Leadership and collaboration in shared virtual environments

Abstract

We present an experiment that investigates the behaviour of small groups of participants in a wide-area distributed collaborative virtual environment (CVE). This is the third and largest study in a series of experiments that have examined trios of participants carrying out a highly collaborative puzzle-solving task. The results reproducing those of earlier studies suggest a positive relationship between place-presence and co-presence, between co-presence and group accord, with evidence supporting the notion that immersion confers leadership advantage

Full text

Leadership and Collaboration in Shared Virtual Environments
Anthony Steed, Mel Slater, Amela Sadagic
Department of Computer Science
University College London
Gower St, London, WC1E 6BT, UK
{A.Steed | M.Slater | A.Sadagic}@cs.ucl.ac.uk
Adrian Bullock, Jolanda Tromp
Department of Computer Science
University of Nottingham
University Park, Nottingham, NG7 2RD, UK
{A.Bullock | J.Tromp}@cs.nott.ac.uk
Abstract
We present an experiment that investigates the
behaviour of small groups of participants in a wide-area
distributed collaborative virtual environment (CVE).
This is the third and largest study in a series of
experiments that have examined trios of participants
carrying out a highly collaborative puzzle-solving task.
The results reproducing those of earlier studies suggest a
positive relationship between place-presence and co-
presence, between co-presence and group accord, with
evidence supporting the notion that immersion confers
leadership advantage.
1. Introduction
This paper describes an experiment that was designed
to investigate what happens when a small group of
strangers meet together to carry out a joint task in a
virtual environment. The issues of interest are the
relationship between computational resources and
leadership, presence of being in a place and co-presence,
the sense of being with other people, and accord between
the group members. This is the third in a series of
experiments examining these issues, in an attempt to
characterise what makes working together in virtual
environments enjoyable and productive, or even
possible.
2. Background
In all three experiments we have used the following
task. Groups of three strangers meet for the first time in
an office environment. The participants have avatars
coloured Red, Green or Blue, and refer to themselves
and the others by colour throughout the experiment.
They meet in a small room, and have to find their way
together to another room that has a series of puzzles
written on pieces of paper stuck around the walls. On
each piece of paper is a set of words or phrases each
prefixed by a number. The participants have to rearrange
all the words with the same number in order to form a
pithy saying.
This task was chosen because it requires collaboration
between the members since it is difficult for one
participant to remember all the words in the phrase. The
task can be divided by each participant examining a
different area of the room.
The first experiment was an exploratory study to
generate hypotheses about how people would conduct
themselves in such a highly collaborative task[5]. For
this purpose, the scenario was carried out by each trio
twice, first in a virtual environment, and then in the real
environment from which the virtual had been modelled.
In the virtual experience, one participant was immersed
using a head-mounted display, and the others were using
desktop machines. The scenario was implemented in the
Dive3.2 software[2]. There were 10 groups in the study.
The results suggested that the immersed person
tended to emerge as leader in the virtual group, but that
this advantage was lost in the real meeting. Group
accord tended to be higher in the real meeting than in the
virtual meeting. Responses such as embarrassment could
be generated in the virtual meeting, even though the
individuals were presented to one another by very simple
avatars. The study also found a positive relationship
between presence of being in a place, and co-presence,
that is the sense of being with the other people. Accord
in the group increased with presence, the performance of
group, and with the proportion of females in the group.
The second experiment demonstrated the feasibility
of carrying out a similar study over a wide-area network
[7]. Four groups of 3 participants completed the virtual
version of the task only. The scenario was implemented
in the Dive3.2 software[2]. There was no immersed
participant and correspondingly no correlation of
leadership with immersion. There was evidence
supporting the earlier results on group-accord.

