C

Group Decision and Negotiation 13: 381–399, 2004 2004 Kluwer Academic Publishers. Printed in the Netherlands

Divergent and Convergent Idea Generation in Teams: A Comparison of Computer-Mediated and Face-to-Face Communication
DAVID S. KERR Department of Accounting, Mays School of Business, Texas A&M University, College Station, TX 77843-4353, USA (E-mail: d-kerr@tamu.edu)

UDAY S. MURTHY School of Accountancy, University of South Florida, 4202 E. Fowler Avenue BSN3403, Tampa, FL 33620-5500, USA (E-mail: umurthy@coba.usf.edu)

Abstract
Many tasks and decisions in business, including management consulting, are performed in group settings. Computer-mediated communication (CMC) tools (e.g., Lotus Notes) are increasingly being used by businesses to support teams in a variety of settings. Considerable research in information systems has demonstrated the advantages of “electronic brainstorming” (EBS) for generic tasks involving only divergent thinking. However, it is unclear whether the benefits of CMC extend to tasks that require both divergent and convergent processes. Per task–technology fit theory (TTF) (Zigurs and Buckland 1998), the use of computer-based group communication support tools, including “chat” systems in wide-spread use today, may be less effective for convergent processes than for divergent processes. This study experimentally compares the performance of computer-mediated and face-to-face (FTF) teams on tasks requiring both divergent and convergent processes. Consistent with theoretical predictions, the results revealed that computer-mediated teams outperformed FTF teams in the divergent aspects of the tasks, while FTF teams outperformed computer-mediated teams in the convergent aspects of the tasks. Key words: computer-mediated communication (CMC), group support systems (GSS), task–technology fit theory, idea generation, divergent and convergent processes

Introduction In many decision-making settings in business, professionals frequently interact in teams to generate ideas and solve problems. The use of information technology to support collaborative teamwork in business has been the subject of ongoing research. While the majority of this research has been in the information systems discipline, there has been some research in related disciplines employing business-oriented contexts and tasks. For example, in the field of accounting, researchers have begun examining the effects of computer-mediated communication (CMC) on various aspects of team decision-making (Bamber, Watson and Hill 1996; Karan et al. 1996; Ho 1999; Murthy and Kerr 2004). CMC tools, also known as

382

KERR AND MURTHY

group support systems (GSS), provide a structured environment that allows participants in a collaborative team to interact simultaneously and anonymously to generate ideas, make decisions, and solve problems (Jessup, Connolly and Tansik 1990). Such interaction is expected to increase in the future as technology and groupware increase the opportunities for simultaneous interaction among team members in geographically dispersed locations. One task that has received much attention in the information systems literature is that of idea generation or “brainstorming.” This research has found that groups brainstorming electronically typically generate more ideas than do groups brainstorming in face-to-face (FTF) settings (Connolly, Jessup and Valacich, 1990; Gallupe, Bastianutti and Cooper, 1991; Gallupe et al. 1992; Valacich et al. 1993; Valacich, Dennis and Connolly 1994). There are two noteworthy characteristics of the types of tasks typically employed in prior research on electronic brainstorming (EBS). First, following Osborn’s (1963) classic brainstorming instructions, participants were told to engage in divergent thinking where “the wilder the idea, the better.” Second, the tasks were generic enough where participants could contribute ideas without having to recall any specialized knowledge.1 By contrast, in many business-consulting settings, team members must generate relevant, cost-effective ideas for solving client-specific problems. Wild ideas that may by appropriate for some brainstorming situations are less likely to be productive in idea-generation tasks aimed at solving a specific business problem. Furthermore, in addressing client-specific business problems, team members must call upon specialized knowledge to increase the relevancy of ideas generated. Such problem-solving-directed collaborative work requires participants to convey alternative solution proposals to the team, following which the team must converge on the key proposal(s) that would best solve the client-specific problem at hand. Prior EBS research has typically employed CMC support systems – chat systems – that facilitate parallel communication (simultaneous input by all team members) and display a real-time log of ideas to all team members, referred to as the “group memory.” While chat systems have been shown to be beneficial in brainstorming sessions employing Osborn’s rules (no criticism allowed, the wilder the idea the better), prior research has not examined whether they are equally appropriate for problem-solving-directed tasks in which wild ideas may be counterproductive. According to task–technology fit (TTF) theory (Zigurs and Buckland 1998), team performance when using CMC is optimized when the toolset provided to team members matches the requirements of the task. For problem solvingdirected collaborative idea-generation tasks, participants must not only generate ideas (a divergent process), but they must also distinguish between relevant and irrelevant ideas (a convergent process). Per TTF theory, for divergent processes, the CMC tool should offer communication support, whereas for convergent processes, the CMC tool should offer information processing support. Relative to face-to-face interaction, CMC has certain process gains, most notably parallel communication support, which facilitate divergent oriented tasks such as brainstorming. However, chat-based CMC also has several process losses relative to face-to-face interaction, such as lack of real-time feedback from others in the team, and increased potential for team members to devote insufficient time to reading and evaluating ideas put forth by others. While these process losses are unlikely to hinder teams when performing divergent-oriented tasks, they are more likely to impair convergentoriented tasks which require generating and identifying relevant, high-quality ideas rather

