Conference PaperPDF Available

Emergent Roles in Decision-Making Tasks using Group Chat



Individuals assume roles in all aspects of life, including in computer-supported collaborative settings. The concept of roles is particularly interesting in settings where no formal roles are defined, as in self-managing virtual teams. In these settings, roles often emerge not only as a result of individual characteristics, but also as participants interact with each other and develop norms of behavior. Using role theory and speech act theory, this study explores the emergence of roles in computer-mediated decision-making groups, using chat transcripts from a lab experiment. Results indicate that four distinct roles emerge as individuals come together in decision-making groups using synchronous computer-mediated communication. These emerging roles have implications for virtual teams in research and practice.
Emergent Roles in Decision-Making Tasks using Group
Jordan B. Barlow
Indiana University
1309 East Tenth Street, BU 670
Bloomington, IN 47405
Individuals assume roles in all aspects of life, including in
computer-supported collaborative settings. The concept of
roles is particularly interesting in settings where no formal
roles are defined, as in self-managing virtual teams. In these
settings, roles often emerge not only as a result of
individual characteristics, but also as participants interact
with each other and develop norms of behavior. Using role
theory and speech act theory, this study explores the
emergence of roles in computer-mediated decision-making
groups, using chat transcripts from a lab experiment.
Results indicate that four distinct roles emerge as
individuals come together in decision-making groups using
synchronous computer-mediated communication. These
emerging roles have implications for virtual teams in
research and practice.
Author Keywords
Computer-mediated communication; virtual teams;
decision-making; roles; role theory; speech acts; speech act
theory; computer-mediated discourse analysis; group
outcomes; cluster analysis.
ACM Classification Keywords
K.4.3 [Organizational Impacts]: Computer-supported
collaborative work
People assume roles in all aspects of life. Roles can be
formal (e.g., manager) or informal (e.g., discussion leader);
they can form naturally (e.g., father) or by appointment
(e.g., professor). Roles are created or emerge in every
situation where humans interact with each other [9, 19].
The concept of role emergence is particularly interesting in
settings where no formal roles are defined. In such
situations, people can fill roles without formal recognition
or even recognizing the roles themselves [19].
One such setting is group decision-making tasks. Roles
often emerge as participants interact with each other and
develop norms of behavior [42]. These roles emerge not
only as a result of differing individual characteristics [e.g.,
51], but also as a result of social interaction within the
group in an effort to facilitate positive outcomes [42].
While past research on emerging roles in face-to-face
decision groups gives some insight into the issue of roles in
decision teams [e.g., 9], research on computer-mediated
teams indicates that members of such teams behave in
different ways than those participating in traditional group
processes [28, 39]. In addition to communication
differences, prior research shows that group tasks requiring
convergence are less likely to be successful virtually than
face-to-face [31].
One way to enrich theoretical and practical understanding
regarding virtual teams is to examine the roles that develop
among such teams. These roles could play an important part
in the behavior and outcomes of such groups [42]. Prior
research on virtual teams shows that the configuration of
individuals in a group influences team outcomes [40]. This
research has shown that the geographic, cultural, and
structural makeup of a team affects group outcomes
including communication effectiveness, coordination, and
performance [27, 28, 40, 51]. Together, this research
indicates that the composition of the team will affect how
the team communicates and performs.
In addition to individual factors (e.g., culture, location,
gender, personality), the makeup of a group in terms of
behavioral roles might affect group outcomes. Group
members have a need for role definition, particularly in
virtual teams where roles tend to be even more ambiguous
[15]. Such role clarity leads to improved team identity [15]
and could also lead to other improved team outcomes. For
example, recent research shows that the emergence of
certain roles in online study groups can ultimately affect the
grades of the group members [11]. Given that roles drive
outcomes in study groups, emerging roles in virtual
decision-making teams could similarly affect group
decisions and other behavioral outcomes such as team
performance, influence, satisfaction, or trust.
Typically, roles are defined as norms or patterns of
behavior [9]. In many computer-mediated environments,
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise,
or republish, to post on servers or to redistribute to lists, requires prior
specific permission and/or a fee.
CSCW ’13, February 2327, 2013, San Antonio, Texas, USA.
Copyright 2013 ACM 978-1-4503-1331-5/13/02...$15.00.
February 23–27, 2013, San Antonio, TX, USA
behavior takes place through written language [23, 29]. In
other words, roles in these settings are based on
communication patterns, such as types of speech acts used,
level of participation in group communication, and who
among participants communicates with others. Such
communication patterns differ depending on situational
variables such as the topic and purpose of the
communication [24]. Thus, while some previous research
has examined role emergence in CMC [3, 16, 20, 35, 44],
none of this research, to my knowledge, has focused on
synchronous online communication nor on virtual teams
working together toward a consensus decision. For
example, many of these studies focus on roles that emerge
in Usenet or Wikipedia. However, individuals
communicating on Usenet or Wikipedia often have
disparate or unfocused goals. The purpose of online
communication in these settings is not necessarily to reach
a consensus on one specific topic. These studies do not
examine nor interpret the roles that emerge as individuals
come together in a common task-focused interaction.
The current study, grounded in role theory and speech act
theory, uses computer-mediated discourse analysis
(CMDA) tools to address the following research question:
What roles, as defined by unique patterns of speech acts
and participation in group communication, emerge in
synchronous computer-mediated decision-making tasks?
The answer to this research question should inform research
on virtual group interventions and outcomes as researchers
and practitioners seek ways to improve decision-making
tasks among computer-mediated teams.
Role Theory
To pursue my research question, I use two complementary
theoretical perspectives. First, I draw from role theory [9].
Role theory defines roles simply as characteristic behavior
patterns [9]. Role theory assumes that people hold social
positions while interacting, and that they hold expectations
for their own and othersbehavior [9]. Further, the concept
of roles provides a means for explaining the relationship
between individuals and social structures by explaining
how social structures are built upon individual roles, or
individual patterns of action [10]. While previous research
on virtual teams has explored the effects of individual
characteristics on team outcomes [e.g., 27, 39, 51], these
studies do not consider the concept of behavioral roles that
emerge during group communication.
Much traditional research on roles focuses exclusively on
leadership roles. The emergence of leadership roles has
been studied in offline groups for decades. Studies by Bales
during the 1950’s examined the types of leaders that
emerge in face-to-face decision-making groups [5]. Bales
concluded that, often, two complementary leaders
emergeone task leader and one social leader. While the
emergence of a strong leader or pair of leaders is
documented in many team settings [e.g., 38, 43, 47, 49], the
present study seeks to explore a variety of roles from a
more general, exploratory perspectiveincluding, but not
limited to, leadership roles.
