Vous êtes sur la page 1sur 34
Polarization & Persuasion Appendix E ‘The spread of answers was analyzed for each task on individual prediscussion preferences. This spread indicated a pronounced difference between the two tasks, illustrating that the task manipulation was successful. In particular, for the intellective task, 78% of the response is clustered on one point of the scale, For the judgmental task, hawever, responses are more spread out with responses scattered throughout the scale without a clear clustering point. See the histograms below. | Pentium Problem - Frequency Percentage Chart Decisions | 0.00% 20.00% ~ 40.00% 60.00% 80.00% Percentage Business Strategy - Frequency Chart Decisions =neeuas 0.00% 10.00% 20.00% 30.00% 40.00% Percentage 198 AIS Quarterty.June 1998 Polarization & Persuasion Appendix D Coding Rules and Sample Excerpts Rules for Content-Coding Transcripts for Persuasive Arguments ] i Categories of PAT | Argument Attributes Definitions Used for Content-Coding Valighty “Truth - Staternent is supported by the parameters defined by the problem orthe statement is supported by the documentation provided to the participants or proof of truthfulness of the statement is provided by participant arthe statement is atherwise verified _ + Fit + Staterent fits {is in line with) views previously expressed by the subject or the statement fits the ‘current discussion thread + Follow ~ Staternent follows trom accepted facts ortollows previously expressed views by the subject * Contribute ~ Staternent supports, represents, or uses in some form ‘ane of the seven options provided as the final decision] Novelty + New way - Statement indicates a new form of organizing the information, * New idea ~ Statement provides information not previously used to conduct the discussion, Selective Excerpts From a Discussion Related to the Pentium Problem Categories of PAT Truth Fit Fallow Contribute New way Selective Protocols From Verbalizations * By hiding the Pentium problem, they (intel) only made themselves look worse, especially when IBM called them on itnationally + The exact problem with the chip was a division error in the 16th decimal place, + Only 1% of the Pentium user population is impacted by this. + The consumers’ trust in Intel has been hindered because of the Pentium incident. Corrective measures must be taken to re-establish the consumer's, perception. * Things like this will eventually alfect our {rust in the company, + Yes, itis a matter of goodwill, = Tithink they should be very... very open with the public. .. as for the chip problem, ask fora voluntary recall for those that are affected “Wouldn't it be botter just (o replace the entire computer. + Pertaps intel should focus on the fact that thay are stil a reputable chip manufacturer and side-step the Pentium issue altogether. + Focus on... Intel is an innovator/leader. + Concentrate cn maintaining status in computer industry + Design marketing campaign to admit mistakes: “Intel we fix our mistakes.” MIS Quaceriyllune 1988 197 Polarization & Persuasion Results of ANCOVA Dependent Variable: Polarization Sum of Mean Source DF | Squares | Square | FV¥alue | Pr>F Medium 1 39.35 89.35 6.7, O01 Yes Group 2 215, 1.08 0.08 0.92 No Task 1 99.12 49.12 6.95 O01 Yes IAI (covariate) 1 3.97 9.97 0.26 0.61 No Med*Group 2 24.48 12.24 0.96 0.98 No Med'Task 1 an7i 774 6.84 0.012 Yes Task*Group 2 17.50 875 0.68) 0.51 No Med'Task'Group| 2 O24 O12, O07 0.99 No 196 MIS Quarteriy(une 1998 Polarization & Parsuasion Appendix C Rationale for a Two-Group Clas: ication In deciding the classes af groups based on their group composition, three types of groups were consit ‘ered, The first category was one that included no individuals from the *assertive-cirecting” type. The secand category of groups included those that had a single assertive-directing member, The third group had two assertive-directing members. These three classifications covered all of the groups: included in this study, ‘.e., there were no groups that included mare than two assertive-cirecting individ. wale in the group. ‘The aim in classifying the groups was 10 find the two exiremes that would allow investigation of the hypatheses, However, veritying whether the presence of more than one assertive-directing individual could have consequences an either group behavior or outcomes was also of interest. Of particular interest was checking whether groups with a single agcertive-directing individual behaved differently when compared with groups with more than one assertive-directing individual. The analysis suggests ‘that in fact there is no discemible difference between groups with a single or mere than one assertive- directing group member. The basis for comparison was the degree to which each of these groups polarize. A t-test was conducted to determine differences in group performanee for group types 2 and 3, The t-test to check whether these two groups are equal resulted in a T-statistic of -.028 and Pr > IT| of 0.77 (standard error of estimate was 2.007). An ANCOVA was conducted using group composition bath as a two-level (including only group types 1 and 2) and a three-level (including group types 1, 2. and 3) variable