presence research in literature we distinguish two main sub models within the model of the RW and the model of the VW: The first mental sub model is called the model of the 'non-self. The non-self is the mental model of the environment as an individual experiences it.
The second mental model is called the 'self'. The self is the mental model that a human develops of him or herself. The mental models of the self in the RW causes Personal Presence and it causes Personal Virtual Presence in the VW. The mental models of the non-self can be divided in a social model and an environmental model, in the real world causing social presence and environmental presence. In the virtual world causing social virtual presence and environmental virtual presence. We assume there could be an important interference between these models.
 

Social Virtual Presence

In order to get a grip on the users needs in this task, several fields of research have been addressed. The result of this research is that Social (Virtual) Presence is thought to be the most important communication need for this type of task.
Social Virtual Presence is understood to mean the feeling that there are other people present in the SVE. This mental sub model, Social Virtual Presence, is a part of the virtual non-self model.
Social Virtual Presence has been found to hold concepts that other fields of research bring forward as being of great importance in a decision making task. Literature on Grounding (Clarc en Brennan, 1991) as well as literature on (informal) communication (Short, William & Christie, 1976 and Kraut, Fish, Root, & Chalfonte, 1990) support these findings.
This sub presence concept is also mentioned by Heeter (1992). Heeter distinguished three types of (virtual) subpresence within the presence concept: Personal Presence, Social Presence and Environmental Presence.
A third argument for the importance of social virtual presence is the fact that even in the first developments of the telecommunication technologies, social scientists have distinguished such a concept. Those scientists saw Social Presence as a measurement of quality of a means of communication ( Short, Williams & Christie, 1976).
The last argument for the fact that social virtual presence is the focus of our attention in specifying the communication needs of the user (and testing if it is possible to improve virtual presence in a decision task) is the popularity of multi-user games (e.g.; Quake). This is because it seems that the success of games depends on the possibility to play with or against other ‘real live’ persons
Social Presence Cues show an evident overlap with Social Context Cues as being described in the deïndividualisation theory in media communication. (Zimbardo & Diender in Forsyth, 1990).
The signals that make a person aware of the presence of other people are imbedded in the so-called Social Presence Cues. Social Presence Cues can take many forms e.g. non-verbal cues, cues related to concepts like proximity and orientation, cues related to physical appearance or eye contact.
The social presence cues can be seen as the operationalisation of all the signals that can originate from other people in the SVE and can improve the sub mental model (social virtual presence) of the non-self and by doing so, can improve virtual presence as a whole.

After considering these parts of the virtual presence theory the following can be summarized.  In order to increase virtual presence, the learning process regarding the mental model of the virtual world  has to go smoothly. Furthermore, increasing virtual presence is task dependent. In case of this decision making task, manipulating social virtual presence is considered most important in trying to increase virtual presence as a whole. Also, if we want to maximize the experience of virtual presence, interference with feelings of RW presence should be kept to a minimum.

In the following experiment we test the hypothesis that if  virtual social presence increases, virtual  presence increases. We will also test if task performance will improve if virtual presence increases. Virtual social presence will be manipulated through social virtual presence cues. This manipulation will also be tested.

Experiment

The Subjects

In total there were 48 subjects. The subjects were divided into two groups (condition A and B) of 24 persons, each group containing 8 subgroups of three persons.
The two groups of  24 subjects were equal on age, gender (12 male/12 female), education and other relevant and irrelevant variables. The subjects were all experienced in browsing the Internet, had never participated in a similar experiment and were familiar with the English language.

The Material

Per group, subjects were seated in separate rooms and were able to communicate with each other through chat, gestures, expressions as well as audio. Subjects were represented by avatars in the virtual environment which looked like a ‘café’. Four multi-media pentium PC’s were used running NT and Win ‘95 operating systems. Software used included among others: Netscape Navigater 3.01 gold clients, Blaxxun Cyberhub and Passport multi-user clients and the corresponding servers. The decision making task was discussion task were people had to reach an agreement.