Several researchers have examined the roles that people
assume as they communicate through computer networks
[3, 16, 20, 35, 44, 45]. These works have identified and
analyzed different types of roles that emerge in
asynchronous communication, including Usenet [20, 44]
and Wikipedia [3, 19, 35]. For example, a myriad of roles
emerge in asynchronous collaborative learning
environments, such as ‘encourager,’ ‘dominator,’ and
‘fellow-traveler [11]. Researchers have used social
network analysis [16] and visualization methods [44] to
understand emerging social roles in Usenet groups such as
‘discussion people’ and ‘answer people.’ Other roles in
Usenet include ‘newbie,’ ‘celebrity,’ ‘elder,’ ‘lurker,’
‘flamer,’ ‘troll,’ and ‘ranter’ [20]. Roles that have emerged
from research on Wikipedia [19, 35, 45] include
‘substantive experts,’ ‘vandal fighters,’ ‘social networkers,’
‘watchdogs,’ ‘starters,’ ‘content justifiers,’ ‘copy editors,’
and ‘cleaners.’
These roles are specific to asynchronous communication
and do not fit well for synchronous decision tasks. For
example, roles such as ‘newbie’ and ‘elder’ refer to people
who are new or well established in the conversation.
However, in synchronous decision tasks, individuals
generally (though not always) start and end the
conversation together. Other roles such as ‘copy editors’
and ‘cleaners’ refer to those who edit the communication of
others, which generally does not occur in synchronous chat.
These examples show that different types of
communication and different types of tasks have different
roles, and the CMC roles that have been defined thus far
likely are not the best fit for synchronous CMC tasks such
as decision-making.
Speech Act Theory
In addition to the role theory perspective as used by these
researchers, I draw from an additional theory base,
particularly suited to synchronous communication, to
explore emerging roles in decision tasksspeech act
theory. Speech act theory [4] posits that speech uttered in a
particular way constitutes an act [29, 41]. Speech acts do
not simply describe the world—they “bring about change to
the world” [2: p. 252]. Speech act theory is particularly
suited to virtual communication because in many computer-
mediated environments, behavior is performed primarily
(and sometimes only) through verbal language [23, 29]. In
other words, to understand the behavior of individuals in
mediated teams, one must understand their speech acts.
Surprisingly, no prior studies on role emergence in CMC
[e.g., 3, 16, 20, 35, 44], to my knowledge, are grounded in
speech act theory. Hence, they do not focus heavily on
February 23–27, 2013, San Antonio, TX, USA
message content to understand roles, but rather focus on
how much and with whom individuals communicate.
However, according to speech act theory, behavior can only
be truly understood by understanding the meaning of
communicative acts. Further, these studies tend to focus
more on social roles than task roles [16] by focusing solely
on participation and interaction patterns rather than
meaning of the messages. In order to understand task-based
roles, such as those that should emerge in decision-making
tasks in CMC, one must understand the speech acts
performed by communicating individuals.
Using these perspectives from role theory and speech act
theory, I propose that roles in virtual group decision-
making tasks can be detected by studying the speech acts of
individuals participating in the task.
To explore emerging roles in synchronous decision tasks, I
used CMDA methods to analyze 26 online group chat
transcripts taken from a lab experiment. The experiment
was originally conducted by the author and other colleagues
to investigate the effects of priming on group decision-
making tasks [7]; analysis of the participants’
communication was not done for that study.
The participants in the lab experiment were 130
undergraduate business students (80 male, 49 female, one
undisclosed gender) working in 26 groups of five. All
students received extra credit for participating in the study.
Participants were instructed to make university admissions
decisions regarding a set of five possible candidates. To
encourage discussion among team members, participants
were not allowed to admit more than three of the five
candidates. This task has been used in previous research on
decision-making in virtual teams [22].
The admissions task is a hidden profile task, meaning the
information given to each individual consists of both
common information, known to all participants, and unique
information, known to only one participant on each team.
As a result, hidden profile tasks encourage team discussion.
Hidden profile tasks are similar to the tasks real groups
face; in real groups, not all participants have complete
information to make the best decision individually.
Experiment Procedures
The experiment began with participants playing a priming
game [13] for eight minutes. In one condition, participants
were primed with achievement and attention words. The
remaining participants were given neutral words. For more
discussion on priming in virtual settings, the reader is
referred to [13]. The priming did not have an effect on
group decision quality.
Participants were then provided four minutes to read the
candidate information and make an individual admissions
decision. Next, participants worked together as a team for
15 minutes using Gmail Chat to arrive at a team decision.
Pseudonyms were used in Gmail Chat so that participants
did not know the identity of their team members or which
people in the room were part of their group. The
participants sat in separate cubicles in front of individual
personal computers during the study so no audible
communication occurred. Finally, participants were asked
to fill out a survey indicating their team’s decision and
individual demographic information.
CMDA Procedures
At the end of the experiment, I downloaded the transcripts
of each session from the Gmail accounts that were created
for the experiment. The 26 transcripts contained 2996
messages, with an average of 115.23 messages per
CMDA methods were used to code and analyze the data.
CMDA is a collection of discourse analysis methods
tailored specifically for computer-mediated communication
[23]. CMDA methods are particularly suited to my research
question because roles reflect patterns of behavior, and
behavior in virtual teams is reflected entirely by the way
group members communicate [23]. Text analysis methods
and other methods based on speech act theory are
increasingly used by information systems researchers in a
variety of contexts [1, 2, 29]. Abbasi and Chen [1] note that
“CMC text analysis is important for evaluating the
effectiveness and efficiency of electronic communication in
various organizational settings, including virtual teams and
group support systems” (p. 812). To explore the emerging
roles of the participants of the lab experiment, I used two
CMDA methodsparticipation analysis and speech act
analysisto measure how much and in what ways
participants communicated.
Participation analysis is a common CMDA method used to
detect roles in asynchronous CMC [e.g., 16, 44]. For
example, work exploring collaboration in Wikipedia largely
defines the roles ‘zealot’ and ‘good Samaritan’ based on the
amount of participation in editing Wikipedia articles [3].
Participation analysis has also been used in studies of
emergent leadership in virtual teams [43, 47, 49]. In this
research, I analyze the number of messages by participant,
the percentage of the team’s messages by each participant,
and the length (in words) of each message.