and found minimal and non-significant differences in the level and direction of polariza- tien, Given these results, the decision was made to club groups with one or two assertive-directing members (group types 2 and 3) into one of the classifications (labeled assertive-directing) we then ‘compared these groups With those that had no-assertive-dieecting members (labeled flexible-cohering) This permitted the study to proceed with a two-level classification of group composition rather than a three-level classification Overall Model Fit Three-Group Classification Dependent Variable: Polarization sumot | Mean | | Source DF Squares Square | F Value Pro Significant? Model 10 ‘307.81 so7e | 240 | 0.019 Yes Error sa | esz10 | 1282 | ComectedToial | 64 | 999.91 MIS Quartedy.lune 1998 185 Polarization & Persuasion Appendix B The Pentium Problem When ward about a flaw in the Inte!’s new Pentium chip began circulating before Thanksgiving 1994, was viewed by Intel, initially, as an inconvenience. Having kept the defect a secret for months, Intel December by IBM's decision to halt shipments of its machine containing the ‘was caught unprepared i Pentium chip. If you were on the baard of directors of Intel advising the CEO, Andrew Grove, what issues would you advise Mr. Grove to consider in developing Intel's policy to deal with the Pentium case. What issues would you cansider critical? Characterize your decision on the following scale: 1 | Complete denial ofthe existence of the problem 2 | Agree that problem MAY exist in a few cases, but indicate that i is noncritieal and thu unimportant. 3 | Agree that problam MAY oxist in a few eases. If justified, send a new Pentium chip. 4 | Testallchips at Intel site. Replace chip if awed. Otherwise retum chip. 5 | Replace Pentium chip only & | Gomplete computer replacement, no restitution. 7 | Complete product recall, replacement of computer pius restitution 194 MIS Quartenyllune 1998 Polartzabon & Porsuasion Appendix A Business Strategy Intel has traditionally pumped megabucks into product development and factories and thus has estab- lished itself as the hardware hegemon of the PC revolution, much as Bill Gates has made Microsoft the software champ. Recent developments in the consumer market for PCs has Intel convinced that, before long, consumers and not corporate customers will keep Intets riches flowing. As a result, Intel needs to evaluate its business strategy. As a consultant advising Andy Grove on this matter, what Intel plots its future direction? sues do you feel should be considered as. ‘What issues would you consider critical? Characterize your decision on the following scale: Pursue your interests strictly as an engineering company. Direct most resources to impraving process elficiency. Direct most resources to product enhancement. Balanced mix of improved process and introducing neve products. Direct most resources to developing new products to meet identilied customer needs Direct most resources to addressing and understanding customer needs without changing the product or the process. Pursue your interests stiictly a8 a consumer-oriented company. MIS Quarterlyllune 1998 183 Poterization & Persuasion Research: Exploring the Black Box,” MIS ‘Quarterly (11:4), 1987, pp. 493-8 12. Valacich, J. S., Paranka, D., George, J. F., and Nunamaker, J. F. “Communication Concurrence and the New Media,” ‘Communication Research {20:2}, 1993, pp. 249-276, Van Someren, M. W., Barnard, Y. F., and Sandberg, J. A. ©. The Think Aloud Method: A Practical Guido ta Modelling Cognitive Processes, Academic Pross, New York, 1994, Vidmar, N. "Effects of Group Discussion on Category Width Judgment,” Journal of Personality and Sociat Psychology (29), 4974, pp. 187-105. Vinokur A., and Burnstein, E, “Novel Argumentation and Attitude Chango: Tho Case of Polarization Following Group Discussion," European Journat of Social Psychology (8), 1978, pp. 335-348, Vinokur A., and Burstein, E. “Effects of Partially Shared Persuasive Arguments on Group-Induced Shifts: & Group-Problem- Solving Approach." Journal of Personality and Social Psychology (29), 1974, pp. 305-315. Weisband, 8. P. “Group Discussion and First Advocacy Effects in Computer-Mediated and Face-to-Face Decision Making Groups,” Organizational Behavior and Human Decision Processes (53). 1992. pp. 952-380. Zigurs. |., Poole, M, S., and OeSanetis, G. “A Study of Inluence in Computer-Mediated Group Decision Making,” MS Quarterly (12:4), 1988, pp. 625-644, Zuber, J. A., Cratt, H. W., and Werner, J "Choice Shift and Group Polarization: An Analysis of the Status of Arguments and Sosial Decision Schemes,” Joumal of Personality and Social Psychology (62) 1982, pp. 80-61 182 MIS Quartenyune 1998 About the Authors Maha El-Shinnawy is assistant professer of information systems in the Mays College and Graduate School of Business Administration at Texas A&M University. She received her Ph.D, in information systems from the Anderson School at UCLA. Her research inter- ests include the patterns of adoption and usage of new communication media and the impacts of information technology on individu- als, groups. and organizations. She has pub lished her research in outlets