DIVERGENT AND CONVERGENT IDEA GENERATION IN TEAMS

383

than just wild ideas. Thus, the purpose of this study is to test the predictions of TTF theory regarding the effects of a popular CMC tool – a chat system – on teams’ performance in problem-solving-oriented tasks involving both divergent and convergent processes. This study extends prior research by comparing, via an experiment, the performance of teams meeting face-to-face (FTF) with those using a CMC chat system in the context of real-world consulting tasks requiring both divergent and convergent processes. The experiment employed two cases requiring participants to make recommendations regarding how a company’s operations could be improved. Each case required participants to both convey alternative ideas and recommendations, a divergent process, and also filter out irrelevant recommendations, a convergent process. Performance in both the divergent and the convergent aspects of the tasks was compared between computer-mediated teams and unassisted (FTF) teams. Treatments were operationalized to mirror as closely as possible real-world FTF and CMC settings, including the use of a scribe in the FTF setting, and a requirement for participants using CMC to identify or “tag” ideas they felt were relevant and should be recommended to the client. Consistent with theoretical predictions, the results reveal that participants who worked in computer-mediated collaborative teams performed better at the divergent aspect of the tasks—they generated a greater number of recommendations than participants who interacted FTF. However, the analysis also revealed that participants who worked in computermediated teams performed worse at the convergent aspect of the tasks, as their recommendations included a higher percentage of irrelevant suggestions than those of participants who interacted FTF. The main message of this study is that chat-based systems, which prior research has shown lead to superior performance by CMC teams on generic brainstorming tasks involving only divergent thinking, are not as well suited for problem-solving-oriented idea generation tasks that require a focus on relevant, cost-effective ideas rather than “wild ideas.” Consequently, prior findings in the literature that CMC is superior to FTF for unfettered brainstorming do not necessarily apply in the context of more focused problemsolving-oriented idea generation in which participants must generate relevantideas in order to solve the problem at hand. The remainder of the paper is organized as follows. The next section provides a background discussion of CMC research examining EBS and TTF theory. The research hypotheses are then presented, stating the expected effects of FTF interaction and CMC in the context of divergent and convergent processes. The research method is discussed next, with a description of the CMC tool used in the experiment. This is followed by the presentation of the results of the experiment. The implications of the results are then discussed, after which the concluding section summarizes the paper and provides some possible directions for future research. Background and Development of Hypotheses Many business decisions are made in team settings. Team discussions often result in the sharing of expertise, the elicitation of many different points of view, and the consideration of alternatives that individuals acting alone might overlook. Most team meetings occur in FTF settings. However, as suggested by Steiner (1972), there are several “process losses”

384

KERR AND MURTHY

associated with meeting FTF which limit the exchange of information. The process losses inherent to FTF settings can be overcome to an extent using CMC. Computer-mediated tools for supporting group work combine communication, computer, and decision technologies to support the formulation of and solution to unstructured or semi-structured problems (Jessup et al. 1990). Chat tools available in virtually all CMC systems overcome the main problem inhibiting idea generation in FTF interacting groups – production blocking – by allowing simultaneous input to the brainstorming session, also referred to as “parallel communication.” Another advantage of CMC is that group members’ typed comments are recorded within the system’s “group memory.” Group members can view prior comments without risk of missing new information (Dennis 1996), and “may process the message at an optimal pace, stopping to consider difficult points” (Petty and Cacioppo 1986, p. 77). Given the aforementioned advantages of CMC over FTF communication, “electronic brainstorming” should be superior to FTF brainstorming. Indeed, a considerable amount of research has confirmed the superiority of computer-mediated brainstorming over FTF brainstorming (Gallupe, Bastianutti and Cooper 1991; Gallupe et al. 1992; Valacich et al. 1993; Dennis and Valacich 1999; Pinsonneault et al. 1999; Satzinger, Garfield and Nagasundaram 1999). It should be noted, however, that prior studies on EBS have almost always employed Osborn’s (1963) four rules for brainstorming, which are as follows: (1) Criticism is ruled out. Adverse judgment of ideas must be withheld (2) “Free-wheeling” is welcomed. The wilder the idea, the better; it is easier to tame down than to think up (3) Quantity is wanted. The greater the number of ideas, the more the likelihood of useful ideas (4) Combination and improvement are sought. In addition to contributing ideas of their own, participants should suggest how ideas of others can be turned into better ideas or how two or more ideas can be joined into still another idea. As is evident from the above rules, the emphasis in brainstorming is on generating the largest quantity of ideas. Thus, the vast majority of the extant research on EBS has simply compared the divergent aspect ofperformance between CMC groups and groups interacting FTF by counting the number of ideas generated. In the current study, we compare the performance of CMC groups and FTF groups in both the divergent aspect of performance, measured by the number of unique ideas generated, as well as the convergent aspect of performance, measured by how well the groups perform at weeding out irrelevant ideas. Barki and Pinsonneault (2001) indicate that little EBS research has assessed idea quality directly, which they suggest is the most important indicator of group brainstorming performance. Using an unbalanced repeated measures experimental design, Barki and Pinsonneault compared the quality of ideas generated by small groups using four brainstorming technologies: nominal, verbal, EBS-anonymous, and EBS-non-anonymous. The study also manipulated three factors thought to improve the effectiveness of EBS groups, i.e., group history, contextual cues, and topic sensitivity. The results indicated that overall, nominal brainstorming groups generated ideas that were at least as good as, if not better than, those generated by EBS groups, while the three conditions manipulated had no effect on the quality of the ideas generated by EBS groups. Although the study focused on idea quality, the tasks used were generic brainstorming tasks (how to reduce violent crimes, how to improve the parking problem on campus, how to increase tourism, and how to solve the AIDS problem). Additionally, Barki and Pinsonneault did not compare EBS relative to FTF

