Directory UMM :Data Elmu:jurnal:I:Information and Management:

Information & Management 37 (2000) 95±100

Flaming among ®rst-time group support system users
Milam Aikena,*, Bennie Wallerb,1
b

a
School of Business Administration, University of Mississippi, University, MS 38677, USA
School of Business, Francis Marion University, P.O. Box 100547, Florence, SC 29501-0547, USA

Received 24 March 1999; accepted 1 August 1999

Abstract
Numerous bene®ts, including increases in ef®ciency, effectiveness, and participant satisfaction, have been noted in the
literature when electronic meetings are used in place of traditional, oral meetings. However, several costs, or process losses,
have also been observed, including an increase in `¯aming' characterized by insults or even obscenities. This paper describes
how ¯aming may be correlated with numerous task and group member characteristics. Results of a case study show that a
large number of ¯ames are unrelated to the topic and that a small minority of those writing comments are responsible for the
majority of ¯ames. No variables were found to be signi®cant predictors, but its incidence was exclusively among males.
# 2000 Elsevier Science B.V. All rights reserved.
Keywords: Electronic meetings; Group support systems; Brainstorming; Disinhibition; Flaming


1. Introduction
Managers spend a large percentage of their time in
meetings, but traditional, oral meetings are often
inef®cient and ineffective. According to a study at
the University of Southern California in Los Angeles,
the average meeting has no written agenda, and the
purpose of the meeting is accomplished only 50% of
the time [13]. Further, 25% of the participants feel
they waste time on irrelevant issues, 33% of the
participants feel pressured to advocate opinions
publicly though they privately disagree with them,
and 33% of the participants feel they have minimal
in¯uence on the meeting.

*

Corresponding author. Tel.: ‡1-662-915-5777.
E-mail address: aiken@bus.olemiss.edu (M. Aiken).
1

Tel.: ‡1-843-661-1501.

Group support systems (GSSs) were developed to
improve meetings. Using this technology, group members are able to participate more and state their
opinions anonymously while all of the discussion
may be recorded automatically on a computer disk
for printout and possible distribution after the meeting
[5].
There are some disadvantages of using the technology, however. In most electronic meetings, participants share comments anonymously or with a pen
name. While this promotes more candid comments, it
also may increase ¯aming, i.e. comments intended to
offend others [15]. While somewhat subjective, at the
extreme ¯aming includes obscenities and other inappropriate comments, and may lead to avoidance of the
technology [16].
While some researchers have proposed possible
causes of ¯aming in the GSS environment and a
few studies have looked at the incidence of ¯aming

0378-7206/00/$ ± see front matter # 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 8 - 7 2 0 6 ( 9 9 ) 0 0 0 3 6 - 1


96

M. Aiken, B. Waller / Information & Management 37 (2000) 95±100

on a group basis, none that we have found have tracked
¯aming on an individual basis, possibly due to limitations of the software. The purpose of this effort was to
examine several user characteristics as they relate to
the number of relevant, irrelevant, and ¯aming comments generated by each participant.

2. Background
There appears to be little consensus on an exact
de®nition of ¯aming [10]. Obscenities and insults are
easily recognized as instances of it, but a large degree
of subjectivity is needed to distinguish impassioned
criticism from insult. At one extreme, a study included
any comments with exclamation points as instances of
¯aming [19], but most researchers de®ne ¯aming as
hostile comments directed at a person rather than an
idea.

Experience with over 1000 participants in 70 meetings has shown that ®rst-time users become disinhibited through the use of an anonymity feature in the
software, and there is frequently laughter about jokes
that all participants view at the same time [6]. Many
®rst-time GSS participants have never been able to
express their ideas with complete anonymity while
simultaneously viewing others' comments and wish to
try out the feature with wild, if not socially-unacceptable views. Continued use of such systems may
reduce the incidence of this type of ¯aming, however,
as the novelty wears off.
Several studies have found more ¯aming in anonymous electronic meetings than in traditional meetings. For example, one study using students who did
not know each other for hypothetical tasks found a
large amount of ¯aming when social control breaks
down [11]. Other studies, however, have found little or
no increases in ¯aming [14,18]. For example, one used
managers and professionals who were about 43 years
old and employed in a conservative corporation; they
found little disinhibited behavior. This may be because
the members were part of a social organization having
ongoing expectations of common group identity and
shared activities to accomplish [9]. Similarly, a study

