A Study of User Attitude Dynamics in a Computer Game

Yang Cao, Golha Sharifi, Yamini Upadrashta, Julita Vassileva

University of Saskatchewan, Computer Science Department,

57 Campus Drive, Saskatoon, Saskatchewan S7N 5A9, Canada

Keywords: Human Factors, Internet and Collaborative Computing, Agents for Internet Computing

Abstract: When designing a distributed system where a certain level of cooperation among real people is important,

for example CSCW systems, systems supporting workflow processes and peer-to-peer (P2P) systems, it is

important to study the evolution of relationships among the users. People develop attitudes to other people

and reciprocate the attitudes of other people when they able to observe them. We are interested to find out

how the design of the environment, specifically the feedback mechanisms and the visualization may

influence this process. For this purpose we designed a web-based multi-player computer game, which

requires the players to represent explicitly their attitudes to other players and allows studying the evolution

of interpersonal relationships in a group of players. Two versions of the game deploying different

visualization techniques were compared with respect to the dynamics of attitude change and type of

reactions. The results show that there are strong individual differences in the way people react to success

and failure and how they attribute blame and change their attitude to other people involved in the situation.

Also the level and way of visualizing the other players’ attitude influences significantly the dynamics of

attitude change.

1 INTRODUCTION

There are many examples of solid user communities

that formed around pieces of technology (e.g.

slashdot.com), but there are many more examples of

failed ones. Exactly what went right in the thriving

communities and what went wrong in the others is

difficult to analyze. In our experience developing

and deploying I-Help (Greer et al., 2001), a multi-

agent environment supporting synchronous and

asynchronous peer-help in a University

environment, we discovered widely varying levels

of user participation in different classes. It seems

that not so much technical, but a complex interaction

of social factors played a significant role, like

rewards (in terms of marks, virtual money or

reputation/visibility in the group), attitudes (pre-

existing interpersonal relationships among users),

and personal beliefs (e.g. altruism). This experience

taught us that it is important to study the

sociological aspects of cooperation, and that the

application should model and support the existing

relationships among people, organizational

structures (Artikis et al., 2002 ; Sierra and Noriega,

2002) and incentives for cooperative action (Golle et

al., 2001). In the study described here we focus on

the following general questions:

• ho

w people develop interpersonal

relationships when interacting in a computer-

based multi-user environment,

• wh

at is the role of individuality in attributing

praise / blame in case of success/ failure, and

• doe

s the design of the environment,

especially the feedback given to the user

about the other users’ attitudes influence the

reciprocation of attitudes quantitatively or

qualitatively.

A multi-player game environment was designed

as a to

ol to study these questions. It requires the

players to represent explicitly their attitudes to the

other players and to change their attitudes towards

the other players depending on the outcome of the

game and their realization of the others’ attitudes

towards themselves. Different ways of visualizing

the others’ attitude (text vs. animated face displaying

emotion - smiley) were applied in two different

versions of the game.

222

Cao Y., Shariﬁ G., Upadrashta Y. and Vassileva J. (2004).

A Study of User Attitude Dynamics in a Computer Game.

In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 222-229

DOI: 10.5220/0002599202220229

 SciTePress

2 RELATED LITARATURE

There are many studies on the evolution of

cooperation the area of groupware and CSCW.

Methods have been proposed to support and manage

collaboration by suggesting appropriate roles,

detecting and helping resolve conflicts, and

assigning tasks depending on the expertise of the

users (Jermann et al., 2001). Enviroments exist that

create awareness of the other participants' actions or

focus of attention (Gutwin et al., 1995), or study the

participation rate and role taking through analysis of

the types of speech acts (Soller 2002, Soller et al.,

2002) and user actions (Muehlenbrock and Hoppe,

1999), and create models of how these acts relate to

effective collaboration and provide guidance about

what acitivities the participants should engage in to

improve collaboration (Barros and Verdejo, 2000).