3. Experimental Design
In this latest experiment there were three participants
in each trial; one at University College London, UK, one
at the University of Nottingham, UK and one at
Integrated Information Systems, Greece. The difference
in display characteristics between participants was re-
introduced with the purpose of examining whether the
network condition would remove the effect of
immersion.
Given the distributed nature of the experiment, a
rigorously timed schedule was prepared and carried out.
The participants spent 4 or 5 minutes training to navigate
using the system. Once this was completed, the task was
explained and the participants were told their colour, and
instructed to refer to each other only by colour.
Throughout the trials the UCL participant was Blue
(immersed), the Nottingham participant was Red (non-
immersed) and the IIS participant was Green (non-
immersed).
Once all sites had completed this set-up stage, the
multi-user session was started in co-ordination between
the sites. The participants entered the environment
(Figure 1) and introduced themselves to one another and
decided how to complete the task (Figure 2). The task
was limited to 15 minutes duration.
A questionnaire was then used to elicit information
regarding the areas of interest outlined in the
introduction: the relationship between co-presence and
presence, the relationship between both types of
presence and immersion, the relationship between both
types of presence and accord within the group, and the
extent to which leadership is conferred by immersion.
Self-reported presence was assessed by six questions
each on a 7-point scale, with the higher score indicating
higher presence. The questions (paraphrased here)
required a response to the extent to which:
1. there was a sense of being in the room which has
pieces of paper with the riddles;
2. there were times during the experience when the real
world of the office in which the experience was
really taking place was forgotten;
3. that the virtual place is thought of as somewhere
visited rather than just images that were seen;
4. the sense of being in the virtual place was stronger
than of being in the real world of the office;
5. the structure of memory of the virtual place is similar
to the structure of memory of real places;
6. the virtual experience was overwhelming
In line with previous experiments where these
questions have been employed (e.g. [6]) the overall
measure was conservatively taken as a count of how
many scores of ‘6’ or ‘7’ were given to the questions.
Reported co-presence was assessed by eight questions
on the theme of ‘being with the other people’. These
were again rated on a 7-point scale, and required a
response to the extent to which:
1. there was a sense of being with the other people;
2. the computer interface seemed to vanish and there
was direct working with the other people;
3. the experience was more like working with other
people rather than interacting with a computer;
4. other people were forgotten and concentration was on
the task as if the individual was the only one;
5. the experience was like some other real experience of
working together with people;
6. a sense of being with other people rather than just
experiencing computer images;
7. the experience resembled being together with others
in a real-world setting;
8. there was a sense of other human beings interacting
with the individual;
Figure 1. Overview of the environment. Figure 2. Red, Green and Blue examining
one of the posters.

Overall co-presence was measured as the number of
‘6’ or ‘7’ scores out of the eight questions.
Accord was assessed by a series of questions relating
to enjoyment, harmony and cooperation of the group.
These questions were rated on a 7-point scale, and
concerned the extent to which:
1. the experience was enjoyed in a similar manner to a
previous real meeting that was enjoyable;
2. there was a desire to meet the other two people;
3. the group was in harmony during the task;
4. the person felt comfortable with the other two;
5. there was a desire for the group to form again;
6. the other two people were cooperative;
7. there was no embarrassment.
Overall accord was measured as the average response
across all of these questions.
There were two questions that related to leadership,
one directly and the other indirectly. Each participant
was asked to score (out of 100) all three participants on
the degree to which that person “was the ‘leader’ or
main organiser”. In addition, there was a similar
question concerning who did most of the talking. An
overall leadership score was constructed for each
individual by taking the average of the assessments of
the other two members of the group. A similar method
was used for ‘talkativeness’.
In order to take into account the impact of personality
on the results there was a final question that assessed the
extent to which the individual generally experienced
social anxiety in everyday life. This employed the
Interaction Anxiousness Scale[3] where fifteen
statements are given, and the participant has to respond
in a range of 1-5, where the higher score corresponds to
a higher social anxiety. An example statement is “I often
feel nervous even in casual get-togethers” for which the
responses range from “the statement is not at all
characteristic of me”(1) to “the statement is extremely
characteristic of me”(5). The average score amongst the
general population for this test is reported as 38, whereas
for those with social anxiety problems the average score
is 55. For the participants in this experiment the average
social-phobic scores amongst the three groups were 32
(Red - Nottingham), 35 (Green - Greece), 36 (Blue -
London).
4. Technical Description
We used the dVS/dVISE 5.0 software from Division
Ltd[1] to implement the scenario. dVS does not support
audio communication between participants, hence the
Robust-Audio Tool (RAT) v.3.0.23 was used alongside
the main application. The environment ran at 20-30Hz
on all client machines.
The machine at UCL was a Silicon Graphics Onyx
with twin 196 MHz R10000 processors, Infinite Reality
graphics and 192M of main memory. The machine at
Nottingham was a Silicon Graphics Indigo2 with a
200MHz R4400 processor, High Impact graphics and
192M of main memory. The machine at IIS was a
Silicon Graphics Octane with a 195MHz R10000 and
128M main memory. The final machine involved was
the dVS and audio server at UCL. This was a Silicon
Graphics Indigo2 with a 200MHz R4400 processor and
64M of main memory. The dVS system uses a central
server to distribute initial scene state, and subsequently
uses peer to peer connections for event delivery.
The participant at UCL used an immersive system
employing a 2 tracker Polhemus Fastrak, Virtual
Research VR4 helmet and a 3D mouse with 5 buttons.
The other two participants used a desktop system with a
1280x1024 screen and a 2D mouse with three buttons.
The experiment was carried out over a network used
for weekly trials by the partners of the COVEN project
[4]. Each site was connected over ISDN to UCL with
mean round trip times over all trials being 100ms
between Nottingham and UCL, 450ms between
Nottingham and IIS and 300ms between UCL and IIS.
5. Results of the Statistical Analysis
There were 20 groups of 3 participants in the study.
The results of 14 participants were eliminated from the
analysis because of incomplete questionnaires. Of these
8 were the Green participants located in Greece. There is
obviously considerable difficulty in conducting a study
involving people with different native languages, and
this is probably reflected in the relatively high
proportion of non-native English speakers with
incomplete questionnaires.
Co-presence and Presence: As found in the previous
experiments there was a significant positive relationship
between presence and co-presence. Since co-presence is
scored as a count out of 8, logistic binomial regression
can be used with co-presence as the response variable,
and the presence score as the explanatory variable. This
results in a significant fit (Chi-squared = 18.6 on 1 d.f.).
This relationship is practically important whatever the
reason for the correlation. If one influences the other, or
if they are both independently caused by the same
underlying factors, it is useful to know that benefiting
one may positively influence the other.
Presence and immersion: There was no reported
significant difference between reported presence and co-
presence between the immersed and non-immersed
participants. Again this is in line with the previous