Previous research on roles in CMC generally combines
participation analysis with another type of analysis
typically social network analysis because of the
asynchronous nature of the studies [e.g., 19]. I chose to use
speech act analysis, a better fit with the theoretical
reasoning and methodological design of my research.
Speech act analysis has been used extensively in linguistics
to analyze the meaning of utterances in speech and writing
February 23–27, 2013, San Antonio, TX, USA
[18, 37]. In speech act analysis, a technique based on
speech act theory [4, 41], researchers read every utterance
in a communication transcript and assign it to a speech act
category describing the behavior portrayed by the utterance
(e.g., “claim”, “question”, “reaction”). To apply speech act
analysis, several taxonomies have been created by linguists
[e.g., 17]. These taxonomies have been used to examine
computer-mediated communication as well; for example,
researchers have applied the Francis and Hunston [17]
taxonomy to understand the seriousness of Internet Relay
Chat [26].
However, due to challenges of previous speech act
taxonomies, such as difficulty in distinguishing between
closely related codes, missing codes, overlapping codes,
and codes specific to certain types of datasets [c.f. 29],
Herring et al. [25] developed a consolidated and simplified
taxonomy specifically for the analysis of CMC. This CMC
act taxonomy, which consists of 16 types of speech acts,
was tested by Herring and her colleagues on blog posts,
threaded bulletin boards, and synchronous chat. They found
it to be easily applied and interpreted. The CMC act
taxonomy has been used in previous research on computer-
mediated communication [30].
I classified each of the 2996 messages of the sample
according to its relevant speech act type using the CMC act
taxonomy. Some messages contained multiple speech acts;
these were divided and coded as separate messages,
resulting in a classification of 3055 total speech acts.
Because speech act analysis can be subjective, a second
coder participated in the analysis to establish inter-rater
reliability. The second coder and I jointly coded one
transcript of 170 speech acts to ensure mutual
understanding of the classification scheme. Next, we
independently coded four more transcripts (458 speech acts,
about 15 percent of the total sample), and obtained a
Cohen’s kappa reliability score of 0.708, indicating
substantial agreement between raters [33]. The rest of the
messages were then coded solely by the author.
Participation and Speech Act Analyses
Summary statistics for the participation analysis (shown in
Table 1) show that the messages in this type of conversation
are typically short, with one-third of messages being only
one or two words in length.
Average message length in words
Average number of messages per participant
1-word messages as % of total messages
1- or 2-word messages of % of total messages
Table 1. Participation summary statistics.
Speech Act Category
Others (< 1%)
Table 2. Speech act summary statistics
In addition to these overall summary statistics, participation
statistics were calculated for each participant of the
experiment in order to compare amounts of participation as
an indicator of behavioral patterns.
The overall results from the speech act analysis (shown in
Table 2) indicate that most of the acts performed by
participants through their communication were sharing
information, making claims about information, agreeing
with others, and asking questions. These speech acts are
consistent with the type of task completed by the groupa
decision-making task. While this research focuses on such
decision-making groups, research on different types of
group tasks (e.g., brainstorming, planning) would likely
result in differences in the types of speech acts used and
length of messaged communicated by participants [24].
Cluster Analysis
Based on the summary results of the participation and
speech act analyses and a qualitative analysis of the text, I
determined a set of six relevant factors of role behavior to
be used in a cluster analysis to explore roles: two
participation factors and four speech act factors.
Though I collected three participation analysis metrics, I
chose to use just two as factors in the cluster analysis:
percentage of team messages contributed by an individual,
and average message length. I chose not to use number of
individual messages because this metric is redundant with
percentage of team messages, which also accounts for
between-group differences in the number of messages
The four speech act factors used were the following:
information shared (percentage of an individual’s speech
acts classified as ‘inform’ speech acts), opinions shared
February 23–27, 2013, San Antonio, TX, USA
(percentage of speech acts classified as ‘claim’ speech
acts), agreement with others (percentage of speech acts
classified as ‘accept’ speech acts), and amount of
discussion guiding (percentage of speech acts classified as
‘inquire,’ ‘manage, ‘request,’ or ‘direct’ speech acts). I
combined the ‘inquire,’ ‘manage,’ ‘request,’ and ‘direct’
speech acts because these were strikingly similar upon
reviewing them in the transcripts. Each of these four factors
accounted for over 10 percent of the overall number of
speech acts, respectively.
I then performed a cluster analysis based on the six factors,
as done in previous CMC role research [35] and virtual
team leadership research [43]. For example, Liu and Ram
[35] used cluster analysis to define roles in Wikipedia
collaboration; the factors for the cluster analysis were
behaviors exhibited by the Wikipedia users (e.g., sentence
insertions, link modifications, etc.). In contrast, my cluster
analysis used factors based on levels of participation and
types of speech acts, as these are the most relevant
indicators of behavior in synchronous virtual
I first used Ward’s method of hierarchical clustering to
identify the proper number of clusters. The results
suggested four distinct groups of participants in the sample
based on the six factors. I then used non-hierarchical (k-
means) clustering to refine the original clusters. No cluster
contained all five members of any group, indicating that the
clusters do not reflect characteristics of teams, but rather
individuals. The means of each of the six factors for each of
the four clusters are shown in Table 3. Distinguishing
values for each cluster are highlighted in light grey (low
values) and dark grey (high values).
Table 4 provides a summary description of the four clusters.
A qualitative reading of the transcripts confirms the
emergence of these four roles as group members interacted
with each other. Again, these roles are specific to a
decision-making task.
The largest cluster was the cluster of organizers, those who
directed the conversation of the group by asking questions
and requesting information. The organizers tended to have
more than the average number of messages, with moderate
amounts of sharing, claiming, and agreeing. Often, teams
had more than one organizer as individuals worked together
to move the discussion forward in a limited period. In fact,
it was not uncommon for groups to have 3-4 organizers.
The next largest group is the listeners, those who
participated less in discussion and primarily agreed with
others. While the listeners tended to share some information
and make claims when they did speak, these individuals did
not guide the discussion.
% of team
% of acts
classified as
% of acts
classified as
% of acts
classified as
% of acts
classified as
Table 3. Cluster analysis results. Distinctly high values in dark gray; distinctly low value in light gray.
Distinguishing characteristics
Role name
# individuals
Mostly share information and agree with others; little guiding of discussion;
few, short messages and few claims about information.
24 (18.5%)
Few messages, mostly agreeing with others; little guiding of discussion; average
amounts of sharing and interpreting information.
37 (28.5%)
Mostly guide the discussion by managing or asking questions; above average
amounts of sharing information, making claims, and agreeing with others.