of mature bank employees generating a strategic plan
using a GSS found no incidences of ¯aming [1].
Social norms and culture may play a large role. For
example, groups from Asia are often more restrained

and consensus-seeking and rarely generate in¯ammatory comments, even in anonymous electronic meetings [3,4]. In addition, the incidence of ¯aming and
irrelevant comments may be reduced signi®cantly by
strong admonitions from the group facilitator to
refrain from such conduct [17]. Although a few theories have been proposed, no studies have found
signi®cant predictors of why ¯aming comments are
made by speci®c individuals.

3. Case study
To examine why ¯aming occurs in an anonymous,
electronic meeting, a case study was conducted.
3.1. Subjects
Two groups of 45 undergraduate Business students
discussed two topics using a GSS as part of a required
assignment in their electronic classroom at their regularly scheduled meeting time. The groups were fairly
homogeneous in age, education, typing speed, and

computer experience. Approximately 40% of each
group were female.
3.2. Task
After a brief explanation of how to use the software,
the subjects spent 10 minutes generating comments on
two topics:
1. `Should the President of the United States be
removed from of®ce? Why or why not?'
2. `How can we solve the parking problem on the
campus?'
Topic 1 was chosen because of the level of controversy surrounding the issue at the time of the
assignment, the subjects were likely to have knowledge of it, and we felt a controversial topic could result
in more ¯aming. Topic 2 was chosen because it was
believed to be less controversial, and it has been used
in prior experiments, e.g. [8].
From experience, we have found that 10 minutes is
adequate time to capture most of a meeting's participants' views. The ®rst group discussed the parking
problem ®rst while the second group discussed the
presidential removal problem ®rst.


97

M. Aiken, B. Waller / Information & Management 37 (2000) 95±100

3.3. Procedure
Following each idea generation task, participants
completed an online questionnaire using a seven-point
Likert scale with 1 equivalent to not controversial, no
knowledge, etc. and 7 equivalent to extremely controversial and a great deal of knowledge:
1. How controversial is this topic?
2. How much do you know about this topic?
3. How familiar are you with electronic communication?
4. How important is this topic?
5. How satisfied are you with the meeting process?
In addition, we recorded the sex of each individual
and the number of the computer that the participant
was using.
3.4. Dependent variable measures
The number of relevant, irrelevant, and ¯aming
comments were measured through a content coding

of transcripts for each participant. The electronic
meeting software automatically recorded comments
from each group member in separate ®les and used the
same identi®er (the computer number) for each member's questionnaire answers. Coding was conducted by
two independent observers.
Relevant comments were deemed to be those ideas
related to the topic and not in¯ammatory (e.g., `Build
a parking garage'). Flaming comments were judged
to be those which might offend somebody in the
group. These comments included those which
insulted another person in the group by name (e.g.,
`Teach_how to drive'), obscenities, and sexual references. Many of these ¯ames were also irrelevant, but
were not counted as such. Irrelevant comments were
counted as those not related to the topic and not
in¯ammatory (e.g., `When can we get out of here?').
Two independent coders evaluated transcripts of the
meetings and agreed on 98% of their selections,
indicating adequate inter-rater reliability.
4. Results
4.1. Independent variables

The results of the questionnaires and the number of
comments generated for both groups are shown in

Table 1
Questionnaire results (overall)
Parking

Controversy
Knowledge
Communication
Importance
Satisfaction
Relevant
Irrelevant
Flames