However, organizational rules alone do not

necessarily yield the desired result, as a self-

organizing dynamic may appear in the organization

which guides the system away from the desired path

(Hummel and Schoder, 1995). Such dynamic most

often results from personal attitudes and

relationships. Most of existing CSCW work is

applied to settings where implicit social structures

already exist, i.e. the users know each other in

advance and have established relationships and

status. With the advance of telework environments,

there will be an increased need for CSCW

environments supporting collaboration between

users who have never met face to face and who

don’t know each other. Building up attitudes and

social relationships in such environments happens

exclusively during the process of collaboration,

mediated through the collaborative environment and

can therefore be strongly influenced by the design of

the environment.

While attitude formation has been studied in the

area of social psychology, the CSCW literature to

date has not paid much attention to the fact that

people often think in terms of relationships with

other people and their attitudes /feelings towards

other people govern to a high extent their actions.

Attitude formation is a complex process which has

been modelled theoretically from different

perspectives. For example, the balance theory

(Prendinger and Ishizuka, 2002; Rist and Schmitt,

2002), symmetry theory, congruence theory and

cognitive dissonance theories take a cognitive stance

and explain how people’s attitudes towards each

other are influenced by their (shared or different)

attitudes to important ideas, events or other people.

A more pragmatic view is that human attitudes

depend on past experience and reciprocation. For

example, if somebody has behaved badly towards

another one in the past, it is very likely that the

second one will develop a dislike to the first one

(without even trying to judge the motives). While

such behavior could be modeled theoretically (e.g.

the reciprocating “tit-for-tat” strategy in the iterated

Prisoner’s Dilemma) (Axelrod, 1984) and can be

implemented practically with machine learning

techniques, attempts to explicitly represent

relationships among users have been made only

recently. Models of trust updated by reinforcement

learning from experience (Yu and Singh, 2002a; Yu

and Singh, 2002b) and /or reputation, using other

agents as a source of indirect experience (gossip)

(Conte and Paolucci, 2002) have been proposed

recently in the area of multi-agent systems. These

studies have been concerned with the emerging

global properties of the system as a result of

introducing trust relationships among agents (e.g.

what types of equillibria can be reached, how robust

is the agent society with respect to “cheaters”).

Interpersonal relationships have been studied on

a global scale by sociologists. Studies of social

networks focus on the patterns of interactions within

a group and analyze particular properties of the

graph formed by the people (nodes) and their

interactions (edges): density, cohesiveness, etc.

There have been studies of CSCL envrionments

using social analysis, for example (Nurmela et al.,

1999), where the social network cohesiveness of the

group is measured to identify the prominent

participants in collaboration.

There has been a lot of interest recently in the

area of social sciences in general, and particularly in

the area of business management in the development

of “social capital” in a community or workplace,

resulting from positive weak ties (Granovetter,

1973) as a way to promote cooperation, information

flow and innovation at the workplace. We believe

that building social capital or positive relationships

can be an important incentive in CSCW systems

where there is no external source of motivation for

the users to collaborate (Vassileva, 2002). While

introducing currency and micro-payments can help

motivate users to help each other (Golle et al.,

2001), many users can actually feel repulsed from a

money-oriented system (Shirky, 2000); something

that we discovered also in our experience with I-

Help (Vassileva, 2002). People can be motivated by

the possibility to create relationships with other

people, and by participating in an active network of

relationships, creating thus a small world where

recognition and being liked by peers are important

factors for the individual (Vassileva, 2002).

We believe that representing and reasoning

expliclty about attitudes and relationships among

users could be applied in the areas of groupware and

CSCW, and this approach can provide a way to

A STUDY OF USER ATTITUDE DYNAMICS IN A COMPUTER GAME

223

handle emerging self-organizing group dynamics.

The design of the rules of interaction in the CSCW

can encourage the development of positive attitudes

and relationships and increase the motivation for the

users to act and to cooperate.

Multi-player computer games provide a good

context for exploring emerging social relationships.

A Swedish research project on a game called

"Kaktus" (Laaksolahti and Persson, 2001), allows

teenage users to experiment different social

behaviours and respond to various social pressures.

“Sims Online”, a multi-player simulation game

allows (according to the advertisement) to: “Build a

network of friends to enhance your power, wealth,

reputation and social standing.” Multi-player action

games such as "Dark Age of Camelot" provide a

even better ground for studying dynamic social

network issues. Recent surveys show that players are

troubled by cheaters and saboteurs. The rules of the

game (team-based player versus player conflict, no

direct communication with the other teams, no

ability to switch teams, etc.) set up a situation where

given perfect game balance, on any given night, a

player may lose 2/3 of his battles. This often leads

to frustration and looking for someone to blame.