DIVERGENT AND CONVERGENT IDEA GENERATION IN TEAMS

385

brainstorming. The current study extends this line of research by employing a problemsolving-oriented business task and comparing the performance of CMC and face-to-face teams. Theory and hypotheses Building on McGrath’s (1984) task circumplex and information/media richness theory (Daft and Lengel 1986), and Zigurs and Buckland (1998) proposed a theory of TTF.2 The theory posits a typology of tasks and CMC technologies and prescribes a best fit between different types of tasks and technology. CMC technologies are defined broadly by Zigurs and Buckland (1998) as a set of communication, information processing, and agenda structuring tools designed to work together to support the accomplishment of group tasks. Communication support tools assist with the communication needs of the group, and include features such as simultaneous and anonymous input and the display of comments made by group members; information processing support tools relate to the aggregation, organization, and evaluation of information input by group members and other information sources; and agenda structuring support tools are used to define and facilitate the process by which group members interact, and include tools used for setting and enforcing agendas. TTF theory posits that a good “fit” between these CMC technology tools and task requirements will result in the best team performance. The question of whether “fit” matters in the use of CMC technologies to support group work has been the subject of recent research. Using the case study method, Pollard (2003) conducted in-depth interviews with adopters of GSS and concluded that that lack of TTF strongly influenced discontinued use of the systems. Dennis, Wixom and Vandenberg (2001) propose a “Fit-Appropriation Model” for interpreting the effects of GSS on performance, arguing that GSS performance is affected by (1) the fit between the task and the GSS structures selected for use (i.e., communication support and information processing support), and (2) the appropriation support the group receives in the form of training, facilitation, and software restrictiveness to help them effectively incorporate the selected GSS structures into their meeting process. Based on a meta-analysis using this model to organize and classify past research, the authors conclude that fitting the GSS to the task had the most impact on outcome effectiveness (decision quality of ideas). Shirani, Tafti and Affisco (1999) conducted an experiment to examine the interaction between task structure and technology to support synchronous and asynchronous group communication. They compared participant performance using email and a GSS and two levels of task structure. The less-structured task involved an analysis of Citicorp’s competitive strategy, whereas the more structured task involved selecting between two operating systems. In addition to measuring the total number of unique ideas generated, ideas were further divided into basic and inferential idea categories. The results indicates that GSS groups generated more total and basic ideas, but groups using email performed a deeper problem analysis as indicated by a higher proportion of inferential ideas generated by these groups. However, it is worth noting that neither the structured nor the unstructured task used in this study required participants to engage in both divergent and convergent thinking. Furthermore, the study did not include a FTF team interaction mode, leaving open the

386

KERR AND MURTHY

question of how the performance of GSS groups or those using email would have compared to teams meeting FTF. As indicated previously, in the type of brainstorming tasks typically used in prior EBS research, participants are told that the more ideas generated the better, the wilder the ideas the better, and to avoid criticizing others’ ideas. For most real-world tasks undertaken by teams in business-consulting settings, however, wild ideas may not be of much help as the team attempts to craft a solution for the client. Such tasks can be characterized as intellective tasks involving an initial process of divergent thinking followed by a process of convergent thinking. In the divergent thinking phase, where various possible solutions are generated and conveyed by each team member, the communication support feature of CMC should maximize the number of ideas generated, owing to the parallel communication and group memory features. In contrast to FTF settings where only one team member can speak at a time, computer-mediated groups should exhibit superior performance in terms of the number of unique ideas generated. In the convergent thinking phase of such tasks in business-consulting settings where ideas must be analyzed to identify relevant, cost-effective ideas that are well suited to solve the client’s problem at hand, TTF theory suggests that CMC teams should be provided with information processing support tools to facilitate convergent thinking. Specifically, computer-mediated teams should be provided with tools that enable them to distinguish between better and worse ideas from the list generated during divergent thinking, which is likely to be a long list due to the parallel communication feature.3 Absent such explicit information processing support, TTF theory suggests that teams interacting via a chat system are likely to perform worse than teams interacting FTF in the convergent thinking aspects of the tasks. In a FTF setting, given that only one person can be speaking at a time, ideas proposed can be instantly evaluated so that those deemed of poor quality are eliminated immediately. Thus, relative to a chat system, the inherent nature of the FTF setting facilitates the convergent aspects of the tasks. Hypotheses In the present study, we extend prior research on computer-mediated brainstorming by investigating the performance of computer-mediated teams relative to FTF teams in tasks involving both divergent and convergent communication processes. The tasks employed in the current study require participants to make recommendations regarding how a company’s internal control system and operations could be improved. The tasks also require participants to distinguish between relevant, cost-effective recommendations and irrelevant recommendations and therefore involve both conveying ideas and converging on relevant ideas. Relatively small teams of three or four persons per team are used, which leads to conservative observations of the benefits of CMC. Per TTF theory as discussed above, we expect CMC to be more effective than FTF communication in the divergent aspect of the tasks (i.e., generating and conveying alternative recommendations), since CMC teams are provided with communication support tools enabling parallel communication and an electronic group memory. Due to the immediate feedback available when discussing ideas face to face, FTF teams are expected to be more effective than computer-mediated teams in

DIVERGENT AND CONVERGENT IDEA GENERATION IN TEAMS

387

the convergent aspect of the tasks. Specifically, given that teams using a CMC chat system lack information processing support to facilitate distinction between relevant and irrelevant ideas while members of FTF teams are able to provide immediate feedback on each idea as it is proposed, we predict that FTF teams will recommend a lower percentage of irrelevant ideas relative to CMC teams. This discussion leads to the following hypotheses. Hypothesis 1. Computer-mediated teams will propose a greater number of recommendations than teams interacting face-to-face (divergence). Hypothesis 2. Computer-mediated teams will propose a greater percentage of irrelevant recommendations than face-to-face teams (convergence). According to Hypothesis 1, computer-mediated teams are expected to propose a greater number of recommendations, since the communication support provided by the chat system facilitates the divergent aspect of the tasks. On the other hand, per Hypothesis 2, since chat systems typically do not provide information processing support to facilitate the convergent aspect of the tasks, and since FTF teams have the advantage of immediate feedback when ideas are proposed, computer-mediated teams are expected to recommend a greater percentage of irrelevant (low-quality or cost-ineffective) ideas than FTF teams. In effect, teams’ performance in the convergent aspect of the tasks is measured in terms of the interacting teams’ degree of success in differentiating between relevant and irrelevant ideas.