results.
Presence and Accord: As found previously,
individual accord is positively associated with co-
presence. The square of the correlation coefficient (R2) is
0.38, and the fit is significant at the 5% level.
Leadership and Immersion: In the first study the
immersed participants tended to emerge as leaders in the
virtual session but not in the real session. In the second
study there were no immersed participants, and no
special pattern of leadership emerged.
In this study the situation is more complex, and still
there is no uncomplicated answer that can be given with
confidence to this question. The Green participant (in
Greece) was almost never the leader. There is strong
evidence that the participants in Greece experienced the
dual problem of language difficulties with also the
poorest audio response. The speed of the network
connection from London to Greece and from Nottingham
to Greece, was 3 to 5 times slower than between London
and Nottingham. Recall also that eight of the twenty
Green participants had been eliminated from the
questionnaire data.
There was a very high correlation, as found in the
earlier studies, between leadership rating and degree of
talkativeness (R2 = 0.82). In fact 85% of the variation in
overall leadership score can be accounted for by the
degree of ‘talkativeness’ (positive correlation) and the
extent of social anxiety (negative correlation).
Independently of any other factors, the more talkative
and the less social discomfort a person generally
experiences, the more likely they were to have been
rated as leader, which is to be expected also in real-life
meetings.
Leadership and degree of talkativeness may be
considered as different aspects of one overall leadership
variable, computed as the average of the two. Using this
as a response in a regression analysis, 40% of the
variation in the response can be accounted for by the
‘colour’, gender and social-discomfort as explanatory
variables. A summary of the regression analysis is shown
in Table 1.
Bearing in mind that average social anxiety score of
32 for Red, the analysis suggests a leadership advantage
for the Blue participants, but not for the female
participants. For the males, substituting in the social
anxiety score of 32 for the Red participants gives a
predicted score of 29. Since the social anxiety score is
not significant for the Blue participants, their predicted
score is 63. In the case of the females, substituting in the
average social anxiety score gives a predicted leadership
rating of 29 for the Red females, and it is also 29 for the
Blue females. It is possible that the use of the head-
mounted display disadvantaged females compared to
males.
Table 1: Regression of Overall Leadership
Male Female
Red 51 -
0.7*social_phobic 64 -
0.7*social_phobic
Green 19+
0.3*social_phobic 12+
0.3*social_phobic
Blue 63–
0.5*social_phobic 29–
0.5*social_phobic
Non significant results are in italics.
6. Conclusions
The results of the statistical study based on the
questionnaire responses offers further support for the
relationship between co-presence and presence, between
co-presence and group accord. The question of
immersion and leadership is still very much an open one,
though there is evidence here supporting the notion that
immersion confers leadership advantage.
Acknowledgments
Manolis Avagianos and Nikos Kladias ran the
experiments at IIS, and Lisa Miller assisted at UCL. The
COVEN project is funded by the European ACTS
Framework, project number AC040.
References
[1] Division Ltd (1997) dVISE VDI Reference, Division
Ltd, Bristol, BS32 4JT, UK, http://www.division.com
[2] Frecon, E. and Stenius M. (1998) DIVE: A Scaleable
Network Architecture for Distributed Virtual
Environments, Distributed Systems Engineering Journal
- Special Issue on Virtual Environments, 5, pp.91-100.
[3] Leary, M. (1983) Social Anxiousness: The Construct
and its Measurement, Journal of Personality
Assessment, 47, 66-75.
[4] Normand, V. et al. (1999) The COVEN Project:
Exploring Applicative, Technical and Usage Dimensions
of Collaborative Virtual Environment, Presence:
Teleoperators and Virtual Environments, 8(2), in press.
[5] Slater, M., Sadagic, A., Usoh, M., Schroeder, R. (1999)
Small Group Behaviour in Virtual and Real
Environments: A Comparative Study. Accepted to
Presence: Teleoperators and Virtual Environments.
[6] Slater, M. Usoh, M., Steed, A. (1995) Taking Steps: The
Influence of a Walking Metaphor on Presence in Virtual
Reality, ACM Transactions on Computer-Human
Interaction (TOCHI) 2(3) September, 201-219.
[7] Tromp, J.G., Steed, A., Frecon, E., Bullock, A., Sadagic,
A., Slater, M. (1998) Small Group Behaviour
Experiments in the COVEN Project, IEEE Computer
Graphics and Applications 18(6), 53-63.