President

Mean


SD

Mean

SD

4.52
4.32
3.23
4.93
4.19
1.48
0.33
0.48

1.86
1.82
1.73
1.78
1.93

1.59
0.56
0.88

5.19
4.20
3.44
4.56
3.87
0.73
0.41
0.77

1.82
1.63
1.97
2.13
1.88
0.84
1.13
1.44

Table 1. As we suspected in designing the tasks,
subjects believed that the removal of the president
was more controversial (F ˆ 4.69, p ˆ 0.032). Parking was believed to be more important, but not signi®cantly so (F ˆ 1.25, p ˆ 0.265). The subjects'
knowledge of both problems was approximately the
same (F ˆ 0.17, p ˆ 0.68), and their satisfaction with
the meeting process discussing each topic was not
signi®cantly different (F ˆ 1.02, p ˆ 0.314). The
question regarding their electronic communication
experience was repeated in the questionnaires.
Although, the mean was slightly different, it was
not signi®cant (F ˆ 0.45, p ˆ 0.504). The results of
the questionnaires and the number of comments generated for group 1 and group 2 are shown in Tables 2
and 3.
4.2. Comment analysis
Table 4 shows the number of relevant, irrelevant,
and ¯aming comments for each group by topic. A
Table 2
Questionnaire results (group 1)
Parking

Controversy
Knowledge
Communication
Importance
Satisfaction
Relevant
Irrelevant
Flames

President

Mean

SD

Mean

SD

4.26
3.66
3.49
5.11
4.14
1.83
0.31
0.34

1.87
1.41
1.77
1.66
1.59
1.72
0.47
0.76

5.38
4.54
3.95
4.84
4.51
0.68
0.62
1.14

1.78
1.56
1.72
2.05
1.77
1.00
1.48
1.78

98

M. Aiken, B. Waller / Information & Management 37 (2000) 95±100

Table 3
Questionnaire results (group 2)
President

Controversy
Knowledge
Communication
Importance
Satisfaction
Relevant
Irrelevant
Flames

Parking

Mean

SD

Mean

SD

5.00
3.87
2.95
4.29
3.24
0.79
0.21
0.42

1.86
1.65
2.10
2.20
1.79
0.66
0.58
0.89

4.79
5.00
2.97
4.74
4.24
1.12
0.35
0.62

1.84
1.95
1.68
1.90
2.26
1.37
0.69
0.99

large amount of `free-riding' was exhibited by the
subjects. Although they were encouraged to write
comments, only 25 of the 45 in the ®rst group wrote
about the parking problem and only 20 wrote about the
president. In the second group, only 22 of 45 wrote
about the president and only 21 about parking. A large
portion of the groups wanted to just read others'
comments.
Flaming comments were written by a small minority of the participants. In the ®rst group, ¯ames were
written by ®ve participants (20% of the total who
wrote comments) when discussing parking and 10
participants (50% of the total who wrote comments)
when discussing the second problem. All ®ve of those
who wrote ¯aming comments for parking also wrote
¯aming comments about the second issue. In the
second group, ¯ames were written by four participants
(19% of the total who wrote comments) when discussing parking and 10 participants (27% of the total
who wrote comments) when discussing the president.
Two of the four who wrote ¯aming comments for the
®rst topic also wrote ¯aming comments for the second
topic. Of those who wrote ¯aming comments, one or
Table 4
Comment analysis

two wrote a disproportionate share. For example, one
participant in group 1 discussing the president wrote
seven ¯ames, three irrelevant comments, and no relevant comments. That is, one of 10 who wrote ¯ames
wrote 22% of the total. Finally, all of the participants
who wrote ¯ames were male. A similar result was
found in a study of Japanese GSS participants in which
females were more polite and readers of their anonymous comments could often distinguish those written
by them [2].
Most of the ¯ames were not relevant to the topic
(60%), and many were insults to a particular person in
the group directly by name (30%) or indirectly (14%),
e.g. `The person who wrote comment number 45 is
a_'. There was no statistically signi®cant difference
between the number of irrelevant ¯ames based upon
which topic was chosen. Some of the ¯ames were
repeated, and many referred to or built upon other
¯ames.
The topic order may have had an effect on the
number of comments generated and the amount of
¯aming. The number of ¯aming and irrelevant comments increased when the ®rst group switched from
the parking problem to the president, while the number of relevant comments decreased. The number of
¯ames and irrelevant comments increased when the
second group switched from the president to the
parking problem, but the number of relevant comments also increased. The subjects may have learned
how to use the software more and its capabilities. In
addition, ¯aming might have spawned more ¯aming in
response.
The number of relevant comments about parking
was signi®cantly greater than the number about the
president (F ˆ 12.65, p ˆ 0.001) but the numbers of
irrelevant and ¯aming comments were not signi®cantly different (F ˆ 0.28, p ˆ 0.599) and
(F ˆ 2.15, p ˆ 0.145), respectively.
4.3. Variable correlations