Over time, teams that were intended to be unified

against a common enemy end up fragmented into

smaller, tighter communities that bicker among

themselves, only to reunite eventually and repeat the

cycle. Could some subtle alteration of the game

rules break this cycle and create large, happy

communities? Or do people naturally seek small

circles of friends and find reasons to isolate

themselves?

We propose a new way of exploring emerging

interpersonal relationships in a computer mediated

environment by using specially designed multi-

player games. In this way we can capture the time

evolution of social networks of real people, not

artificial agents, as with social simulation. The

players form relationships (even though only for a

short period of time, in the context of the game) and

are more willing to reveal their attitudes to each

other in a context of a game than in a real

environment. While it can be argued that the context

of the game is different than the context of a real

world collaboration environment, we believe that

most multi-player games reveal individual

characteristics of the players that can be seen also in

their real-world encounters. The game allows to

study the individual differences in the way people

change their attitudes, which can help in desgining

individualized feedback in CSCW environments.

The next section describes the design of a web-based

multi-player game called “Who likes me”.

3 GAME DESIGN

We want to study the evolution of personal

relationships among a group of people using a multi-

player web-based game. The rules of the game

require the users to express and modify explicitly

their attitude to the other players as a level of liking

or disliking.

In each round of the game a player picks a

destination player and has to send him/her a signed

packet containing 100 units. However, the packet

can’t be sent directly to the destination, but by

passing to one of the other players (the most liked

one). If the selected player likes the originator of the

packet, it passes it directly to another player (his/her

most liked player), but if s/he doesn’t like the

originator he/she will take part of the packet

proportional to the level of dislike and then pass it

further. This process continues until the packet

reaches the destination or is destroyed by the other

players. After each rounds of the game, the player

gets system feedback about what proportion of his

packet reached the destination, feedback about the

other players’ attitudes towards him/her and is able

to change his/her attitudes to the other players. After

each player completes a given number of rounds

(e.g. 10), the one who achieved the highest number

of transported successfully units wins the game.

The success of a player in the game is

determined by the attitude of the other players to

him/her. It is advantageous if the player has a

reciprocated positive relationship with at least one

other player. However, this is not enough, since if

the “friend” of the player passes his/her packet to

another one who dislikes him/her, the packet can be

destroyed nevertheless. Only through mutual liking

and cooperation can all players achieve high scores

(though in this case other factors will define who

wins the game, e.g. who sends packets faster).

However, the uncertainty in the other players’

attitude towards oneslef and the desire for

reciprocation after unsuccessful rounds make the

players increase or decrease their level of liking,

which makes the game dynamic, unpredictable and

interesting. Strategizing successfully in such a

complex situation is practically impossible.

3.1 Game Rules

The requirement for the game is that there should be

at least three people to play. The game starts by

player A signing in the system. Player A will be

provided with the list of pseudonyms of the current

players and will be required to enter his/her attitude

(how much he/she likes each of other player) as a

natural number from 5 (strong like) to 1 (strong

ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING

224

dislike). Player A can start to play a round of the

game by choosing one of the players as a

destination. Player A sends a packet with containing

100 units to destination. The packet continually

passes among the group of remaining players, until

it reaches the destination (fully or partially) or is

destroyed. Each intermediate player, receiving A’s

package takes away a number of parts proportional

to the level of dislike it holds towards A. The round

finishes when the packet reaches the destination

player or is destroyed. At the end of the round, the

player gets feedback about the success of his/her

package and feedback generated by the system about

the attitudes of the other players towards him/her.

We chose to provide only a rough summary of the

relations of the other players towards the player,

deduced from the observation of how the packet

travels and how much it looses. Only summary

information “likes” or “dislikes” is presented to the

player, but it is not clear if, for example, “dislikes”

means 3 or 1. We designed two different ways of

presenting the feedback to the user in two different

versions of the game – textual and graphical (see

Figure 1). After seeing the feedback, the player can

change his/her attitudes to any of the other players

(if s/he wishes) and play another round.