44 (33.8%)
Mostly make claims, but do not share many facts; little agreement with others;
many messages, often long, but only moderate discussion guiding.
25 (19.2%)
Table 4. Interpretation of clusters / roles.
February 23–27, 2013, San Antonio, TX, USA
Fewer individuals filled the role of sharers, those who
mostly shared facts about the candidates but did not
interject many strong opinions. These individuals answered
questions and shared what they knew so that the team
would have enough information to make a good decision.
However, they had fewer, shorter messages and usually did
not propose solutions.
The last emerging role was that of opinionaters, those who
primarily tried to convince others of their own opinions.
Most of the individuals in this role shared a variety of
opinions, but were slow to accept the opinions of others,
some even showing high tendencies toward stubbornness in
the transcripts. These individuals used many messages,
often long messages, to try to persuade their teammates of
their opinions.
The roles detected through this cluster analysis reflect the
roles that emerge in a decision-making task by groups
communicating through chat software. Different roles
would likely emerge in situations where a different type of
task or technology was used.
The results show that distinct roles emerge in synchronous
CMC, likely as a result of individual characteristics, social
interaction, and group needs [42]. These roles emerged in
text-based communication; similar roles may emerge in
other types of CMC, though the results cannot be perfectly
generalized from this study. However, the techniques
proposed by my research to detect roles could be used in
future studies examining role emergence in other computer-
mediated groups using a different technology or performing
a different type of task.
The results of this research should be considered in light of
some limitations. First, the transcripts for this study were
taken from a lab experiment using student participants.
While the results are useful for a variety of settings,
including student groups, careful consideration should be
used in applying the results to other contexts where
individuals have different characteristics than students.
Because roles and behavior are dependent on a multitude of
factors, including gender, education, and culture [27], roles
may differ in other contexts outside of student groups.
However, because the results demonstrate general human
behavior, similar results could potentially hold in other
Care should also be taken in generalizing the results to
other types of CMC or other types of group tasksthis
study only examined text-based chat for decision-making
groups. Other types of communication technologies that
more closely resemble face-to-face communication may
have different results, especially for different types of tasks.
However, a main contribution of this research is a
demonstration of an appropriate technique for detecting
roles in CMC. This technique can be applied to CMC in
other contexts in future research.
Another limitation of the study is the small sample size.
Because the data for the research were taken from a
previous study, more data could not be collected. Using a
larger sample, more research should further examine and
verify the roles in synchronous CMC and then investigate
how these roles truly affect group outcomes and
interventions. Notwithstanding, this study is an important
contribution in that role emergence was shown, and a
pattern of investigating roles in synchronous CMC was set
Finally, this study only examines roles in groups of
individuals who are interacting for the first time and have
no anticipation of interacting together in the future.
However, research shows that both individuals and groups
adapt to computer-mediated communication over time and
shift behaviors to meet the unique demands of CMC [36,
48]. Future research should examine whether and how roles
in synchronous decision tasks persist or change as groups
adapt over time.
Interpretation of Findings
Interpretation of Individual Roles
Each of the roles detected in this study reflects distinct
patterns of behavior exhibited by participants during a
decision-making process in a text-based communication
The most common role in the groups was the organizer.
The organizers in a group were those who showed
leadership behaviors as they directed the discussion, asked
questions, and structured the communication. Often, teams
would have more than one organizer as individuals worked
together to move the discussion forward in a limited period
of time. Some groups even had three or four organizers.
This complex pattern of leading the discussion and
communication is consistent with prior CMC research
suggesting that a simple pattern of single or complementary
leadership does not always hold in online contexts [e.g., 11,
19, 50] and emergent leadership patterns are different in
virtual teams than in face-to-face teams [38]. Further
research supports the notion that influence and leadership
behaviors are displayed by individuals of all social
positions [50]. In online contexts, even when task-related,
the simple emergence of one or even two distinct leaders is
not a given [38]; rather, individuals participating in CMC
show more complex patterns of leading and following by
assuming different types of roles.
Synchronous online decision-making groups also tend to
have many listeners, those who were mostly passive and
agreeing with others. This finding is consistent with
research showing that social loafing tends to emerge in
groups larger than two to three individuals [12]. These
listeners were generally doing more than just lurking; when
February 23–27, 2013, San Antonio, TX, USA
asked for information or an opinion they would share it,
often elaboratinglisteners tend to share longer messages
than others. However, the amount of participation of these
individuals, along with the high amount of agreement,
showed a mostly passive role. Possible reasons that an
individual would assume a listener role include a quiet
personality, or an interest in the discussion while others
dominate as organizers or opinionaters. In other words, for
some listeners, it may be the case that they shared more
messages at the beginning, but the percentage of team
messages typed decreased as others in the group dominated
and the listeners resorted to mostly agreement statements.
Further, it may be that listeners type longer messages than
others in an effort to share more information and opinions
at one time to compensate for a lack of consistent active
speaking during the discussion.
Fewer individuals assumed the role of sharers, those who
simply shared facts but did not attempt to widely influence
the discussion. Like listeners, sharers fell into a largely
passive role. Sharers typed sparse, short messages, often in
response to a question put forth by an organizer. Sharers
avoid giving their opinions and only respond when needed.
While all roles are affected by individual characteristics,
possible reasons that people assume a sharer role during the
task could be lack of care about the task, or too many other
people in the group dominating the discussion.
Opinionaters were those whose primary goal was to
convince others of their own opinions. Most of the
individuals in this role shared a variety of opinions, but
were slow to accept the opinions of others, some even
showing high tendencies toward stubbornness in the
transcripts. While the number of speech acts categorized as
‘reject’ was low in the overall sample, speech acts
connoting a rejection of others’ ideas was highest in this
cluster. Further, opinionaters shared relatively low amounts
of factual information with the team, relying on opinion
more than evidence to argue a case.
Interpretation of Role Structure
A variety of all four roles emerged in groups at all levels of
decision quality; that is, teams who performed better on the
task did not have a unique role structure. The emergence of
roles is affected by individual characteristics, but the
transcripts indicated that in some cases, people assumed
roles as a result of the group interaction. For example, if a
group happened to have many organizers or opinionaters
dominating the discussion, others who may have assumed
these roles became more passive and assumed a listener or
sharer role instead.
The roles are also dependent on the task that was used. For
example, if the task had been more emotional than
university admissions, there may have been more
opinionaters. If the task had focused more on encouraging
participations, there might have been fewer listeners.