Method Participants and design The research hypotheses were tested through an experiment using students in a mastersdegree program in business. The students were enrolled in a corporate auditing course and were formed in teams of four, or in a few cases three, at the beginning of the semester. Students were assigned to teams such that each team was as similar as possible to all other teams along four dimensions: team members’ (1) aptitude (measured by students’ overall grade point average at the beginning of the semester), (2) background (prior or current employment in an auditing related job, including internships), (3) membership in student business organizations, and (4) gender (no all-male or all-female teams). Prior to the experiment, participants had met several times with their teams to discuss and complete various course projects unrelated to the experiment. Thus, the participants interacted in “established teams” with a shared history rather than ad hoc teams formed merely for the purpose of the experiment. Subjects participated in the experiment in fulfillment of a course requirement. Participants were told that a portion of their course grade would be based on their performance on the experimental task, which served to motivate them to take the experimental task seriously. Teams were assigned to experimental conditions randomly. Forty-six participants in 10 teams of four and 2 teams of three worked in both computer-mediated and FTF communication environments. All teams completed one case in the CMC mode and a second case in the FTF communication mode. Based on a

388

KERR AND MURTHY

Figure 1. Experimental design.

counter-balanced design, half of the teams participated in the computer-mediated setting before they met FTF, and half participated in the computer-mediated setting after they met FTF. The experimental design is illustrated in Figure 1. The experimental task involved two cases in which teams focused on internal control design issues – a non-trivial, relatively unstructured task that consulting and auditing teams are likely to perform in groups, and which has received little attention in prior research. For their first case, participants were given a detailed flowchart illustrating the purchasing function for a company; see Figure 2. Participants were asked to study and discuss with their team the existing internal controls over purchases and then develop a list of recommendations regarding how company management could improve operations. Half of the teams completed this case in the computer-mediated mode and half completed the case FTF. In the second case, participants were given a narrative description of a company’s cash collections function and controls over cash; see Figure 3. Again, participants were asked to study and discuss with their team the existing internal controls and develop their list of management recommendations. In both cases, participants were told that their lists of recommendations would be evaluated on the bases of quality and quantity. Teams therefore had an incentive to differentiate between relevant (high-quality, cost-effective) and irrelevant (low-quality or cost-ineffective) recommendations proposed by team members, and to include only relevant recommendations on their lists. The experimental instructions thus encouraged both divergent thinking (quantity of recommendations) as well as convergent thinking (quality of recommendations). To ensure that participants had some degree of specialized knowledge relevant to the experimental task, participants were limited to masters-degree program students enrolled in a corporate auditing course. The topics of internal control evaluation and tests of controls and transactions had been covered in class, and participants had been tested on these topics. The experimental task focused on evaluation of the operations and design of internal control over purchasing and cash collections functions—areas with which these participants were familiar. Unlike tasks used in prior electronic brainstorming studies, the tasks used in the current study required participants to recall specialized knowledge (i.e., relevant internal

DIVERGENT AND CONVERGENT IDEA GENERATION IN TEAMS

389

Figure 2. Purchasing function case.

controls for the purchasing and cash receipt functions) in order to suggest improvements in the design of the internal control systems. Both cases were completed by all participants in a single session. Teams in the FTF discussion mode were seated around a table, with each team in a separate conference room.

390

KERR AND MURTHY

The Art Appreciation Society operates a museum for the benefit and enjoyment of the community. You have been asked by the board of directors of the Art Appreciation Society to provide suggestions for improvements in their admission procedures and controls over cash admission fees. During hours when the museum is open to the public, two clerks who are positioned at the entrance collect a five-dollar admission fee from each nonmember patron. Members of the Art Appreciation Society are permitted to enter free of charge upon presentation of their membership cards. At the end of each day, one of the clerks delivers the proceeds to the treasurer. The treasurer counts the cash in the presence of the clerk and places it in a safe. Each Friday afternoon, the treasurer and one of the clerks deliver all cash held in the safe to the bank and receive an authenticated deposit slip, which provides the basis for the weekly entry in the cash receipts journal. The board of directors of the Art Appreciation Society has identified a need to improve their internal control structure over cash admission fees. The board has determined that the cost of installing turnstiles, sales booths or otherwise altering the physical layout of the museum will greatly exceed any benefits which may be derived. However, the board has agreed that the sale of admission tickets must be an integral part of its improvement efforts.∗
∗

American Institute of Certified Public Accountants adapted

Figure 3. Cash collections case.

One member in each team was selected to serve as the team’s scribe to prepare the team’s list of recommendations. The use of a scribe in the FTF condition was intended to mimic similar real-world FTF meetings, where a scribe would likely exist. Participants were not permitted to use any reference materials. All FTF teams were monitored by a facilitator from a control room via closed-circuit television. Teams in the computer-mediated discussion mode used the chat feature of the VisionQuest group support system developed by Collaborative Technologies Corporation. VisionQuest supports same time/same place discussion, same time/different place discussion, and different time/different place discussion. The mechanics of using the VisionQuest software were explained to participants, and assistance was available for any participant needing help. Each team member was assigned a computer terminal in a large computer laboratory, with their other team members dispersed throughout the room. In effect, the setting simulated a “same time/different place” setting since verbal interaction between participants was not allowed. VisionQuest enables team members to offer their comments simultaneously. The program splits each team member’s computer screen so that each person can read the comment he or she is forming and sending as well as the other team members’ comments. Since the participants were only using the chat tool in VisionQuest, which involved typing their comments into the input window and pressing ‘Enter’, the brief training they received on using the system was deemed sufficient. The VisionQuest session log captured each team’s discussion and list of recommendations. In order to distinguish between discussion and subsequent statements of recommendations and to facilitate comparisons of CMC teams’ performance with that of FTF teams, participants were instructed to indicate which recommendations they would propose to the client’s management by preceding those recommendations with the code “reco” (short for “recommendation”). The requirement to type “reco” in front of ideas deemed worthy of being “submitted” by the team was aimed