such as The Information Society, The internationat Jounal of Huma Computer Studies, and the Proceedings of the Intemational Conference on Information Systems, She currently serves as associate editor on the editorial board of IEEE Transactions on Professional Communication, Ajay S. Vinze is an associate professor of management information systems in the School of Accountancy and Information Management at Arizona State University Priot to joining ASU, he sétved on the MIS faculty at Texas A & M University, He received his Ph.D. in MIS from the University of Arizona, Tucson, in 1988, His research interests include business applications of artificial intelligence technolagy and the sludy of computer-supperted collaborative work. His publications have appeared in leading MIS journals like Information Systems flesearch, Decision Sciences Journal of Management Information Systems, IEEE Transactions on Systems, Man and Cybemeties, Omega, International Journal of Man-Machine Studies, and Information & Management. Betare joining the academic environment. he was an MIS consultant based in the Philippines. He is @ member of the Assaciation of Gomputing Machinery, The Institute of Management Sciences, the American Assaciation of Artificial Intelligence, and the IEEE Computer Society. Myers, D. G., and Lamm, H. “The Group Polarization Phenomenon.” Psychological Bulletin (83), 1976, pp. 602-627. Neter. J., Wasserman, W., and Kutner, MH. Applied Linear Statistical Models: Regression, Analysis of Variance, and Experimental Designs, Irwin, Homewood, Hk, 1990. Nunamaker, J. F., Jr., Applegate, L., and Konsynski, B, "Facilitating Group Creativity: Experience With a Group Decision Support System,” Journal of Management Information Systems (3:3). 1967, pp. 5-18, Nunamaker, J. F,, Vogel, D. R., Heminger, A Martz, B., Grohowski, R., and McGoff, C “Experiences at IBM With Group Support Systerns: A Field Study." Decision Support Systems (5:2). 1989, pp. 183-196. Nunamaker, J. F., Dennis, A: R., Valacich, J. S., Vogal, D. R., and George, J. F “Electronic Meeting Systems to Support Group Work: Theory and Practice at Arizona,” Communications of the ACM (34:7), 1991, pp. 40-61 Pavil, C, “Another View of Group Polarizing: The ‘Reasons for’ One-Sided Oral Argumentation,” Communication Research (21:5), 1994, pp. 625-642 Pinsonneault, A, and Kraemer, K, L, “The Impact of Technological Support on Groups: An Assessment of Empirical Research," Dacision Support Systems (5). 1989, pp. 197-216. Poole, M. S., Seibold, D. R., and McPhee R. D. “Group Decision Making as a Structuration Process," Quarterly Journal of Speech (71). 1985, pp. 74-102 Poole, M. S., Holmes, M., and DeSanctis, 6. “Gontlict Management in a Computer- Supported Meeting Envirenment," Management Science (37:8), 1991 pp. 926-953. Porter, E. H., and Maloney, 8. E. The Strength Deployment inventory. Personal Strengths Publishing, Inc,, Pacific Palisades, CA, 1989. Pruitt, D. G. “Choice Shifts in Group: Discussion: An Introductory Review." Journal of Personality and Social Psychology (20:3), 187 1a, pp. 389-260 Pruitt, D. G. “Conclusions: Toward an Understanding of Choice Shifts in Group. Polarization & Persuasion Discussion,” Journaf of Persanality and Social Psychology (20:3), 197 1b, pp. 495-510. Reder, S., and Cankiin, N. F. “Selection and Effects of Channels in Distributed Communication and Decision-Making Tasks: A Theoretical Review and a Proposed Research Paradigm,” paper pre- sented at the Thirty-Seventh Annual Conference of the Intemational Communication Association, Montreal Canada, 1997 Sambamurthy. V.. and Poole, M.S. “The Effects of Variations in Capabilities of GDSS Designs on Management of Cognitive Contlict in Groups,” information Systems Research (3:3). 1992. pp. 224-251 Shavelson, A. J. Statistical Reasoning for the Behavioral Sciences (2nd ed.), Allyn and Bacon Inc., Needham Heights, MA. 1988. Sia, C. L., Tan. C. Y.. and Wei, K. “Will Distributed GSS Groups Make More Extreme Decisions? An Empirical Study.” in Proceedings of the Seventeenth Inter- national Conference on Information Systems, J. |, DeGrass, S, Javepnaa, and A. Srinivasan (eds.), Cleveland, OH, 1996, Pp. 326-398, Siegel, J.. Dubrovsky, V., Kiesler, $., and MeGuire, T. W. ‘Group Processes in Gomputer-Mediated Communication,” Organizational Behavior and Human Decision Processes (37), 1986, Stoner, J. A. F. “A Comparison of Individuat and Group Decisions Involving Risk.” unpublished master's thesis, MIT, Schoo! of Industrial Management, Cambridge, MA. 1961 8, C,, Wei, K.. and Watson, A. T. Dampening Status Influence Using a Group Support System: An Empirical Study.” in Proceedings of Fourteenth International Conference on Information Systems, J}. DeGross, B. P. Bostrom, and D. Robey (eds.), Orlando, FL. 1993, pp. 7-88. Toch, R. The Social Psychology of Sociat Movements, Bobbs-Mertil, Indianapélis, IN, 1965 Todd P., and Benbasat, |. "Process Tracing Methods in Gecision Support Systems MIS Quosteryésune 1998 191 Polarization & Persuasion Isenberg, D. J. "Group Polarization: A Critical Review and Meta Analysis,” Journal of Personality and Social Psychology (506), 1986, pp. 1141-1151 Jarvenpaa, S. L., Flag, V. S., and