The Design

The main goal of the experiment is to examine the effect of manipulating social virtual presence cues. This was done by stating the following conditions:
Condition A Minimize Social Presence Cues / Condition B Maximize Social Presence Cues
The manipulation of social presence cues could be established by choosing an ‘independent-groups design’. Generally speaking it can be stated that tiredness and practice were ‘approximately’ equally present in both conditions.
Further, a combination between a random-groups design and a matched-groups design has been chosen. Subjects were matched on the following relevant subject variables: gender and experience. On all other relevant subject variables like intelligence and susceptibility-for-virtual presence a random draw out of the population should cause an equal distribution among the conditions.

The Procedure

Before the test of a condition started, two pilot studies were taken to test the stability of the experimental setting.
The scenario: In order for the subjects not to meet each other before the experiment started, they arrived at the experimental location with 5 minutes in between. The subjects then read the manual and completed the pretest questions, the susceptibility-for-virtual presence questionnaire and gave an individual opinion on the topic on which they had to make a group decision. Then the subjects started to practice with the SVE application. After that the group decision had to be made. When the decision making task was completed, the virtual presence questionnaire and the social virtual presence questionnaire were filled in. Also, the subjects had to answer what their individual opinion on the discussion topic now was. Finally the subjects were told that the experiment was over, but that they might have to wait before they were escorted out of the building. In the mean time they could either go back to the SVE or read the reading material they brought.

Operationalisation of the Dependent Variable: Measuring Virtual Presence in a Decision Making Task

Measuring Susceptibility for Virtual Presence

The susceptibility of a person for virtual presence was measured with a susceptibility questionnaire, derived from earlier research done by Psotka (1993). This analyses has been repeated in this experiment.

Measuring Virtual Presence as a Whole

The amount of virtual presence a person experienced performing the task in the SVE was measured with a virtual presence questionnaire, also derived from earlier research done by Psotka (1993). Psotka found a very strong correlation (.82) between the susceptibility-for-virtual presence and the feeling of virtual presence. This correlation has also been calculated in this experiment, see results.

Measuring Comeback Rate

Comeback rate was used as another instrument to measure virtual presence. Comeback rate was measured as follows:
Subjects were told to wait after the experiment ‘officially’ had come to an end. Before hand the subjects were told to take something to read with them. They were given the choice between reading or re-entering the SVE. The leader of the experiment recorded if the subject came back to the SVE or preferred to read.

Measuring Virtual Social Presence

Social virtual presence was measured using questions derived from Short et al. (Short, Williams & Christie). Short describes two social presence measuring methods which we combined into one single social virtual presence questionnaire.

Measuring Task Performance

The decision making task could be measured by making use of research done within the context of the deïndividualisation theory. This is possible because Social Presence Cues and Social Context Cues are similar.

Deïndividualisation

The deïndividualisation theory gives an explanation for the fact that people sometimes behave in an extreme way when they communicate through a medium.
Deïndividualisation, in terms of communicating through a medium, can be caused by the fact that a person is anonymous.
Anonymity means: a person cannot be traced to his true identity e.g. because he doesn't give his or hers home address and by the lack of Social Context Cues provided by the means of communication (Jessup, Connolly & Tansik, 1990).
The theory further describes what a person feels if he/she experiences deïndividualisation: Low self awareness and Altered experiencing.
Finally deïndividualisation behavior is being described as extreme, atypical and polarized (Zimbardo & Diender in Forsyth). This behavior can lead to risky decisions or unconventional decisions (Kiesler &  Sproull, 1992 and Williams, 1977).
Based on the literature in this field, we found that Social Context Cues can be considered to be the same as Social Presence Cues. Tractability and social context cues are used to manipulate anonymity. Thus, if we manipulate social presence cues and tractability, social virtual presence could be considered the same concept as anonymity (in media research). This means that if we manipulate social presence cues and tractability, virtual presence can be investigated the same way as deïndividualisation. In this presence context we will say that social virtual presence will be manipulated and not anonymity.