Different types of tasks such as brainstorming or planning,
rather than decision-making, could result in different types
of roles altogether. The interpretation of these roles is based
on the decision-making task. However, the methods used to
understand the roles can be applied to a wide variety of
Practical Implications
The results of this study have implications for both research
and practice. Awareness of the unique task-based roles that
emerge during decision-making tasks using synchronous
text-based CMC is important for groups that commonly use
technology for such tasks. Understanding the types of
behaviors that emerge in these settings is useful to better
structure and facilitate communication among individuals.
While determining exactly which roles are present in a
given group is currently a manually intensive process that
occurs after the fact, groups can be informed that, generally
speaking, the four roles presented here are common in
decision-making groups.
Research on Role Outcomes
This research also has implications for research on the
relationship of roles to group outcomes. Previous research
indicates that inclusion or exclusion of certain types of
individuals can affect team processes and outcomes [21].
However, until recently, little research was done in
exploring relationships between roles and team-level
outcomes [42], and scholars have called for more research
examining how the presence or absence of roles affects
group outcomes such as productivity and longevity [19].
For example, the presence of certain roles in collaborative
learning environments affects group project grades [11]; the
quality of a Wikipedia article depends on the roles of the
contributors and how they collaborate with each other [35].
Such relationships between roles and group outcomes
should be further examined for virtual teams in decision-
making tasks. Having clarity of roles in virtual teams will
lead to improved team identity [15] and other group
Several virtual group decision-making outcomes could be
affected by the emergence of roles. For example, roles may
affect the extent to which individuals change their decisions
to conform to the group. Because individuals in different
roles communicate (i.e., behave) differently, the behavior of
some individuals may be more influential than the behavior
of others. In this case, the sharers and listeners of the
groups tend to be more passive than other roles. Because
individuals in these roles largely agree with others without
interjecting strong opinions, they could be more likely to
change their original decisions to match the opinions of
others in the group as they give up their own opinions to
agree with the statements of others. Future research should
more thoroughly examine what types of group outcomes are
most influenced by these roles. Further, if roles affect group
outcomes, and group performance is correlated with
February 23–27, 2013, San Antonio, TX, USA
individual characteristics such as gender, educational
background, and culture [e.g., 27, 46, 51], behavioral roles
may act as a mediator between individual characteristics
and group outcomes. Future research should further explore
this relationship.
Research on Factors Influencing Role Formation
Next, if roles affect group outcomes, it may be important to
shape role formation in group to achieve better outcomes of
group decision-making. Traditional role research has
viewed roles as somewhat static within individuals. For
example, researchers argue that every user in online Q&A
services behaves consistently and uses patterns of
communication regardless of circumstances [6]. Indeed,
roles are known to be influenced by individual
characteristics [42]. However, research also shows that
individual roles are adapted in real-time depending on the
context and situational demands [42]. This opens up the
possibility that individuals and organizations can use
interventions to change the emerging roles in group
contexts. In the context of synchronous CMC decision-
making tasks, some interventions, whether intentional or
resulting from context, could change the way that groups
interact together and adapt their individual roles.
Researchers in the computer-supported collaboration and
human-computer interaction fields have developed
technological interventions to improve group collaboration
[e.g., 8, 14, 32, 34], and more research is needed to
understand how such interventions could affect role
emergence in groups. For example, collaboration
researchers developed a system that allows group members
to give anonymous feedback in a backchannel in addition to
the regular communication of the group [8]. This
backchannel communication allows for more contributions
from participants who may otherwise participate less. A
group using such a system may have fewer people
assuming a ‘listener’ role. Other researchers have proposed
real-time systems that summarize individual participation to
the group by highlighting the number of comments by
individuals or summarizing contributions with keywords
[14, 32]. These systems give feedback while groups are
communicating, and have been shown to change group
communication styles and facilitate cooperation among
group members [32]. Technological interventions have also
been designed to give real-time feedback to group members
about the type of language they use during group discussion
[34]. These intervention systems have the power to affect
the roles that emerge during computer-mediated decision
tasks. Future research should connect these two streams of
researchwhile work on group intervention systems shows
improvements in group processes, more research is needed
to better understand exactly how such interventions affect
the roles of individuals in the group.
In addition to technological interventions, other factors,
such as unrelated interactions with other people before the
meeting, could subconsciously encourage team members to
listen and agree rather than interject and actively
participate. For example, in the lab experiment from which
transcripts were used for this study [7], individuals were
primed with attention and achievement words in an effort to
change the behavior of individuals to pay more attention to
other group members. Unfortunately, the sample size of the
data is too small to test whether the priming effect changed
which roles emerged in the teams (e.g., whether the priming
caused more listeners to emerge). Future studies should
examine direct and indirect effects (such as priming) on
role emergence.
Research Methods for Detecting Roles
Finally, this study has implications for the methods in
carrying out further research on roles in virtual settings.
This research successfully used speech act analysis, a form
of CMDA, to detect and interpret roles in synchronous
computer-mediated communication. Further research
should build on this work by continuing to use speech act
analysis and other CMDA methods that are theoretically
related to the research question regarding roles in online
communication. A study of speech acts directly reflects
participant behavior in computer-mediated settings, and
such techniques should be applied in studies of online
behavior. While the current study examines roles for
decision-making tasks, the procedures used can be applied
to a variety of CMC settings.
The author sincerely thanks Alan Dennis and Valerie
Bartelt, who designed the lab experiment; Lingyao Yuan
and Rosh Dhanawade, who helped in proctoring the
experiment; Kayleen Barlow, the second coder of the
speech act analysis; and Susan Herring and David Wilson,
who provided helpful feedback on previous versions of this
1. Abbasi, A. and Chen, H. CyberGate: A design
framework and system for text analysis of computer-
mediated communication. MIS Quarterly 32, 4 (2008),
2. Ågerfalk, P.J. Getting pragmatic. European Journal of
Information Systems 19, 3 (2010), 251-256.
3. Anthony, D., Smith, S.W., and Williamson, T.
Reputation and reliability in collective goods: The case
of the online encyclopedia Wikipedia. Rationality and
Society 21, 3 (2009), 283-306.
4. Austin, J.L. How to Do Things with Words. Harvard
University Press (1962), Cambridge, MA, USA.
5. Bales, R.F. Task roles and social roles in problem-
solving groups. In Role Theory: Concepts and Research,
February 23–27, 2013, San Antonio, TX, USA
B.J. Biddle and E.J. Thomas Eds. John Wiley & Sons,
Inc. (1966), New York City, NY, USA, 254-263.