DIVERGENT AND CONVERGENT IDEA GENERATION IN TEAMS

391

at mimicking the use of a “tagging” mechanism for comments.4 Only recommendations preceded with “reco” were included in the analyses, since using the entire chat session transcript to determine a team’s recommendations would have biased the results in favor of CMC teams on the divergent aspect of the tasks. Comments in the computer-mediated discussion mode were made anonymously. Although team members knew who was in their team, they were unaware of the individual who made each comment. All teams were allowed 30 minutes to complete each case, which appeared to be sufficient time in both interaction modes. Dependent measures The two primary dependent measures in the study were (1) the total number of management recommendations proposed by each team, after eliminating duplicates, and (2) the percentage of irrelevant recommendations proposed by each team (number of irrelevant recommendations divided by total number of recommendations proposed). A third measure was also obtained, that of the net number of relevant recommendations remaining after eliminating proposed recommendations that were irrelevant. To determine the numbers of both relevant and irrelevant recommendations proposed by each team, teams’ lists of recommendations were given to two independent coders. Both coders were graduate students with professional auditing experience. Each coder received two grading keys—one for each case. The grading keys were prepared prior to administration of the experiment and contained a list of relevant management recommendations for each case. In addition to the recommendations listed on the keys, each coder was instructed to classify as relevant other recommendations proposed by teams if, in the coder’s opinion, the recommendations were unique (i.e., not simply a prior recommendation that had been re-worded) and relevant (i.e., the coder believed the recommendation was appropriate and should be communicated to the client’s management). All teams’ lists of recommendations were masked to hide from the coders whether the list was created by a computer-mediated team or a FTF team. Analyses of inter-coder reliability revealed a correlation of 0.818. The number of management recommendations was aggregated across the two cases for each communication mode. In addition, measures of participants’ satisfaction levels with the overall team experience, satisfaction with the outcome, and perceived team efficiency were obtained as participants completed two debriefing questionnaires; one immediately after the FTF communication condition, and one immediately following the CMC condition. Responses to each question were based on a seven-point Likert scale. The questions are shown below:
How would you describe your experience when working with your group? (scale anchored with “very dissatisfied” and “very satisfied”). How satisfied were you with your group’s list of recommendations to the client? (scale anchored with “very dissatisfied” and “very satisfied”). How efficient was your group at discussing the control weaknesses and developing recommendations? (scale anchored “very efficient” and “very inefficient”).

392

KERR AND MURTHY

Table 1. Descriptive statistics for dependent variables: computer-mediated teams vs. FTF teams Dependent variables Number of all recommendations Mean Standard deviation Percentage of recommendations that were irrelevant Mean Standard deviation Number of relevant recommendations Mean Standard deviation 9.7 3.5 7.2 2.2 18.3 12.2 9.5 8.1 11.9 4.5 7.9 2.2 CMC teams Face-to-face teams

Results As discussed in the previous section, each team completed two cases using both FTF interaction and computer-mediated interaction. While the variable of interest in the research design was the interaction mode, the possibility of case effects cannot be ignored. Accordingly, case effects were investigated by analyzing the difference in the mean number of management recommendations between the case completed first versus the case completed second. The mean numbers of relevant recommendations generated by teams for the first and second cases were 7.92 and 8.92, respectively. The difference is not statistically significant ( p = 0.443). This analysis was also performed separately for FTF teams and computer-mediated teams. Again, there were no significant case effects (FTF: p = 1.00; CMC: p = 0.338). Given the absence of significant differences in the numbers of recommendations generated for case 1 vis-a-vis case 2, data from the two cases were combined in the subsequent analyses. Analysis of variance (ANOVA) was used to test each of the two research hypotheses. Descriptive statistics for each of the dependent variable are presented in Table 1, arranged in the table according to participants’ team-interaction mode. ANOVA results for CMC and FTF teams are presented in Table 2. Effect of communication mode on idea-generation performance Hypothesis 1 predicted that computer-mediated teams would propose a greater number of management recommendations relative to FTF teams. As shown in Tables I and II, computer-mediated teams outperformed FTF teams on the divergent-thinking aspect of the task, proposing a significantly greater number of management recommendations (CMC = 11.9, FTF = 7.9). This difference is statistically significant (F = 7.547, p = 0.012). Thus, Hypothesis 1 is supported. Hypothesis 2, by contrast, predicted that FTF teams would perform better at the convergent-thinking aspect of the tasks than computer-mediated teams, as measured by

DIVERGENT AND CONVERGENT IDEA GENERATION IN TEAMS

393

Table 2. ANOVA results: computer-mediated teams vs. face-to-face teams Sum of squares All recommendations Between groups Within groups Total Percentage of irrelevant recommendations Between groups Within groups Total Relevant recommendations Between groups Within groups Total 37.5 184.3 221.8 1 22 23 37.5 8.4 4.48 0.046 4.7 23.6 28.3 1 22 23 4.7 1.1 4.37 0.048 96.0 279.8 375.8 1 22 23 96.0 12.7 7.55 0.012 df Mean square F Significant value

the percentage of irrelevant recommendations proposed. As shown in Table I, computermediated teams had a mean of 18.3% irrelevant recommendations, while only 9.5% of the recommendations proposed by FTF teams were deemed irrelevant. This difference (8.8%) is statistically significant, as indicated in Table II (F = 4.371, p = 0.048). Hypothesis 2 is therefore supported. As reported above, computer-mediated teams proposed a greater number of recommendations than FTF teams, but a greater percentage of those recommendations were judged to be irrelevant. Therefore, it was also of interest to determine which communication mode resulted in the larger number of proposed relevant recommendations; i.e., whether the greater overall number of recommendations proposed by computer-mediated teams was driven by the fact that those teams proposed a larger percentage of irrelevant recommendations. As shown in Tables I and II, computer-mediated teams generated a significantly greater number of net relevant management recommendations (CMC = 9.7, FTF = 7.2). This difference is statistically significant (F = 4.476, p = 0.046). Thus, even though FTF teams performed better than CMC teams in the convergent aspect of the tasks, a greater number of relevant recommendation were preposed by CMC teams relative to FTF teams. Effect of communication mode on participants’ satisfaction levels Apart from the number of recommendations proposed by computer-mediated teams relative to FTF teams, we also sought to determine the effect of CMC use on participants’ satisfaction with their team experience and outcomes. Given the increasing interest in and availability of tools to support team work, generally categorized as groupware, the extent of user satisfaction is of interest to business organizations contemplating the use of, or currently using, CMC for collaborative work. Participants’ mean team-experience satisfaction levels, mean outcome satisfaction levels, and mean perceived efficiency levels in