3.2 Agents Represent Players

Personal agents represent each player in the game,

thus saving the player from having to consider

individually each passed packet and ensuring

consistency in the forwarding of packages according

to the attitudes of the user towards the other players.

The personal agent maintains a list of attitudes {a

, …, a

} of the player towards the other k players.

A number a

∈ {1,2,3,4,5} where 1 (negative,

dislike) to 5 (positive attitude, like) represents each

attitude. The player assigns each the value of his/her

attitude to each of the other players, thus

"instructing" his/her agent how to play the game on

her behalf. During the course of the game, the agents

decide to whom to pass each packet sent to them and

how much to take away from it, depending on the

value of the attitude of the user towards the

originator of the package. The packet is sent to the

agent of the most liked player M | a

= max

, a

…, a

}. If the player dislikes completely the

originator R of the package, i.e. a

= 1, the agent

will destroy the packet, i.e. it will not pass it further.

Otherwise, the agent takes away n parts of the

package where n = 5 – a

and a

is the value of the

attitude of the player to the originator R of the

package. The agents do not reveal the attitudes of

their players to either other agents or to the system.

In summary, the rules for the agents to play are:

1. To preserve privacy the system is not allowed to

access the players' attitudes.

2. The agent that starts the round cannot send its

packet directly to the destination.

3. An agent of player A will not send a package to

the agent of a player B that A dislikes (i.e. there is

= 1 in A's attitude model).

4. Each agent of player A selects to pass the package

to the agent of the player M to whom the user with

the highest attitude value i.e. M | a

= max{a

, a

…, a

5. To prevent infinite loops in the game:

 The agent will not send the packet back to

its sender or to the owner of the packet.

 The agent selects a new agent to send the

packet when it receives the packet from the

same two previous senders twice.

6. If the player's packet is destroyed, the player's

agent will not pass any packet to the first agent that

received its previous packet.

7. If the player dislikes everybody (i.e. his/her

attitude to every other player is 1), it cannot play.

8. The initial packet for each round of play has a

value of 100.

9. The agent who receives the packet will destroy

the packet if its player dislikes (at level 1) the

packet's owner.

10. The agent who receives the packet will

decrement the value of the packet by 5 minus the

value of its attitude to the packet's originator.

4 EXPERIMENTAL RESULTS

We carried out experiments with the game to test the

following hypotheses:

• Individuals react differently, but consistently to

success and failure when changing their

attitudes to the other people involved in the

situation;

• People reciprocate the attitudes of other people,

when they become aware of them;

• The way feedback about other people’s attitudes

is given plays a role in the way people

reciprocate and in the dynamics of the attitudes.

A STUDY OF USER ATTITUDE DYNAMICS IN A COMPUTER GAME

225

Figure 1: System feedback about the other players' attitudes towards the player (textual and smiley versions)

To test the third hypothesis, we experimented

with two versions of the game, one with textual

feedback and one with feedback visualized with

smileys (shown in Figure 1). The preliminary results

generated by two approximately 45-minute

experiments with the two versions of the game are

summarised below.

Six participants played fifty rounds of the text

version of the game in total (i.e. packages sent by

different players) and answered survey forms in the

end. Seven different participants played fourty

rounds each with the smiley version (i.e. seventy

rounds in total). The participants had different

gender, age, and ethnic background. The group

using the smiley version was formed by computer

science graduate students, while the group with the

text version was mixed. In each set of experiments

the participants did not know each other (aliases

were used). The players were given a general

introduction about the game and the basic rules.

While the rounds of the game were not

synchronized across the players, there were five to

six players playing at the same time. The routes for a

packet to reach its destination were different for the

different rounds. The shortest route was a package

passed and destroyed by one player and the longest

one involved all six players several times and

reaching the destination without being destroyed.

Three kinds of results were possible in each round:

the packets reached the destination completely; the

packets did not reach the destination because they

were destroyed, and the packets reached the

destination partially.

The following cases reoccurred during the game:

• When everyone in the group strongly disliked

the originator, the packet couldn't be send to the

destination (direct consequence from rule 8).