6. Barash, V., Smith, M.A., Getoor, L., and Welser, H.T.
Distinguishing knowledge vs social capital in social
media with roles and context. In Proc. ICWSM 2009,
AAAI Press (2009), 1-4.
7. Bartelt, V., Dennis, A.R., Yuan, L.I., and Barlow, J.B.
Individual priming in virtual team decision-making.
Group Decision and Negotiation (accepted pending
minor revisions).
8. Bergstrom, T. and Karahalios, K. Vote and be heard:
Adding back-channel signals to social mirrors. In
INTERACT 2009, International Federation for
Information Processing (2009), 546-559.
9. Biddle, B.J. Recent developments in role theory. Annual
Review of Sociology 12 (1986), 67-92.
10. Callero, P.L. From role-playing to role-using:
Understanding role as resource. Social Psychology
Quarterly 57, 3 (1994), 228-243.
11. Chang, C.-K., Chen, G.-D., and Wang, C.-Y. Statistical
model for predicting roles and effects in learning
community. Behaviour & Information Technology 30, 1
(2011), 101-111.
12. Chidambaram, L. and Tung, L.L. Is out of sight, out of
mind? An empirical study of social loafing in
technology-supported groups. Information Systems
Research 16, 2 (2005), 149-168.
13. Dennis, A.R., Minas, R.K., and Bhagwatwar, A.
Sparking creativity: Improving electronic brainstorming
with individual cognitive priming. In Proc. 45th HICSS,
Computer Society Press (2012), 139-148.
14. Dimicco, J.M. and Bender, W. Second messenger:
Increasing the visibility of minority viewpoints with a
face-to-face collaboration tool. In Proc. 9th Conf. IUI,
(2004), 232-234.
15. Fiol, C.M. and O'connor, E.J. Identification in face-to-
face, hybrid, and pure virtual teams: Untangling the
contradictions. Organization Science 16, 1 (2005), 19-
16. Fisher, D., Smith, M.A., and Welser, H.T. You are who
you talk to: Detecting roles in Usenet newsgroups. In
Proc. 39th HICSS, Computer Society Press (2006), 1-
17. Francis, G. and Hunston, S. Analysing everyday
conversation. In Advances in Spoken Discourse
Analysis, M. Coulthard Ed. Routledge (1992), London,
UK, 1-34.
18. Fraser, B. Hedged performatives. In Speech Acts: Syntax
and Semantics, P. Cole and J.L. Morgan Eds. Academic
Press (1975), New York, NY, USA, 187-210.
19. Gleave, E., Welser, H.T., Lento, T.M., and Smith, M.A.
A conceptual and operational definition of 'social role' in
online community. In Proc. 43rd HICSS, Computer
Society Press (2009), 1-11.
20. Golder, S.A. and Donath, J. Social roles in electronic
communities. In AoIR Conf. Internet Research 5.0,
(2004), 1-25.
21. Guzzo, R.A. and Dickson, M.W. Teams in
organizations: Recent research on performance and
effectiveness. Annual Review of Psychology 47 (1996),
22. Heninger, W.G., Dennis, A.R., and Hilmer, K.M.
Individual cognition and dual-task interference in group
support systems. Information Systems Research 17, 4
(2006), 415-424.
23. Herring, S.C. Computer-mediated discourse analysis:
An approach to researching online behavior. In
Designing for virtual communities in the service of
learning, S.A. Barab, R. Kling, and J.H. Gray Eds.
Cambridge University Press (2004), New York, NY,
USA, 338-376.
24. Herring, S.C. A faceted classification scheme for
computer-mediated discourse. Language@Internet 4, 1
(2007), 1-19.
25. Herring, S.C., Das, A., and Penumarthy, S., 2005. CMC
act taxonomy.
26. Herring, S.C. and Nix, C.G. Is "serious chat" an
oxymoron? Pedagogical vs. social uses of Internet Relay
Chat. In AAAL Conf., (1997), 1-19.
27. Hinds, P., Liu, L., and Lyon, J. Putting the global in
global work: An intercultural lens on the practice of
cross-national collaboration. The Academy of
Management Annals 5, 1 (2011), 135-188.
28. Hinds, P.J. and Bailey, D.E. Out of sight, out of sync:
Understanding conflict in distributed teams.
Organization Science 14, 6 (2003), 615-632.
29. Janson, M.A. and Woo, C.C. A speech act lexicon: An
alternative use of speech act theory in information
systems. Information Systems Journal 6, 4 (1996), 301-
30. Kapidzic, S. and Herring, S.C. Gender, communication,
and self-presentation in teen chatrooms revisited: Have
patterns changed? Journal of Computer-Mediated
Communication 17, 1 (2011), 39-59.
31. Kerr, D.S. and Murthy, U.S. Divergent and convergent
idea generation in teams: A comparison of computer-
mediated and face-to-face communication. Group
Decision and Negotiation 13, 4 (2004), 381-399.
32. Kim, T., Hinds, P., and Pentland, A. Awareness as an
antidote to distance: Making distributed groups
cooperative and consistent. In 15th ACM Conf. CSCW,
ACM (2012), 1237-1246.
February 23–27, 2013, San Antonio, TX, USA
33. Landis, J.R. and Koch, G.G. The measurement of
observer agreement for categorical data. Biometrics 33,
1 (1977), 159-174.
34. Leshed, G., Perez, D., Hancock, J.T., Cosley, D.,
Birnholtz, J., Lee, S., Mcleod, P.L., and Gay, G.
Visualizing real-time language-based feedback on
teamwork behavior in computer-mediated groups. In
CHI 2009, ACM (2009), 537-546.
35. Liu, J. and Ram, S. Who does what: Collaboration
patterns in the Wikipedia and their impact on article
quality. ACM Transactions on Management Information
Systems 2, 2 (2011), article 11.
36. Majchrzak, A., Rice, R., Malhotra, A., King, N., and Ba,
S. Technology adaptation: The case of a computer-
supported inter-organizational virtual team. MIS
Quarterly 24, 4 (2000), 569-600.
37. Mclaughlin, M. Chapter 4: Acts. In Conversation: How
Talk is Organized Sage (1984), Thousand Oaks, CA,
USA, 133-168.
38. Misiolek, N.I. and Heckman, R. Patterns of emergent
leadership in virtual teams. In Proc. 38th HICSS, IEEE
Press (2005), 1-10.
39. Mortensen, M. and Hinds, P.J. Conflict and shared
identity in geographically distributed teams.
International Journal of Conflict Management 12, 3
(2001), 212-238.