Huber, G. P, "Computer Support for Meetings of Medium-Sized Groups Working on Unstructured Problems: A Field Experiment," MIS Quarterly (12:4), 1989, pp. 645-666. Jessup, L. M., Connolly, T., and Galeghar, J. ‘The Effects ot Anonymity on GSS Group Process with an Idea-Generating Task,” MIS Quarterty (14:3), 1990, pp. 313-321 Kaplan, M. F., and Miller. C. E. “Group Decision Making and Normative Versus Informational Influence: Effects of Type of Issue and Assigned Decision Rule.” Journal of Personality and Social Psychology (53), 1987, pp. 306-313, Kelley, H. H., and Thibault, J. W. "Group Problem Solving,” In Handbook of Sociat Psychology (2nd ed.) (4), G. Lindzey and E. Aronson (ads.), Addison-Weslay Cambridge, MA, 1969. Kiesler, S., Siegel, J., and MoGuire, T. W. “Social Psychological Aspects af Computer: Mediated Communication.” American Psychologist (39:10), 1984, pp. 1123-1134 Kirk, R. E. Experimental Design (2nd ed.) Grooks‘Gole, Belmont, GA, 1982. Kirkpatrick, D. "Intel Goes for Broke,” Fortune, May 16, 1994, pp. 62-68. Kogan, N., and Wallach, M, A, “The Risky-Shitt Phenomencn in Small Decision Making Groups: A Test of the Information- Exchange Hypothesis,” Journal of Experimental Social Psychology (3), 1967. pp. 75-85. Kovacs, 2. “Group Decision Making Under Uncertainty: Group Sucture and the Shift Phenomenon,” in Contemporary Problems in Group Decision Making, H. Brandstatter, J. H. Davis, and G. Stoeher-Kreichgauer (eds.), Academic Press, New York, , 1982, pp. 201-214. Lamm, H. “Will an Observer Advise High Risk Taking After Hearing a Discussion of the Decision Problem?" Journal of Personality and Social Psychotogy (6). 1967. pp. 487-471, 190 MIS Quarterlyilune 1998 Lamm, H. “A Review of Our Research on Group Polarization: Eleven Experiments on the Effects of Group Discussion on Risk Acceptance, Probability Estimation, and Negotiation Positions,” Psychological Feports (62). 1988, pp. 807-813. Lamm, H., and Myers D, G. "Group-Induced Polarization of Allitudes and Behavior.” in Advances in Experimental Social Psychology (2), L. Berkowitz (ed.). Academic Press, New York, 1978 pp. 148-198, Laughlin, P, R. "Social Combination Processes of Cooperative Problem Solving Groups on Verbal Intellective Tasks.” in Pragress in Sociat Psychology, M. Fishbein (ed.}, Erfbaum, Hillsdale, NJ, 1980, pp. 127-185. Laughlin, P. A., and Earley, P. C. “Social Combination Models, Persuasive Arguments Theory, Social Comparison Theory. and Choice Shift,” Journal of Personality and Social Psychology (42). 1982, pp. 273-280, Laughlin, PR, and Ellis, ALL ‘Demonstrability and Social Combination Processes on Mathematical intellective Tasks,” Journal of Experimental Social Psychology (22), 1986, pp. 177-189. Lim, L. H.. and Benbasat, I. “A Theoretical Perspective of Negotiation Support Systems," Journal of Management Iniormation Systems (9:3), 1993, pp. 27-44 MeGrath, J. E. Groups: Interaction and Performance, Prentice-Hall. Englewood Cliffs, NU, 1984. MeGuire, T. W.. Kiesler, $.. and Siegel, J “Group and Computer-Mediated Discussion Effects in Flisk Decision Making.” Journal of Personality and Social Psychology (82'S). 1987, pp. 917-930. Muehieman, J. T., Bruker, C., and Ingram, C. M. “The Generosity Shift,” Journal of Persanaity and Social Psychology (34:3), 1976, pp. 344-251. Myers, D. G. “Polarizing Effects of Social Interaction," In Contemporary Problems in Group Decision Making, H. Brandstatter, J H. Davis and G, Stocher-Kreichgaucr (eds), Academic Press, New York. 1982. Pp, 125-161 Benbasal, |., and Lim, LH. "The Effects of Group, Task, Context and Technology Variables on the Usefulness of Group Support Systems,” Small Group esearch 1244), 1993, pp. 430-462 Brawn, R. Social Psychofagy, Froo Pross, New Yark, 1965. Burstein, E. “An Analysis of Group Detision Involving Risk (the “Risky Shift.” Human Refations (22), 1969, pp. 381~396, Burstein, E, "Persuasion as Argument Processing,” in Group Decision Processes, M. Braandstatter, J. H, Davis, and G Stocker-Kreichgauer (eds.), Academic Press, London, 1982, pp, 109-122 Buller, J. K., Jr., and Crino, M. 0. “Effects of Initial Tendency and Real Risk 9a Choice Shifi.” Organizational Behavior and Human Decision Processes ($3), 1992, pp. 14-4 Connolly, T., Jessup, L. M.. and Valacich, J. S. “Etiects of Anonymity and Evaluative Tone on Idea Generation in Computer-mediated Groups," Management Science (36:6). 1990, pp, 689-703. Dennis, A. F., George, J. F.. Jessup, L. M., Nunamaker. J. F., and Vogel, 0. A. “information Technology to Support Electronic Meetings," MIS Quarterly (12:4), 1888, pp. 581-624. Dennis, A. R., Haley, B. J., and Vandenberg, A. J. "A Meta-Analysis of Effectiveness, Efficiency, and Participant Satistaction in Group Support Systems Research,” in Proceedings of the Seventeenth Intemational Conlerence on Information Systems, J. |. DeGross, 8. Jarvenpaa, and A. Srinivasan {eds.), Cleveland, OH, 1996, pp. 278-289. DeSanctis, G., and Gallupe, R. B. "A Foundation for the Study of Group Decision Support Systems,” Management Science (93:5), 1987, pp. 589-609. Diehl. M., and Stroebe, W. “Productivity Loss in Brainstorming Groups: Toward the Solution of a Riddle,” Journar of Porsonatty