• The shortest route of a package happens in two

cases: when a sender passes the packet to the

most liked other player and that player dislikes

strongly the sender, the packet is destroyed

immediately. The second case is when the

sender selects a player who likes strongly the

destination-player, because it passes the packet

directly to the destination (see rule 4).

• The longest route happens in a group where no

one prefers the destination player to the other

players (if the sender likes at least one player

and no one dislikes strongly the originator). In

this situation, the message is passed

continuously in the entire group according to

rule 5 until it finally reaches the destination.

• If a sender has good relations with others and

s/he selects to pass the package to a player who

also has good relations with others, the packet is

delivered to destination successfully. However,

if the sender selects a player who doesn't like

the others, the packet will not be sent to

ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING

226

destination even the sender has good relations

with others.

These cases were possible to deduce from the

rules of the game and the questionnaires showed that

players were aware of them when interpreting the

results of each round and were trying to strategize.

The only way for players to strategize was by

changing their attitudes towards the other players,

using information from the system about the success

of their package sending at each round and the

system-generated record of the like/dislike attitudes

of the other players towards them. In the next

sections, we present the main observations from the

two experiments.

4.1 Setting the Initial Attitudes to the

Other Players

Across the two versions, players were fairly

consistent in choosing their intitial attitude (positive,

negative or neutral). In the text version, 45% of all

initial attitude choices (i.e. each player’s choice of

attitude towards each other player) was positive

(levels 4 or 5), while in the smiley version, 49% of

the initial choices of attitude were positive. In both

the text version and smiley version the initial

negative attitude selection percentage was 17%. In

the text version 38%, and in the smiley version

34%, of the initial choices were netural. From these

numbers it seems that the players had neutral to

positive attitude disposition at start. Next we shall

see that they were fairly conservative in changing

their attitudes.

4.2 Dynamics of Attitude Change

There were 174 opportunities for attitude change in

total in the text version and 234 opportunities in the

smiley version. The total number of opportunities is

calculated as the sum of all feedback stages for each

player multiplied by the number of other players at

each stage. The breakdown of different scales of

attitude changes is consistent across the two

versions. Most players keep attitudes to other

players constant most of the time – 66.7% and 70%

of all opportunities for change of attitude for each

player (to all other players at each round) were not

used in the text version and the smiley version of the

game, respectively. Gradual change with one level

of liking/disliking makes 11% of all changes of

attitude in the textual version; it is slightly less

common (with 4%) than radical change of attitude

(with 2 or 3 levels). Gradual change (12% of all

changes of attitude) is slightly more common (with

1%) than the radical change in the experiment with

the smileys. Drastic change (from level 5 to level 1

or reverse) makes around 6% in both versions (6.3%

in the text feedback version and 6.4% in the smiley

version). In the cases when drastic change of attitude

took place, it was mostly negative (64% of all

drastic changes in the text version and 87% in the

smiley version were negative).

4.3 Typical Reactions

One typical reaction is drastically reducing the level

of liking to the most liked person after a partial or

complete failure to deliver the packet.

This reaction was observed particularly

frequently for specific players (e.g. all six drastic

changes made by Goofy and all four drastic changes

made by Daisy in the smiley version were negative

and came in response to partial failure to deliver a

packet,see figure 2). Three of the four drastic

changes made by Abraham in the text version were

of this type (figure 3). HQ had five drastic changes

of attitude, two of which were negative and three –

positive (figure 3).

Figure 2: The evolution of attitudes of Goofy and

Daisy in the smiley version of the game

Another characteristic reaction was to blame

everyone for failing to deliver a packet, as did

Abraham in the textual feedback version (figure 3).

He reacted to the fact that his package was destroyed

by changing his attitude to all other players to

“strong dislike” towards the end of the game. After

realizing that he will not be able to play anymore, he

changed his attitudes to the other players assigning

random values. He commented in the questionnaire

afterwards that he was annoyed with the other

players and didn’t know what he should think about

them in the end of the game. Two players

demonstrated a similar drastic reaction also in the

experiment with the smiley feedback – see the

evolution of Goofy’s and Daisy’s attitudes shown in

A STUDY OF USER ATTITUDE DYNAMICS IN A COMPUTER GAME

227

figure 2. Goofy drastically reducing his attitudes to

all players after a partial success and had to increase

them again (to randomly chosen levels) to be able to

play. Daisy reduced drastically her level of liking to

three other players (Goofy, Mickey and Minnie),

who were among the four most liked players after a

series of consequtive partial deliveries.