394

KERR AND MURTHY

the computer-mediated condition were compared with the corresponding levels in the FTF condition. Participants in computer-mediated teams were generally less satisfied with their team experience (CMC = 5.5, FTF = 6.0 (larger numbers reflect greater satisfaction levels based on a seven-point Likert scale); t-value = 2.06, df = 45; p = 0.023), were less satisfied with their team’s list of client recommendations (CMC = 5.0, FTF = 5.9; t-value = 3.27, df = 45; p = 0.001), and believed their teams were less efficient (CMC = 3.5, FTF = 2.3 (smaller numbers reflect greater perceived efficiency); t-value = 3.93, df = 45; p < 0.001) than participants in FTF teams.

Discussion Most business organizations today have adopted some form of CMC, particularly chat-type programs, to support communication among team members in geographically dispersed locations (Menezes 1999; Lamont 2000). Prior research on CMC has demonstrated that the use of chat-type CMC systems can enhance teams’ performance in divergent ideageneration tasks relative to FTF interaction. However, it is less clear whether the benefits of chat-type CMC systems extend to tasks requiring convergent processes. This study extends prior research on the performance of CMC-mediated teams relative to FTF teams by examining teams’ performance on two highly realistic business-consulting tasks requiring both divergent and convergent processes. Consistent with the predictions of TTF theory, the results reveal that CMC-mediated teams outperformed FTF teams on the divergent aspects of the tasks examined in the current study, recommending a greater number of ideas regarding how the client’s operations and internal control systems could be improved. Results also revealed that, consistent with theory, FTF teams outperformed CMC-mediated teams on the convergent aspect of the tasks, proposing a lower percentage of irrelevant client recommendations. Further analyses revealed that although CMC-mediated teams proposed a greater percentage of irrelevant recommendations than FTF teams, CMCmediated teams proposed a greater number of relevant recommendations that FTF teams. These findings have important implications for firms contemplating the use of CMC technology to provide electronic support for the teamwork of their professionals. For tasks requiring teams to generate alternative solutions, the findings suggest that the use of a CMC chat systems results in better divergent thinking, leading to the generation of a significantly greater number of ideas relative to meeting FTF. The results revealed, however, that CMC-mediated teams were less successful than FTF teams at weeding out low-quality, irrelevant alternatives. Nevertheless, the net number of relevant recommendations proposed by CMC-mediated teams exceeded that of FTF teams. Our findings therefore suggest that consulting teams should consider using CMC chat systems to first generate as many ideas as possible, and then meet in the richer FTF communication environment to differentiate between the more-relevant and less-relevant ideas. Given that CMC can facilitate team meetings where members interact from remote locations, the costs associated with computer-mediated team interaction can be substantially lower than that for FTF meetings. This ability to support teams with geographically dispersed members, combined with the superior effectiveness of computer-mediated teams relative to FTF teams as demonstrated

DIVERGENT AND CONVERGENT IDEA GENERATION IN TEAMS

395

in this study, indicates that firms should strongly consider increasing the use of such technology. The design of our experiment did not allow us to isolate the effects of CMC on reducing specific process losses commonly associated with FTF interaction, such as production blocking and evaluation apprehension. The computer-mediated teams interacted anonymously in our study. However, since the participants were peers and there was no hierarchical power structure in the teams, we do not believe that anonymity is likely to have had a material effect on the results. Furthermore, there is evidence in the literature that the reduction in production blocking is the major factor in improving the idea-generation performance of computer-mediated teams relative to FTF teams (Gallupe et al. 1994). Future research could attempt to distinguish between the effects of CMC on reducing production blocking and evaluation apprehension. For example, Connolly, Jessup and Valacich (1990) found that groups working anonymously and with a critical confederate produced the greatest number of ideas. The task instructions in the current study informed participants that their recommendations would be evaluated both on quantity and quality. We believe that these instructions, and the analyses performed, allowed us to measure both the divergent aspect of the tasks (i.e., quantity – the total number of recommendations after eliminating duplicates) and the convergent aspect of the tasks (i.e., quality – proposing relevant high-quality, cost-effective recommendations). However, it should be noted that the teams were not specifically directed to engage first in a phase of divergent thinking followed by a phase of convergent thinking. To evaluate the robustness of our results, future research could compare the performance of computer-mediated teams and FTF teams where the meeting agenda enforces specific phases of divergent and convergent thinking. Along these lines, it should be noted that one member in the FTF teams was assigned to be the team scribe, which increased the realism of the FTF setting, but may have somewhat hampered the ability of the scribe to contribute to the team’s discussion. Although there was no group scribe in the CMC condition, only recommendations preceded with the word “reco” were counted, which to some extent leveled the playing field between the two interaction modes. While it is difficult (and not necessarily desirable) for FTF- and CMC-treatment conditions to be perfectly comparable, we do not believe the experimental procedures were biased in favor of any one treatment condition. In line with previous findings in the CMC literature, participants who met FTF were generally more satisfied with the teamwork experience and with the outcomes of their teamwork. Interestingly, although computer-mediated teams were more effective at developing relevant recommendations, participants perceived their teams to be more efficient, and were more satisfied, when interacting FTF. It is important to recognize that the participants in the current study did not have extensive prior experience with the VisionQuest system. Although CMC technology mitigates some of the process losses inherent in FTF interaction, it introduces certain process losses of its own. For example, most individuals find it easier to talk than to type. It is also important to recognize that CMC technology permits meetings that could not be conducted without the technology (different time/different location meetings). It should also be noted that all participants were students. As Connolly, Jessup and Valacich (1990) suggest, laboratory