Relevant

Irrelevant

Flaming

Group 1
Parking
President

54
19

6
21

6
32

Group 2
President
Parking

23
31

5
8

6
11

A Pearson correlation analysis was conducted on
the data, and signi®cant results (at a ˆ .05) were
found among several variables. Subjects with more
knowledge about the topics perceived them as more
controversial (R2 ˆ 0.395, p ˆ 0.0001) and more
important (R2 ˆ 0.309, p ˆ 0.0002), had more electronic communication experience (R2 ˆ 0.205,

M. Aiken, B. Waller / Information & Management 37 (2000) 95±100

p ˆ 0.0139), were more satis®ed with the meeting
process (R2 ˆ 0.244, p ˆ 0.0032), and generated more
relevant comments (R2 ˆ 0.186, p ˆ 0.0255). Those
with more electronic communication experience also
believed the topics were more important (R2 ˆ 0.204,
p ˆ 0.014), were more satis®ed with the meeting
process (R2 ˆ 0.173, p ˆ 0.0377), generated more
relevant comments (R2 ˆ 0.266, p ˆ 0.0064), and
generated more irrelevant comments (R2 ˆ 0.164,
p ˆ 0.0492). The perceived level of importance was
also signi®cantly correlated with the perceived level of
controversy (R2 ˆ 0.369, p ˆ 0.0001). The topic was
signi®cantly correlated with the level of controversy
(R2 ˆ 0.179, p ˆ 0.032) and the number of relevant
comments (R2 ˆ ÿ0.286, p ˆ 0.001). Several of these
results are due to cross-correlation, however.
None of the variables were found to be signi®cant
predictors of ¯aming, although satisfaction with the
meeting process (F ˆ 2.12, p ˆ 0.055) and importance of the topic (F ˆ 2.13, p ˆ 0.054) were close
to our threshold level a ˆ 0.05. There were two postfacto predictors of ¯aming, however: the number of
relevant comments (F ˆ 3.84, p ˆ 0.001; R2 ˆ 0.062)
and the number of irrelevant comments (F ˆ 7.79,
p ˆ 0.001; R2 ˆ 0.356, p ˆ 0.001).

5. Discussion
Kitchens [12] proposed that ¯aming is a function of
gender, age, race, alcohol consumption, locus of control, importance of the topic, satisfaction with the
meeting, anonymity, purpose of the ¯ame, presence
of formal rules, and authority. We did not test all of
these, but found that gender plays an important role:
all of the ¯ames were generated by males. We found
no signi®cant correlation between ¯aming and satisfaction with the meeting, however. Anecdotal evidence from conducting numerous GSS meetings
also suggests that strong admonitions from the group
leader or facilitator often suppresses the incidence of
¯aming. In addition, group members must also be
encouraged to write ideas to overcome free riding [7].
Thus, the group facilitator's leadership has a strong
role to play.
Most of the ¯ames were not related to the topic,
suggesting that the topic may not matter. Rather,
¯ames are probably due to the characteristics (such

99

as gender, level of maturity, hostility, etc.) of the
individual writing them. Only a small minority of
those writing comments generated ¯ames.
Although there were several interesting correlations
among the variables (e.g., those with more knowledge
about the topic and those with more electronic communication experience were more satis®ed with the
meeting), we found no signi®cant predictors of ¯aming. Those who ¯amed wrote more comments, on
average, but they were generally irrelevant to the
topic.

6. Conclusion
Group support systems may increase meeting
ef®ciency, effectiveness, and satisfaction, but with
anonymity come a few unwanted effects, such as
¯aming; i.e., comments written to offend others in
the group. Warnings against such behavior may reduce
their occurrence, but little is known about why they
appear.
This study has investigated the affect of topic
knowledge, importance, and level of controversy,
along with familiarity with electronic communication,
satisfaction with the meeting process, and gender on
the incidence of ¯aming by individuals who have
never used the system before. Only gender played a
signi®cant role.