Figure 3: The evolution of HQ’s and Abraham’s levels of

attitude towards the other players (textual version)

4.4 Reciprocation

Comparing the evolution of attitudes of two players

towards each other (Figure 4) we see that some of

them follow a pattern of reciprocity, delayed with

several minutes because of the delay in feedback

(only after a round of game the participant can see

the system’s evaluation of the others’ attitudes

towards him/her) and the asynchrnous rounds across

the players. We observed a pronounced difference

between the two versions. To measure the

reciprocation in attitudes between each couple of

players, we mapped the evolution of the mutual

attitudes of every pair of players as shown on Figure

4 and counted the changes in the same direction (i.e.

converging) over the total number of attitude

changes. Applying this measure for each pair of

players, we obtained an average of 43.7% (median

50%) reciprocating changes across the players the

text feedback version and average of 77% (median

73%) of reciprocating changes in the smiley version.

This shows that the smiley feedback visualising the

attitude of the the other players stimulates

significantly more reciprocation expressed in

changing the attitude in the same direction. The

reason is probably that the smiley visualization of

the attitudes of the other players is more intuitive

and requires less cognitive processing, thus allowing

a faster, more spontaneous reaction. In contrast, the

textual version required more cognitive processing

and probably some of the attitudes remained not-

noticed by the players, who focussed their attention

on one or two other players only.

Figure 4: Evolution of the mutual attitudes between two

players in the textual and the smiley version.

5 DISCUSSION AND FUTURE

Even though this experiment is too small to bring

conclusive results, it gives some evidence in support

of our three hypotheses and demonstrates the wealth

of data that can be retrieved from a simple multi-

player game. Our results indicate that individuality

plays an important role in how people change

attitudes in response to events resulting from the

attitudes of other people. Probably people differ

also in the way they assign blame for a situation,

which they can not understand because of the

complex interaction of the factors involved. One

reaction is to blame everyone involved; another – to

blame the closest person involved. Such individual

differences need to be considered when designing

feedback about the actions of other players, for

example, providing less feedback for users who tend

to react drastically or selecting appropriate

visualization to encourage cooperation among users.

In the future we will repeat the experiment analyzing

the data available to each player at each point when

they decide to change attitude and use a think-aloud

protocol.

6 CONCLUSIONS

This paper argues for the importance of considering

interpersonal relationships emerging among the

users of multi-user applications, and for the use of

computer games to investigate emerging user

attitudes towards each other. Interpersonal

relationships among users emerge in any social

system, including those mediated by technology, and

they play an important role in the patterns of

ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING

228

interaction among people. There are not enough

studies of how people actually develop attitudes to

each other in the context of a computer supported

interaction environments and how these attitudes

evolve in time in response to system-mediate events

and realizing others’ attitude towards oneself. The

way the system mediates the user’s perception of

success and failure, as well as the attitudes of other

users influences the way people act. We propose

using specifically designed computer games as tools

to investigate the dynamics of interpersonal attitudes

and we show an example of such a game, together

with the intial experimental results. Clearly, more

work is needed to generate constructive results to

guide system design, and we will be working in

cooperation with social psychologists towards this

goal.

REFERENCES

Artikis, A., Pitt, J., Sergot, M., 2002. Animated

Specification of Computational Societies. Proc.

Autonomous Agents and Multi-Agent Systems

Conference, AAMAS’2002, ACM Press, 1053-1061.

Axelrod, R., 1984. The evolution of cooperation. New

York: Basic Books.

Barros, B., Verdejo, M. F., 2000. Analyzing student

interaction process in order to improve collaboration:

the DEGREE approach. International Journal of AI in

Education, 11, 221-241.

Conte, R., Paolucci, M., 2002. Reputation in Artificial

Societies. Social Beliefs for Social Order, Kluwer.

Golle, Ph., Leyton-Brown, K., Mironov, I., 2001.

Incentives for Sharing in Peer-to-Peer Networks.

Proceedings Electronic Commerce EC'01, October

12-17, 2001, Tampa, Florida, ACM press, 264-267.