and Social Psychology (83:3), 1987, pp. 497-509. Drucker, P. F. “The Coming of the Naw Organization," Harvard Business Review (1), 1988, pp. 45-53, Ebbesen, €. B., and Bower, A. J, “Proportion of Risky to Gonservative Arguments in a Polarization & Persuasion Group Discussion and Choice Shifts,” Journal of Personality and Social Psychology (29), 1974, pp. 318-327. Ellis, C. A., Gibbs, $, J.. and Rein, GL. “Groupware: Some Issues and Experiences,” Communications af the ACM (34:1), 1991, pp. 38-68 Ericsson, K. A., and Simon H. A. Protocol Analysis: Verbal Reports as Data (Revised Edition), MIT Press, Cambridge, MA, 1883 Fjermestad, J., Starr, R. H., and Turotf, M, “An Integrated Framework for the Study of Group Decision Support Systems," Proceedings of the Twenty-Sixth Annual Hawaii International Conference on System Sciences (4, 1993, pp. 179-108. George, J. F., Easton, G. K., Nunamaker, J. F., and Northcraft, 6. 8. “A Study of Collaborative Group Work With and Without Computer-Based Support.” Information Systems Research (1:4), 1990, pp. 394-415, Gersiek, C. G, “Time and Transition in Work Teams: Toward @ New Model of Group Development," Academy of Management Journal (31:1), 1988, pp. 9-41 Green, D., and Conelley, E. "Groupthink and ‘Watergate,” paper presented at the annual meeting of the American Psychological Association, 1974. Heath, C., and Gonzalez, R. “Interaction with ‘Others Increases Decision Confidence but Not Decision Quality: Evidence against Information Collection Views of Interactive Decision Making,” Organizational Befavior and Human Decisfon Processes (61:3), 1998, pp. 905-926 Hinsz ¥.B., and Davis J. H. “Persuasive Arguments Theory, Group Polarization, and Choice Shilts,” Journat of Personality and Social Psychology (10:2), 1984, pp. 260-268. Hof, A. D. “The Education of Andrew Grove,” Business Week, January 16, 1995, pp. 60-62 Hollingshead, A. B., MeGrath, J. E., and O'Connor, K. M. “Group Task Pestormance and Communication Technology: & Longitudinal Study of Computer Mediated Versus Face-to-Face Work Groups,” Smail Group Research (24:3}, 1993, pp, 307-333. MIS Quarteryidune 1998 189 Polarization & Persuasion calls for leadership, expertise, or persuasive: hess from a certain set of individuals to forge congensus around an extreme and possibly unpopular outcome, GSS is clearly not the medium to consider. However, when a democ: atic process and outcome is desirable, GSS may be a more appropriate choice. ‘These results also open up multiple avenues for fulure. research. One such avenue is relat ed to the nature of persuasive arguments Sirtlarities have been documented in the par: Suasiveness of argument pools at the aggre: gate level; however, distinctions have also been found that can be made at a finer level of granularity. The rules that have been devel: ‘oped will allow researchers to further pursue this line of inquiry. Future researchers wauld be well advised to use process tracing meth- ‘ods to further study persuasiveness by exam- ining the elements that comprise it, what affects the nature of these elements, and final ly the impact of each of these elements on the ‘group outcome of polarization. ‘Another avenue for future research is the velar tive importance of persuasive arguments vs. ‘social comparison in explaining group polariza- tion in a contemporary context. Such a com- parison has been attempted previously in FTF settings (Butler and Crino 1992), While same group polarization researchers have clearly indicated that persuasive arguments is the dominant explanation for group polarization the introduction of GSS in a group decision making situation also has significant effects on social comparisan processes. This Is illustrat- ed in the results reported here with respect to the high level of polarization for judgmental tasks in FTF seltings vs. the very low levels in GSS settings. Because judgmental tasks have been shown to rely on social comparison, the reduction in normative influence resulted in a dramatic reduction in polarization. This clearly reopens the debate regarding the relative explanatory power of different theoretical per- spective and is another interesting future teseaich area for group polarization and GSS researchers. The results provide contributions to re searchers in the aréas of group polarization 188 MIS Quarteriytiune 1988 and GSS, Researchers in the area of group polarization need to re-examine how groups function in a more contemporary context. The ‘theoretical underpinnings of group polarization need to be revisited in order to understand the changes in the nature of normative and infor mational influences and their impact on group polarization. Technology reduces the role of normative influence in a group setting, causing participants to focus on the information content rather than interaction dynamics; alters the nature but not necessarily the extent of the persuasive arguments exchanged; and changes the extant of the decision outcome. The results also provide impetus for GSS researchers to examine the phenomenon of group polarization as an important outcome of group decision making. This study is an initial step in bringing together these two streams of research that build on each other and also influence one another. Acknowledgements This research is partially unded by the Center for Management of Information Systems (CMIS) at Texas AEM University. We would like to thank the Center as well as the review- ‘rs, associate editor, and especially the editor- inchiel, Bob Zmud, for their helpful feedback and suggestions. References Ackolf, A. L. "Management Misinformation Systems,” Management Seionce (14:4), 1967, pp. 18-21 Ancona, D. G., and Nadler. 