References
[1] M. Aiken, D. Hawley, Z. Min, H. Sloan, How to improve
bank meetings, Journal of Retail Banking 16(4) (1994/1995)
21±25.
[2] M. Aiken, C. Hwang, J. Martin, A Japanese group decision
support system, International Journal of Computer Applications in Technology 9(5/6), 1996, pp. 233±238.
[3] M. Aiken, C. Hwang, J. Paolillo, L. Lu, A group decision
support system for the Asian Pacific Rim, Journal of
International Information Management 3(2), 1994, pp. 1±13.
[4] M. Aiken, D. Kim, C. Hwang, L. Lu, A Korean group
decision support system, Information & Management 28,
1995, pp. 303±310.
[5] M. Aiken, J. Krosp, A. Shirani, J. Martin, Electronic
brainstorming in small and large groups, Information &
Management 27, 1994, pp. 141±149.
[6] M. Aiken, M. Vanjani, J. Paolillo, A comparison of two
electronic idea generation techniques, Information & Management 30, 1996, pp. 91±99.

100

M. Aiken, B. Waller / Information & Management 37 (2000) 95±100

[7] R. Albanese, D. Van Fleet, Rational behavior in groups: the
free riding tendency, Academy of Management Review 10,
1985, pp. 244±255.
[8] R. Gallupe, A. Dennis, W. Cooper, J. Valacich, L. Bastianutti,
J. Nunamaker, Electronic brainstorming and group size, The
Academy of Management Journal 35(2), 1992, pp. 350±369.
[9] S. Hiltz, M. Turoff, K. Johnson, Experiments in group
decision making, 3: disinhibition, deindividuation, and group
process in pen name and real name computer conferences,
Decision Support Systems 5, 1989, pp. 217±232.
[10] J. Kayany, Contexts of uninhibited online behavior: flaming
in social newsgroups on usenet, Journal of the American
Society for Information Science 49(12), 1998, pp. 1135±
1141.
[11] S. Kiesler, J. Siegel, T. McGuire, Social psychological aspects
of computer-mediated communication, American Psychologist 39, 1984, pp. 1123±1134.
[12] F. Kitchens, Flaming in electronic media, Proceedings of the
29th Annual Conference of the Southwest Decision Sciences
Institute, 3±7 March 1998, pp. 41±45.
[13] A. LaPlante, '90s style brainstorming, Forbes ASAP, 25
October 1993, pp. 41±61.
[14] M. Lea, R. Spears, Computer-mediated communication, deindividuation and group decision making, International
Journal of Man-Machine Studies 34(2), 1991, pp. 283±302.
[15] J. Nunamaker, A. Dennis, J. Valacich, D. Vogel, J. George,
Electronic meeting systems to support group work, Communications of the ACM 34(7), 1991, pp. 40±61.
[16] B. Reinig, R. Briggs, S. Brandt, J. Nunamaker, The electronic
classroom on fire: why it happens and how to put out the
flames, Proceedings of the 30th Annual Hawaii International
Conference on System Sciences 2, 1997, pp. 639±647.

[17] B. Reinig, R. Briggs, J. Nunamaker, Flaming in the electronic
classroom, Journal of Management Information Systems
14(3) (1997/1998) 45±59.
[18] R. Rice, Computer-mediated communication system network
data: theoretical concerns and empirical examples, International Journal of Man-Machine Studies 32(6), 1990, pp. 627±
648.
[19] J. Siegel, S. Dubrovsky, S. Kiesler, T. McGuire, Group
processes in computer-mediated communication, Organizational Behavior and Human Decision Processes 37, 1986, pp.
157±187.
Milam Aiken received a BS in Engineering and an MBA from the University
of Oklahoma, a BA in Computer Science
and a BS in Business from the State
University of New York, and a PhD in
Management Information Systems from
the University of Arizona. He is an
Associate Professor of Management Information Systems at the University of
Mississippi.
Bennie Waller received a BS in Marketing from Longwood College, an MBA
from the University of North CarolinaWilmington, and a PhD in Finance from
the University of Mississippi. He is an
Assistant Professor at Francis Marion
University.