396

KERR AND MURTHY

experimentation with student participants is a useful step in exploring the effects of IT-based GSS on idea-generation. Remus (1986) and Briggs, Balthazard and Dennis (1996) provide empirical support for the use of business graduate-student participants as surrogates for business professionals. Apart from the aforementioned implications for organizations and business professionals, we believe the findings are also relevant for educators considering the use of cooperative learning environments to enhance education and learning, as suggested by Cottell and Millis (1992, 1993). Our findings reveal that information technology support for interacting teams improved participants’ effectiveness in generating ideas, thus improving the potential for participants learning and understanding of the problem. Kerr and Murthy (1994), which focused on the effects of CMC on individual learning, found that the use of CMC-enhanced learning relative to FTF cooperative interaction. The decreasing cost and increasing availability of CMC software should spur educators to investigate the use of CMC to support group work in various academic courses.

Summary and conclusion This study examined the effects of CMC using tasks involving both divergent and convergent processes. Prior empirical research on the effects of CMC has tended to focus on simplistic tasks, rather than on realistic business-related problem-solving-oriented tasks for which organizations are increasingly using technology like LotusNotes. Based on TTF theory, this study compared computer-mediated and FTF teams in terms of their relative performance in both divergent and convergent processes. Performance was measured in terms of the number of management recommendations proposed by each team regarding how operations, including the design of internal controls, could be improved. The tasks employed were very realistic business tasks, necessitating a focus on meaningful, cost-effectiveideas rather than just “wild ideas.” In line with the predictions of TTF fit theory, participants working in computer-mediated teams performed better at the divergent aspect of the tasks, recommending a greater number of ideas, while participants working FTF performed better at the convergent aspect of the tasks, proposing a lower percentage of irrelevant recommendations. Assuming that many decisions in consulting and business settings will continue to be made by the teams, the findings of the present study indicate that computer-mediated team interaction has promising potential for improving performance relative to unassisted team interaction, at least for divergent aspects of tasks. In contrast, yet consistent with TTF theory, this study also found that FTF teams performed better than computer-mediated teams when the teams’ task was more equivocal; i.e., weeding out irrelevant ideas and recommendations. These findings have important implications for organizations such as consulting and public accounting firms that are already using CMC technology to provide electronic support for the teamwork of their professionals. Given that FTF teams were found to outperform computer-mediated teams in the convergent aspect of the tasks, the use of CMC may not be appropriate for performing highly ambiguous tasks or in situations involving conflicts of interest where it may be more difficult for the team to converge on a solution.

DIVERGENT AND CONVERGENT IDEA GENERATION IN TEAMS

397

In this study, the discussions of teams meeting FTF were not recorded. As a result, it was not possible to determine which treatment (FTF or CMC) resulted in a greater number of “lost opportunities,” i.e., good ideas that were raised by a team member but did not make the final list of proposals submitted by the team. Future research could examine this issue by recording FTF sessions and identifying high-quality ideas mentioned (typed) but rejected (not tagged) by FTF (CMC) teams. Other issues ripe for future research include investigating the effects of experience and team size on the effectiveness, efficiency, and satisfaction of computer-mediated team members when performing tasks other than idea generation. Future research could also explore whether the CMC effects observed in a “same time” meeting environment are any different in a “different time” setting, and whether differences observed (if any) are a function of the characteristics and requirements of the task. Acknowledgements We thank workshop participants at Texas A&M University and Temple University for helpful comments on an earlier version of this paper. Notes
1. The tasks typically involved student participants brainstorming on questions such as how the parking problem on campus could be resolved, how the air quality in the city could be improved, or what one would do with an extra thumb. 2. Goodhue (1995), and Goodhue and Thompson (1995) propose a theoretical model of TTF in the context of individual performance. By contrast, the Zigurs and Buckland (1998) model of TTF is couched in the context of group performance. However, the basic notion of the need for an appropriate “fit” between the technology (whether used by an individual or a group) and the task is consistent across both models. 3. Nunamaker et al. (1991) and Dennis, Wixom and Vandenberg (2001) suggest that information overload can be exacerbated by parallel communication, since participants can simultaneously input their ideas into the system. The notion that groups will have difficulty coping with a large number of ideas as they attempt to distinguish between better and worse ideas is consistent with suggestions in the information overload literature (Carey and Kacmar 1997; Orwig, Chen and Nunamaker 1997). 4. The operationalization employed in the study for both FTF and CMC conditions was aimed at mirroring realworld settings, with the use of a scribe in FTF teams and the requirement for CMC team members to “tag” comments they felt were of high-quality and cost-effective.

References
Bamber, E. M., R. T. Watson, and M. C. Hill. (1996). “The Effects of Group Support System Technology on Audit Group Decision Making,” Auditing: A Journal of Practice & Theory. 15(1), 122–134. Barki, H. and A. Pinsonneault. (2001). “Small Group Brainstorming and Idea Quality–Is Electronic Brainstorming the Most Effective Approach?, Small Group Research 32(2), 158–205. Briggs, R. O., P. A., Balthazard and A. R. Dennis. (1996). “Graduate Business Students as Surrogates for Executives in the Evaluation of Technology,” Journal of End User Computing 8(4), 11–17. Carey, J. M. and C. J. Kacmar. (1997). “The Impact of Communication Mode and Task Complexity on Small Group Performance and Member Satisfaction?,” Computers in Human Behavior 13(1), 23–49.