0. A. “Top Hats and Executive Tales: Designing the Senior Team,” Sloan Management Review, Fall 1999, pp. 19-28. Baron, A. S., and Roper. G. “Reaffirmation of Social Comparison Views of Choice Shifts: Averaging and Extremity Effects in Autokinetic Situation,” Journal of Personality and Social Psychology (33) 1976, pp, 521-590, variations in task characteristics will cause cif- ferences in the process and outcomes of group decision making. To generalize these Tindings, future researchers shoukd examine a wider spectrum of communication media and tasks since differences may exist between classes of media, within types of GSS. as well as among tasks. Moreover, with respect to group composition. the work presented here Used a unidimensional approach to classitying groups by focusing on interpersonal styles The lack of findings suggests that future researchers should examine other facets such as environmental and task related factors when operationalizing the group composition construct. Furthermore, an experimental sot ling provided the appropriate environment to examine the questions posed. While there is no reason to expect different results in a simi- lar setting, caution shauld be exercised in making generalizations, Conclusions The objective of this paper was to study group polarization. Building on prior efforts by group polarization researchers and integrating work in the area of group support systems, the study sought to examine and explain the group polarization phenomenon in a contemporary context. The model used to study group polar: ization and the process that precedes it uses variables (task characteristics and group com: position) from the group polarization literature and intwoduces technology as @ new variable affecting group dynamics. While the results did ot support all of the hypotheses, there were several interesting findings related to the medi um of communication and its interaction with task in explaining the process and outcomes of group decision making, The results have implications for bath practi tioners and researchers alike. Historically. decision makers have been concerned with the tendency of team-based decisions to become more extreme than individual-based decisions. The resuls of this study confirm the reality of polarization in graup settings. More interestingly, the results also indicate that Polarization & Persuasion sometimes GSS may dramatically reduce the level of group extremization. For practitioners, these results are particutarty significant in situ- ations where a democratic process of decision making is desirable. In such situations. GSS allows the group fo reach consensus that is, mote representative af individuals’ collective judgment rather than the extreme views of a minority faction, Clearly in organizations there can be situations where managers would pre- fer a democratic outcome: in such situations, GSS could be @ medium of choice, On the other hand, law levels of polarization may not always be a positive outcome. For example, in some situations, the democratic process of reaching group consensus may not be the most effective, especially in situations where it ig necessary to draw and identify individual member expertise and allow such expertise to dictate the decisional outcome. Decision mak- ers should thus identify their particular situa- tion with the knowledge that the choice of com- munication medium may alter the outcome. The study of the process of group decision making in terms of persuasiveness is a major contribution of this paper. Protocol analysis, while a painstakingly detailed method of analy- sis, proved to be a useful technique in allowing the process to be unraveled, Pratocol analysis revealed an intriguing result: the level of per suasiveness was not as different as was expected in the context of the dramatically ait ferent levels of polarization reported. Since ersuasiveness is considered a primary deter hinant of group polarization, this result is par- ticullarly curious. For practitioners considering GSS as a means for supporting teams in orga- nizations, this medium offers a hast of bene- fits. However, the data indicate that GSS is not as persuasive a medium of communication as its ETF counterpart. While the findings do not refute previous research efforts that have shown GSS to produce a larger number of alternatives. this study indicates that @ larger number of alternatives does not necessarily translate into a more persuasive stream of arguments. in general, GSS settings do not provide greater parsuasiveness or polarization than FTF settings, This finding is of importance to practitioners in determining their choice of communication medium, When the situation