Measuring Deïndividualisation / Social Virtual Presence

In the research that is conducted in the field of deïndividualisation, anonymity is being taken as the independent variable and is being measured by (among others) the dependent variable: ‘Group-polarisation’.
Group-polarisation hypothesis: the  "average post group response will tend to be more extreme in the same direction as the regroup responses" (Meyers & Lamm, 1976 in Forsyth).
It was thought that polarization processes could depend on the topic of discussion. In order to control this an emotional- and a business decision making task have been chosen. These so called ‘risky shift’ tasks originate from the research on group dynamics within social psychology  (Short, Williams & Christie).

Operationalisation of the Independent Variable

In the experiment an effort was made to manipulate the degree of social virtual presence. This was done by maximizing social virtual presence cues in condition B. For example: a subject had to make him/herself known to the other subjects by filling in and exchange referential cards with personal information in condition B. The other signals which are thought to increase or decrease virtual presence, had to be held constant over both conditions e.g. personal presence, social presence environmental presence, personal virtual presence, environmental virtual presence, the will, the ability and the learning process. The real world mental models that are thought to inhibit virtual presence, are held constant in condition A and B. All subjects were interacting in the same desktop situation. Manipulating social virtual presence meant that Personal- and Environmental Virtual Presence were also, however slightly, being manipulated. (This will be illustrated in table 2). The susceptibility for virtual presence (will and ability) was thought to be equal in both condition because of the randomization of the subjects. The learning process was being controlled in the following manners:
One: there was time to practice. This time to practice was longer in condition B, because this condition was more complex. And the subjects that were used had to have had some experience with browsing the Internet. Two: the 'speed' of the SVE was held constant in both conditions and was optimized.
 

 
Table 2: Operationalisation of Virtual Presence
 

Results

The correlation between the social virtual presence score and the virtual presence score on the questionnaires was found to be significant (0,458). This supports our presence theory. Namely, social virtual presence is a sub mental model of virtual presence. If social virtual presence increases, virtual presence should also increase, and the other way around.
However, the social virtual presence manipulation failed. Social virtual presence as well as virtual presence didn't score significantly different in the two conditions. Neither did the scores on extremity of behavior differ significantly.
Further, this research found that the research findings of Psotka could not be replicated. Not in the case of the factor analyses nor in the case of the correlation between the virtual presence questionnaire and the susceptibility-for-virtual presence questionnaire (0,247). The reliability of the questionnaires introduced by Psotka have to be considered very low: virtual presence questionnaire (Alpha: 0,6679) and the susceptibility-for-virtual presence questionnaire (Alpha: 0,4530).
A significant finding is the fact that people ‘come-back’ more in the condition where there are more social presence cues (0,23). Although this could be a interesting measurement for virtual presence, virtual (social) presence did not significantly differ in the same way as 'come-back' rate.

Discussion

The first point of comment is on the fact that there was a significant correlation between virtual presence and social virtual presence. However the social-presence-cues manipulation conducted in this experiment failed. Social virtual presence nor virtual presence scored significantly higher in condition B, the condition with the maximized social-presence-cues.
The second point of discussion is why the manipulation the social presence cues did not work. This might be caused by the experimental stage the questionnaires are in, but it might also be caused by the malfunctioning technical situation. These matters will be discussed more thoroughly in the following.
The third point of discussion is why Psotka’s findings were not reproduced in this experiment. In general we think it is important that presence research describes clearly the type of task people performed. Good comparisons in virtual presence research can only be made when the task has been made explicit. In the research Psotka performed, the task has not been described, which makes a good comparison impossible. Furthermore, the research Psotka performed included HMD’s, as this research used low immersive desktop conditions. According to our theory on virtual presence this means that there was less interference between real world mental models and virtual world mental models and could therefore make a great difference in experiencing virtual presence.  Here also the technical situation needs to be discussed, as well as the used questionnaires.
Our experiment has shown that virtual presence research lacks good presence measuring questionnaires.  And in order to be able to use other virtual presence measurements (task-performance or come-back rates) presence research needs to have some sort of measurement to hold on to. This could well be a task independent virtual presence questionnaire. An important point of comment on the virtual presence- and susceptibility-for-virtual presence questionnaires is their unreliability , res. alpha: 0,6922 and 0,4530. Also the factor analyses that Psotka conducted could not be repeated as well as the correlation between the two questionnaires. Perhaps a set of constructs, based on an integration between our proposed virtual presence theory and Psotka's, could lead the way towards better presence questionnaires.
Finally in explaining the results of this experiment the technical malfunctioning could be of major importance. The three most important technical problems were: speed of the interaction, halted application processes and audio communication quality.
The speed of the interaction, although on a LAN, was very slow. This drawback could have overruled the manipulation of social virtual presence cues. The slow interface impeded the learning process that is thought to underlie the development of the mental models of the virtual environment. Regarding the  halted application processes of the SVE software (bugs/errors) which needed correction during the experiment, the interference from the (strengthened) real world mental models on the virtual world models increased. An especially important technical drawback was the non-co-temporal audio implementation. Subjects were not given the opportunity to send audio signals that were  received by the other subjects without considerable delay of time. The SVE state of the art software for the internet had not yet fully developed in March 1997.