Granovetter, M., 1973. The Strength of Weak Ties.

American Journal of Sociology, 78, 1360-80.

Greer, J., McCalla, G., Vassileva, J., Deters, R., Bull, S.,

Kettel, L., 2001. Lessons Learned in Deploying a

Multi-Agent Learning Support System: The I-Help

Experience. Proceedings of AI in Education AIED'01,

San Antonio, IOS Press: Amsterdam, 410-421.

Gutwin, C., Stark, G., Greenberg, S., 1995. Support for

Workspace Awareness in Educational Groupware. In

Proc. CSCL'95, 1st Conference on Computer

Supported Collaborative Learning, 147-156.

Hummel, T., Schoder, D., 1995. Supporting Lateral

Cooperation through CSCW Applications: An

Empirically Motivated Explanatory Approach.

Supplement to the Proceedings of the Fourth

European Conference on Computer Supported

Cooperative Work (ECSCW), 11-15 September 1995,

Stockholm, Sweden

Jermann, P., Soller, A., Muehlenbrock, M., 2001. From

Mirroring to Guiding: A Review of State of the Art

Technology for Supporting Collaborative Learning.

Proceedings of the First European Conference on

Computer-Supported Collaborative Learning,

Maastricht, The Netherlands, 324-331.

Laaksolahti, J., Persson, P., 2001. "Kaktus" project,

http://www.sics.se/humle/projects/Kaktus/

Muehlebrock, M., Hoppe, H.U., 1999. Computer-

Supported Interaction Analysis of Group Problem

Solving. Proceedings Computer Supported

Collaborative Learning Conference,CSCL'99, 398-

405.

Nurmela, K., Lehtinen, E., Palonen, T., 1999. Evaluating

CSCL Log Files by Social Network Analysis. In

Proceedings of the Computer Support for

Collaborative Learning (CSCL) 1999 Conference, C.

Hoadley & J. Roschelle (Eds.) Dec. 12-15, Stanford

University, Palo Alto, California. Mahwah, NJ:

Lawrence Erlbaum Associates.

Prendinger, H., Ishizuka, M., 2002. Evolving Social

Relationships with Animate Characters. Proceedings

of Animating Expressive Characters for Social

Interactions, AISB’02 Convention, London, April.

Rist, T., Schmitt, M., 2002. Avatar Arena: An Attempt to

Apply Socio-Physiological Concepts of Cognitive

Consistency in Avatar-Avatar Negotiation Scenarios.

In Proceedings of Animating Expressive Characters

for Social Interactions, AISB’02 Convention, London.

Shirky, C., 2000. In Praise of Freeloaders. The O'Reilly

Network. Available on line at: <http://www.oreillynet

.com /pub/a/p2p/2000/12/01/shirky_freeloading.html>

Sierra, C., Noriega, P., 2002. Electronic Institutions:

Future Trends and Challenges. Proceeding Workshop

on Cooperative Information Agents CIA’02, Madrid.

Soller, A., 2002. Computational Analysis of Knowledge

Sharing. In Collaborative Distance Learning Doctoral

Dissertation. University of Pittsburgh.

Soller, A., Wiebe, J., Lesgold, A., 2002. A Machine

Learning Approach to Assessing Knowledge Sharing

during Collaborative Learning Activities. Proceedings

of Computer-Support for Collaborative Learning,

CSCL2002, Boulder, CO, 128-137.

Vassileva, J., 2002. Motivating Participation in Peer to

Peer Communities. In P. Petta, S. Ossowski and F.

Zambonelli (eds.) Proceedings of the Workshop on

Emergent Societies in the Agent World, ESAW'02,

Madrid, <http://www.ai.univie.ac.at/%7Epaolo/

conf/esaw02/esaw02accpapers.html >

Yu, B., Singh, M., 2002a. Distributed Reputation

Management for Electronic Commerce.

Computational Intelligence.

Yu, B., Singh, M., 2002b. Emergence of Agent-Based

Referral Networks. Proceedings of Autonomous

Agents and Multi-Agent Systems Conference

AAMAS’02, ACM Press, 1268-1269.

A STUDY OF USER ATTITUDE DYNAMICS IN A COMPUTER GAME

229