MIS Quarterlyilune 1988 187 Potanzation 4 Persuasion style of the participants while controlling for the environmental (group size and history) and task-related factors (one-time task). The find: ings do not rule out the importance of group composition; rather, the findings or lack theract suggest to future GSS researchers that factors Controlled for in this experiment, such as group size and group history, may have a more sig- nificant impact an group. process and aut- comes than the factors that were manipulated, Fulure researchers would be well served by including other environmental or task-related factors when addressing the group composi: tion construct. In assessing the 1ask-medium interaction, a significant interaction was observed for bath persuasiveness and polarization. The results indicate that the medium of communication moderates the relationship between task char- acteristics and both group polarization and persuasive arguments. A relationship that is interesting to note from the results is that there was significantly less variance in either per- suasiveness or polarization in the GSS setting by task, In FTF situations, however, there were considerably larger differences in bath the process and autcome by task. GSS reduces the differences that researchers In group deci sion making have previously noted in process and outcomes by task, While the study did not predict or hypothesize such an effect for GSS, this finding is notewarthy for researchers inter ested in group polarization. As predicted in hypothesis Ha, in GSS set- fings, groups engaged in intellective tasks were able to bring the correct answer to the table and reach agreement with a less persua- sive argument pool. GSS is a more efficient wm for allowing individuals to bring their viewpoint, especially it itis the correct one, ta, the forefront without the distortions that are introduced by the normative pressures of a FTE situation. For judgmental tasks, there was ‘no significant difference in the level of persua- siveness in the argument pool. However, one ‘of the most dramatic findings in this study the large difference in polarization between GSS and FTF groups for judgmental tasks. Judgmental tasks entailed extremely high lev- ‘els of polarization in FTF settings and yet sus- 186 MIS Quarierlyilune 1998 tained very low levels of polarization in GSS settings. With similar levels of persuasiveness for both media, GSS was able to significantly reduce the level of polarization for judgmental tasks. A clear explanation for this finding is based on previous research which indicates that polarization for judgmental tasks is @ fune- tion of normative influences rather than infor- mational influences (Laughlin and Earley 1982). When these judgmental tasks are removed from the rich environment atfarded by the FTF medium into an anonymous GSS set ng where normative influences are subdued, ‘a dramatic drop in the degree of polarization ‘was noted, This effect was not as pronounced for intellective tasks since these tasks are not dependent on normative influences. These findings suggest that the type of task ‘groups are engaged in is critical when making choice regarding the medium used to com- plete the task. For judgmental tasks, groups. need to carefully examine their choice since the medium used may result in decisions that may be dramatically different from the initial majority view. In these situations, the outcome ‘of the group decision making process changes greatly depending on the level of normative and social influence present. In cages whore it is desirable to. allow the social and normative influences to flourish and guide the group in a certain direction, the traditional FTF medium, as has been illustrated, will lead to such out- comes. On the other hand, where consensus building and a democratic autcome ie impar- tant, then GSS would be the more appropriate medium, This study focuses on examining the effect of medium, task characteristics, and group com- position an group polarization, In conducting the experiment. the study was restricted to ‘only two levels for each condition examined, GSS was considered representative of new group communication media and FTF was equated to the more traditional mode of com- munication. it should be noted, however, that there are numerous technologies avaliable ta Support group decision making and ather more traditional communication media. In addition, this experiment considered oniy two tasks. Prior researchers have indicated that rather than an extreme position. The findings also support the contention that anonymity, enables group members to feel protected to.a large extent from the social pressures of face- toeface settings (Kiesler et al. 1984). As such, GSS reduced normative influence and increased people's consideration of the major- ity view instead of being persuaded by a sin- ‘gle dominant member, thus resulting in lower levels of polarization The second question related to whether the persuasiveness of arguments precoding the group decision is different when comparing FTF and GSS groups. Interestingly, the GSS. selting did not result in a larger degree of per: suasiveness of arguments