Further Research

Although the conducted experiment in this paper didn't turn out to deliver very significant results, this may not be a reason why further research on the proposed theory on virtual presence should not be performed. The theory provides a strong framework in which fundamental virtual presence research on very different levels can be conducted. For example with different tasks and different I/O devices. We will continue on this theory on virtual presence, in our next virtual presence research which will take place in a CAVE.
To be able to do further work on virtual presence, is necessary to improve the questionnaires e.g.; the virtual presence questionnaire, the virtual social presence questionnaire and  susceptibility-for-virtual presence questionnaire. These should be improved  in validity and reliability.
 

References

Clark, H. H., & Brennan, S.E. (1991). Grounding in communication. In R. M. Baecker (Eds.), Readings in Groupware and Computer-Supported Cooperative Work: assisting human-human collaboration (pp. 222-233). San Mateo: Morgan Kaufmann Publishers, 1993.

Daft, R. L. & Lengel, R. H.(1991). Organizational information requirements, media richness, an structural design. Management Science, 32, 554-571.

Ellis S. R. (1996). Presence of mind : A reaction to thomas sheridan
article. Presence: Teleoperators and Virtual Environments, 5(2).

Eveland, J. D. & Bikson, T. K. (1988). Work group structures and computer support: A field experiment. In R. M. Baecker (Eds.), Readings in Groupware and Computer-Supported Cooperative Work: assisting human-human collaboration pp. 431-442). San Mateo: Morgan Kaufmann Publishers, 1993.

Finholt, T., & Sproull, L. (1990). Electronic groups at work. In R. M. Baecker (Eds.), Readings in Groupware and Computer-Supported Cooperative Work: assisting human-human collaboration (pp. 431-442). San Mateo: Morgan Kaufmann Publishers, 1993.

Forsyth, D. (1990). Group dynamics. Monterey, CA: Brooks/Cole.

Frank, F. & Anderson, L. R. (1971). Effects of task and group size upon
group productivity and member satisfaction. Sociometry, 34, 135-149.

Heeter, C. (1992). Being there: The subjective experiene of presence. Presence: Teleoperators and Virtual Environments, 1(2), 262-271.

Held, R. M. & Durlach, N. I., (1992). Telepresence. Presence: Teleoperators and Virtual Environments, 1(1), 108-112.

Hollan, J. & Stornetta, S. (1992). Beyond being there. In R. M. Baecker (Eds.), Readings in Groupware and Computer-Supported Cooperative Work: assisting human-human collaboration (pp. 842-848). San Mateo: Morgan Kaufmann Publishers, 1993.

Jessup, L.M., Connolly, T. & Tansik, A. (1990). Toward a theory of automated group work: The deindividuating effects of anonymity. Small Group Research, 21(30), 333-348.

Kiesler , S., & Sproull, L. (1992). Group decision making and communication technology. In R. M. Baecker (Eds.), Readings in Groupware and Computer-Supported Cooperative Work: assisting human-human collaboration (pp. 842-848). San Mateo: Morgan Kaufmann Publishers, 1993.