40. O'Leary, M.B. and Mortensen, M. Go (con)figure:
Subgroups, imbalance, and isolates in geographically
dispersed teams. Organization Science 21, 1 (2010),
41. Searle, J.R. A taxonomy of illocutionary acts. In
Expression and Meaning: Studies in the Theory of
Speech Acts, J.R. Searle Ed. Cambridge University Press
(1979), Cambridge, UK, 1-29.
42. Stewart, G.L., Fulmer, I.S., and Barrick, M.R. An
exploration of member roles as a multilevel linking
mechanism for individual traits and team outcomes.
Personnel Psychology 58, 2 (2005), 343-365.
43. Sudweeks, F. and Simoff, S. Leading conversations:
Communication behaviours of emergent leaders in
virtual teams. In Proc. 38th HICSS, IEEE Press (2005),
44. Welser, H.T., Gleave, E., Fisher, D., and Smith, M.A.
Visualizing the signatures of social roles in online
discussion groups. Journal of Social Structure 8, 2
(2007), 1-32.
45. Welser, H.T., Kossinets, G., Smith, M.A., and Cosley,
D. Finding social roles in Wikipedia. In Annual Meeting
of the ASA, AllAcademic (2008), 1-8.
46. Woolley, A.W., Chabris, C.F., Pentland, A., Hashmi, N.,
and Malone, T.W. Evidence for a collective intelligence
factor in the performance of human groups. Science 330,
29 Oct 2010 (2010), 686-688.
47. Yoo, Y. and Alavi, M. Emergent leadership in virtual
teams: What do emergent leaders do? Information and
Organization 14, 1 (2004), 27-58.
48. Zhang, P. and Sun, H. The complexity of different types
of attitudes in initial and continued ICT use. Journal of
the American Society for Information Science and
Technology 60, 10 (2009), 2048-2063.
49. Zhang, S. and Fjermestad, J. Bridging the gap between
traditional leadership theories and virtual team
leadership. International Journal of Technology, Policy
and Management 6, 3 (2006), 274-291.
50. Zhu, H., Kraut, R., and Kittur, A. Effectiveness of
shared leadership in online communities. In 15th ACM
Conf. CSCW, (2012), 1-10.
51. Zolin, R., Hinds, P., Fruchter, R., and Levitt, R.E.
Interpersonal trust in cross-functional, geographically
distributed work: A longitudinal study. Information and
Organization 14, 1 (2004), 1-2.
February 23–27, 2013, San Antonio, TX, USA
... The social processes associated with collaboration technology and how individual team members interact can influence team performance [24,40,53]. Collaboration technology research has only recently begun to consider how deeper-level individual characteristics of team members [105] such as intelligence [6] and personality [120] affect team processes and outcomes. However, research in CMC has examined intelligence and personality separately. ...
Full-text available
Teams have increasingly turned to computer-mediated communication (CMC) to work when team members cannot all be in the same physical space at the same time, leading to the need to better understand what influences group performance in these settings. We know that team member intelligence and personality affect team performance when teams work face-to-face, but their effects are not yet clear when teams use text-based CMC, which has different characteristics than face-to-face communication. We conducted a laboratory study of 61 teams working on a decision-making task using text-based CMC. We found that team mean extraversion had a large negative effect, and team mean neuroticism had a medium-sized negative effect on team performance. Team mean intelligence had no effect. We recommend that managers consider the effects of extraversion when selecting team members and focus on selecting more introverted team members if the team is likely to extensively use text-based CMC. Likewise, managers should consider extraversion when designing teamwork processes for virtual teams; if a team has many members who are high in extraversion, the team should use text-based CMC sparingly. We also recommend that researchers use extraversion as a control factor in future research studying text-based CMC because extraversion has a large effect on team outcomes and, left uncontrolled, could increase unexplained error variance and overshadow the focus of the research study.
... In HCI, there has been a sizable literature examining human social roles in various sociotechnical contexts. Previous work has examined social roles in family life [5,12,73], the workplace [14] and online support communities [159], and researchers have investigated gender roles regarding emerging technologies [106,134,160]. Research in this area has been marshaled toward the automated recognition of social roles, for example in the workplace [7,29,76,155], in teaching [65], and across contexts [6,7,105]. ...
Full-text available
Out of the three major approaches to ethics, virtue ethics is uniquely well suited as a moral guide in the digital age, given the pace of sociotechnical change and the complexity of society. Virtue ethics focuses on the traits, situations and actions of moral agents, rather than on rules (as in deontology) or outcomes (consequentialism). Even as interest in ethics has grown within HCI, there has been little engagement with virtue ethics. To address this lacuna and demonstrate further opportunities for ethical design, this paper provides an overview of virtue ethics for application in HCI. It reviews existing HCI work engaging with virtue ethics, provides a primer on virtue ethics to correct widespread misapprehensions within HCI, and presents a deductive literature review illustrating how existing lines of HCI research resonate with the practices of virtue cultivation, paving the way for further work in virtue-oriented design.
... Previous work has explored models and techniques to identify and support different roles in ICTs of various application domains, including collaboration tools [4], access control systems [5], knowledge co-production platforms [3], and software engineering tools [1], [27]. For example, Arazy et al. [3] identified seven roles of Wikipedia contributors, such as all-round contributors and layout shapers, through a clustering analysis of user actions in Wikipedia articles. ...
... Previous work has explored models and techniques to identify and support different roles in ICTs of various application domains, including collaboration tools [4], access control systems [5], knowledge co-production platforms [3], and software engineering tools [1], [27]. For example, Arazy et al. [3] identified seven roles of Wikipedia contributors, such as all-round contributors and layout shapers, through a clustering analysis of user actions in Wikipedia articles. ...
Full-text available
Contributors to open source software (OSS) communities assume diverse roles to take different responsibilities. One major limitation of the current OSS tools and platforms is that they provide a uniform user interface regardless of the activities performed by the various types of contributors. This paper serves as a non-trivial first step towards resolving this challenge by demonstrating a methodology and establishing knowledge to understand how the contributors' roles and their dynamics, reflected in the activities contributors perform, are exhibited in OSS communities. Based on an analysis of user action data from 29 GitHub projects, we extracted six activities that distinguished four Active roles and five Supporting roles of OSS contributors, as well as patterns in role changes. Through the lens of the Activity Theory, these findings provided rich design guidelines for OSS tools to support diverse contributor roles.