398

KERR AND MURTHY

Connolly, T., L. M. Jessup, and J. S. Valacich. (1990). “Effects of Anonymity and Evaluative Tone on Idea Generation in Computer-Mediated Groups,” Management Science 36(6), 689–703. Cottell, P. G. and B. J. Millis. (1992). “Cooperative Learning in Accounting,” Journal of Accounting Education 10, 95–111. Cottell, P. G. and B. J. Millis. (1993). “Cooperative Learning Structures in the Instruction of Accounting,” Issues in Accounting Education 8(1), 40–59. Daft, R. L. and R. H. Lengel. (1986). “Organizational Information Requirements, Media Richness and Structural Design,” Management Science 32(5), 554–571. Dennis, A. R.. (1996). “Information Exchange and Use in Group Decision Making: You Can Lead a Group to Information, But You Can’t Make It Think,” MIS Quarterly 20(4), 433–457. Dennis, A. R. and J. S. Valacich. (1999). “Research Note. Electronic Brainstorming: Illusions and Patterns of Productivity,” Information Systems Research 10(4), 375–377. Dennis A. R, B. H. Wixom and R. J. Vandenberg. (2001). “Understanding Fit and Appropriation Effects in Group Support Systems via Meta-Analysis,” MIS Quarterly 25(2), 167–193. Gallupe, R. B., L. M. Bastianutti, and W. H. Cooper. (1991). “Unblocking Brainstorms,” Journal of Applied Psychology 76(1), 137–142. Gallupe, R. B., Cooper, W. H., Grise, M. L., and L. M. Bastianutti. (1994). “Blocking Electronic Brainstorms,” Journal of Applied Psychology 79(1), 77–86. Gallupe, R. B., Dennis, A. R., Cooper, W. H., Valacich, J. S., Bastianutti, L. M. and J. F. Nunamaker, Jr. (1992). “Electronic Brainstorming and Group Size,” Academy of Management Journal 35(2), 350–369. Goodhue, D. L. (1995). “Understanding User Evaluations of Information Systems,” Management Science 41(12), 1827–1844. Goodhue, D. L. and R. L. Thompson. (1995). “Task-Technology Fit and Individual Performance,” MIS Quarterly 19(2), 213–236. Ho, J. (1999). “Technology and Group Decision Process in Going-Concern Judgements,” Group Decision and Negotiation 8(1), 33–49. Jessup, L. M, Connolly, T. and D. A. Tansik. (1990). “Toward a Theory of Automated Group Work: The Deindividuating Effects of Anonymity,” Small Group Research 21(3), 333–348. Karan, V., Kerr, D. S., Murthy, U. S. and A. S. Vinze. (1996). “Information Technology Support for Collaborative Decision Making in Auditing: An Experimental Investigation,” Decision Support Systems 16, 181–194. Kerr, D. S. and U. S. Murthy. (1994). “Group Support Systems and Cooperative Learning in Auditing: An Experimental Investigation,” Journal of Information Systems 8(2), 85–96. Lamont, I. (2000). “The Coolest Kind of Collaboration,” Network World 17(46); Framingham (November 13), 117–119. Menezes, J. (1999). “Knowledge Apps Hit the High End,” Computing Canada (25)27; Willowdale (July 9), 19–22. McGrath, J. E. (1984). Groups: Interaction and Performance, Englewood Cliffs, NJ: Prentice Hall. Murthy, U. S. and D. S. Kerr. (2004). “Comparing Audit Team Effectiveness via Alternative Modes of ComputerMediated Communication,” Auditing: A Journal of Practice & Theory, 23(1), 141–152. Nunamaker J. F., Dennis A. R., Valacich J. S., Vogel D. R., and J. F. George. (1991). “Electronic Meeting Systems to Support Group Work,” Communications of the ACM 34(7), 40–61. Orwig R. E., Chen H. C. and J. F. Nunamaker. (1997). “A Graphical, Self-Organizing Approach to Classifying Electronic Meeting Output,” Journal of the American Society for Information Science 48(2), 157– 170. Osborn, A. F. (1963). Applied Imagination: Principles and Procedures of Creative Problem-Solving, New York: Charles Scribner’s Sons. Petty, R. E. and J. T. Cacioppo. (1986). “Communication and Persuasion, New York, Springer-Verlag. Pinsoneault, A., Barki, H., Gallupe, R. B. and N. Hoppen. (1999). “Research Note. The Illusion of Electronic Brainstorming Productivity: Theoretical and Empirical Issues,” Information Systems Research 10(4), 378–380. Pollard C. (2003). “Exploring Continued and Discontinued Use of IT: A Case Study of OptionFinder, A Group Support System,” Group Decision and Negotiation 12(3), 171–193. Remus, W. (1986). “Graduate Students as Surrogates for Managers in Experiments on Business Decision Making,” Journal of Business Research 14, 19–25.

DIVERGENT AND CONVERGENT IDEA GENERATION IN TEAMS

399

Satzinger, J. W., Garfield, M. J. and M. Nagasundaram. (1999). “The Creative Process: The Effects of Group Memory on Individual Idea Generation,” Journal of Management Information Systems 15(4), 143– 160. Shirani, A. I., Tafti, M. H. A. and J. F. Affisco. (1999). “Task and Technology Fit: A Comparison of Two Technologies for Synchronous and Asynchronous Group Communication,” Information & Management 36(3), 139–150. Steiner, I. D. (1972). Group Process and Productivity, San Diego, CA: Academic Press. Valacich, J. S., Paranka, D., George, J. F. and J. F., Nunamaker Jr. (1993). “Communication Concurrency and the New Media: A New Dimension for Media Richness,” Communication Research 20(2), 249–276. Valacich, J. S., Dennis, A. R. and T. Connolly. (1994). “Idea Generation in Computer-Based Groups: A New Ending to an Old Story,” Organizational Behavior and Human Decision Processes 57(3), 448–467. Zigurs, I. and B. K. Buckland. (1998). “A Theory of Task-technology Fit and Group Support Systems Effectiveness,” MIS Quarterly 22(3) 313–334.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.