Kraut, R. E., Fish, R. S., Root, R. W., and Chalfonte, B. L. (1990). Informal communication in organizations: form, functions, and technology. In R. M. Baecker (Eds.), Readings in Groupware and Computer-Supported Cooperative Work: assisting human-human collaboration (pp. 431-442). San Mateo: Morgan Kaufmann Publishers, 1993.

Laughlin, P. R., & Earley, P. C. (1982). Social combination processes, persuasive arguments theory, social comparison theory and choice shift. Journal of Personality and Social Psychology, 42, 273-280.

Loomis, J. M., (1992). Distal attribution and presence. Presence: Teleoperators and Virtual Environments, 1(1), 113-119.

McGrath, J. E. (1984). Groups: Interaction and Performance. In R. M. Baecker (Eds.), Readings in Groupware and Computer-Supported Cooperative Work: assisting human-human collaboration (pp. 165-168). San Mateo: Morgan Kaufmann Publishers, 1993.

McGuire, T., Kiesler, S. & Siegel, J. (1987). Group and computer-mediated discussion effects in decision making. Journal of Personallity and Social Psychology, 52(5), 917-930.

Psotka, J. M. Davison, S. & Lewis, S. A., (1993). Exploring immersion in virtual space. In R. Stuart & G. P. Panos (Eds.), Virtual Reality Systems tm, 1(2), 70-82.

Resnick L. B., Levine J. M. & Teasley S. D. (1991). Socially Shared Cognition.

Sheridan T. B. (1992). Musings on telepresence and virtual presence. Presence: Teleoperators and Virtual Environments, 1(1), 120-126.

Short, J., Williams, E. & Christie, B. (1976). The social psychology of telecommunications. London: John Wiley & Sons, 1976.

Short, J., Wiliams, E. & Christie, B. (1976). Visual communication and
social interaction. In R. M. Baecker (Eds.), Readings in Groupware and Computer-Supported Cooperative Work: assisting human-human collaboration (pp. 153-161). San Mateo: Morgan Kaufmann Publishers, 1993.

Short, J., Wiliams, E., & Christie, B. (1976). Communication modes and task
performance. In R. M. Baecker (Eds.), Readings in Groupware and Computer-Supported Cooperative Work: assisting human-human collaboration (pp. 169-175). San Mateo: Morgan Kaufmann Publishers, 1993.

Schloerb, D. W. (1995). A quantitative measure of telepresence. Presence: Teleoperators and Virtual Environments, 4(1).

Sheridan T. B. (1996). Further musings on the psychophysics of
presence. Presence: Teleoperators and Virtual Environnements, 5(2).

Steuer, J. (1992). Defining vrtual reality: Dimensionsdetermining telepresence. Jounal of Communication, 42(4), 73-93.

Suchman, L. & Trigg, R. (1991). Understanding practice: video as a
medium for reflection and design. In R. M. Baecker (Eds.), Readings in Groupware and Computer-Supported Cooperative Work: assisting human-human collaboration (pp. 205-217). San Mateo: Morgan Kaufmann Publishers, 1993.

Tromp, J. G., (1995). Virtual reality in the office: Evaluation of Massive by measuring telepresence and immersion. Supplement Proceedings ECSCW'95, Poster at European Computer Supported Cooperative Work. September, Stockholm.

Tromp, J. G., (1995). Presence, telepresence and immersion: The cognitive factors of embodiment and interaction in virtual environments. Proceedings of FIVE'95, Framework for Immersive Virtual Environments. December, London.

Vliet, G. van. (1982). Denken en doen bij experimenteel onderzoek. Groningen: Wolters-Noordhoff.

Williams, E (1977). Experimental comparisons of ‘face-to-face’ and mediated communication: A review. Psychological Bulletin, 84(5), 963-976.

Winograd, T., & Flores, F. (1986). Understanding computers and cognition. Massachusetts: Addison-Wesley Publishing Company.

Zeltzer, D., (1992). Autonomy, interaction and presence. Presence: Teleoperators and Virtual Environments, 1(1), 127-132.