... This is done by coding based on a definite interpretation after identifying when a conversation occurred and who spoke, temporal sequence of speaking, and orientation of an utterance. In this method, decision making or informal communication are studied in monolingual chat as a computer-mediated communication tools [7], [8], [9]. On the other hand, in the case of mistranslations in a multilingual environment, it is more difficult to identify block of speech and intention than in a monolingual environment. ...
Conference Paper
Full-text available
Due to globalization, cross-lingual communicating using machine translation has been becoming widespread. However, machine translation results in mistranslations. When collaborating on the Internet, it may be more difficult to effectively communicate with others than face-to-face as it may be difficult to get a handle on the context of the conversation. We analyzed chat data in a machine-translation-mediated multilingual gaming simulation on the Internet. We analyzed the chatting data on the basis of conversation tags that are added by players. We extracted players' action and context protocols in analysis, revealing common communication patterns. We found that mistranslation tends to occur when their conversation are complex, such as containing conversational congestion. This is usually overcome by communicating on common ground after experiencing conversational congestion.
Collaborative Sequencing (CoSeq) is the process by which a group collaboratively constructs a sequence. CoSeq is ubiquitous, occurring across diverse situations like trip planning, course scheduling, or book writing. Building a consensus on a sequence is desirable to groups. However, accomplishing this requires groups to dedicate significant effort to comprehensively discuss preferences and resolve conflicts. Furthermore, as numerous decisions must be assessed to construct a sequence, this challenge can be exacerbated in CoSeq. However, little research has aimed to effectively support consensus building in CoSeq. As a first step to systematically understand and support consensus building in CoSeq, we conducted a formative study to gain insights into how visual awareness may facilitate the holistic recognition of preferences and the resolution of conflicts within a group. From the study, we identified design requirements to support consensus building and designed a novel visual awareness technique for CoSeq. We instantiated this design in a collaborative travel itinerary planning system, Twine, and conducted a summative study to evaluate its effects. We found that visual awareness could decrease the effort of communicating preferences by 21%, and participants' comments suggest that it also encouraged group members to behave more cooperatively when building a consensus. We discuss future research directions to further explore the needs and challenges in this unique context and to advance the development of support for CoSeq tasks.
Full-text available
Over the past fifteen years, the Internet has triggered a boom in research on human behavior. As growing numbers of people interact on a regular basis in chat rooms, web forums, listservs, email, instant messaging environments and the like, social scientists, marketers, and educators look to their behavior in an effort to understand the nature of computer-mediated communication and how it can be optimized in specific contexts of use. This effort is facilitated by the fact that people engage in socially meaningful activities online in a way that typically leaves a textual trace, making the interactions more accessible to scrutiny and reflection than is the case in ephemeral spoken communication, and enabling researchers to employ empirical, micro-level methods to shed light on macro-level phenomena. Despite this potential, much research on online behavior is anecdotal and speculative, rather than empirically grounded. Moreover, Internet research often suffers from a premature impulse to label online phenomena in broad terms, for example, all groups of people interacting online are “communities”; the language of the Internet is a single style or “genre.” Notions such as community and genre are familiar and evocative, yet notoriously slippery, and unhelpful (or worse) if applied indiscriminately. An important challenge facing Internet researchers is thus how to identify and describe online phenomena in culturally meaningful terms, while at the same time grounding their distinctions in empirically observable behavior.
John Searle's Speech Acts made a highly original contribution to work in the philosophy of language. Expression and Meaning is a direct successor, concerned to develop and refine the account presented in Searle's earlier work, and to extend its application to other modes of discourse such as metaphor, fiction, reference, and indirect speech arts. Searle also presents a rational taxonomy of types of speech acts and explores the relation between the meanings of sentences and the contexts of their utterance. The book points forward to a larger theme implicit in these problems - the basis certain features of speech have in the intentionality of mind, and even more generally, the relation of the philosophy of language to the philosophy of mind.
Role theory concerns one of the most important features of social life, characteristic behavior patterns or roles. It explains roles by presuming that persons are members of social positions and hold expectations for their own behaviors and those of other persons. Its vocabulary and concerns are popular among social scientists and practitioners, and role concepts have generated a lot of research. At least five perspectives may be discriminated in recent work within the field: functional, symbolic interactionist, structural, organizational, and cognitive role theory. Much of role research reflects practical concerns and derived concepts, and research on four such concepts is reviewed: consensus, conformity, role conflict, and role taking. Recent developments suggest both centrifugal and integrative forces within the role field. The former reflect differing perspectival commitments of scholars, confusions and disagreements over use of role concepts, and the fact that role theory is used to analyze various forms of social system. The latter reflect the shared, basic concerns of the field and efforts by role theorists to seek a broad version of the field that will accommodate a wide range of interests.
The quality of Wikipedia articles is debatable. On the one hand, existing research indicates that not only are people willing to contribute articles but the quality of these articles is close to that found in conventional encyclopedias. On the other hand, the public has never stopped criticizing the quality of Wikipedia articles, and critics never have trouble finding low-quality Wikipedia articles. Why do Wikipedia articles vary widely in quality&quest; We investigate the relationship between collaboration and Wikipedia article quality. We show that the quality of Wikipedia articles is not only dependent on the different types of contributors but also on how they collaborate. Based on an empirical study, we classify contributors based on their roles in editing individual Wikipedia articles. We identify various patterns of collaboration based on the provenance or, more specifically, who does what to Wikipedia articles. Our research helps identify collaboration patterns that are preferable or detrimental for article quality, thus providing insights for designing tools and mechanisms to improve the quality of Wikipedia articles.
The bulk of our understanding of teams is based on traditional teams in which all members are collocated and communicate face to face. However, geographically distributed teams, whose members are not collocated and must often communicate via technology, are growing in prevalence. Studies from the field are beginning to suggest that geographically distributed teams operate differently and experience different outcomes than traditional teams. For example, empirical studies suggest that distributed teams experience high levels of conflict. These empirical studies offer rich and valuable descriptions of this conflict, but they do not systematically identify the mechanisms by which conflict is engendered in distributed teams. In this paper, we develop a theory-based explanation of how geographical distribution provokes team-level conflict. We do so by considering the two characteristics that distinguish distributed teams from traditional ones: Namely, we examine how being distant from one's team members and relying on technology to mediate communication and collaborative work impacts team members. Our analysis identifies antecedents to conflict that are unique to distributed teams. We predict that conflict of all types (task, affective, and process) will be detrimental to the performance of distributed teams, a result that is contrary to much research on traditional teams. We also investigate conflict as a dynamic process to determine how teams might mitigate these negative impacts over time.