as had been pre- dicted. There was no difference in persuasive: ness at an aggregate level by medium of com fnunication. Further analysis of the data at a finer level of granularity, i.e., by the two. dimansions of novelty and validity, ilustrates. that the same features of GSS that may be allowing individuals to contribute more total arguments and enhance the lavel of creativity of those arguments may simultaneously be causing a reduction in the validity of the poo! ‘of arguments. Specilieally, the findings reveal that creativity of arguments is indeed higher in a GSS setting, but validity of arguments is the same in both settings (validity: F = 0.08, p < 0.8032; novelty: F = 7.65, p < 0.0077). This resuls in a similarity in the aggregate level of persuasivoness despite differences in specific elements of the persuasiveness of arguments. Studying persuasiveness al an aggregate lavel assumes that the different dimensions of ersuasiveness occur in unison and affect the ‘outcome in the same way, The study suggests that, while it may have been appropriate to study lovel of persuasiveness. as a good indi cator of polarization in FTF settings, the nature ol persuasiveness may be just as important when considering this process in 2 GSS context With similar levels of persuasiveness, the sig: nificant reduction in polarization in GSS set: tings is intriguing, One explanation for this reduction in group polarization could be that the nature ol the arguments exchanged in a GSS setting is inherently different from that in Polanzanon & Parsuasian a FIF seiting as discussed above. On the other hand, recall that in posing the research questions, two principal forces that cause polarization—informational and normative influences—were indicated. @ reduction in either of these forces would result in a corre sponding reduction in group polarization (lsenterg 1986), In FTF settings, groups expe- rience the full range of normative and informa: tianal influences. In GSS settings, however, while social comparison is not totally efirminat- ed, normative impacts are greatly reduced causing a lessening in the group polarization experienced. In addition, this reduetion in social pressure experienced may have also Caused individuals to rely more on the argu- ments and information presented rather than simply on the initiator of the argument. By being receptive to all comments made, and focusing on the content of the discussion rather than the individuals engaged in the dis cussion, GSS groups may converge toward the average of initial individual decisions rather than being swayed by dominant others, How does "communication medium” maderate the relationship of group composition or task characteristics with group polarization and per- suasiveness of arguments? In the study, medi- um did nat moderate the relationship between group composition and the dependent vari: ables. The lack of findings related to group ‘composition in the study was surprising since prior meta-analyses studies {Benbasat and Lim 1983; Dennis et al. 1996) have reported significant effects for group composition from the perspective of group size, group history. and the existence of formal hierarchy withia groups. CCiassitying groups is a complex task. In prior ‘GSS work, three classes ol factors have been ‘considered in defining group composition: inal. vidual attributes such as interpersonal styles, personally traits, motivations, and oxporionce: eavitonmental altributes such as group size, history, location, and context; and factors associated with the immodiate task or goal such as whether the task is ongoing or a one- time task, formal or informal. This study etassi- fied the groups based on individual attributes and, in particular, focused on the interpersonal AMS Quaneryésune 1998 185 Polarization & Persuasion Medium eT hterective a Judgmental | oss | + susgrentat| a etectve Figure 3. Medium*Task Interaction (Polarization) group polarization in both traditional face-to- face settings as well as those mediated by group support systems. It posits that the medi um of communication may interact with task and group composition to alter the nature of the group interaction and the extent of group outcomes. The results suggest that group polarization is a function of more than just task characteris ties and group compesition with medium of communication playing a central rale. The findings clearly indicate that medium of com- munication has a strong eflect on group polar 188 MIS Quartoryune 1998 ization, The first question posed was: Do groups polarize in both FTF and GSS sot tings? The findings in this context provide sup- port for the established literature and confirm that group polarization accurs in lace-to-face settings. In newer forms of graup communica- tion using technology support like GSS, how- ever, the study finds that group polarization ‘occurs to a significantly lesser degree than in face-to-face situaliens, These results are in line with the predictions that the GSS setting will result in equality of participation and a more balanced argument set causing group consensus to accur toward the group average Polarization & Persuasion Table 7. Planned Contrasts for Group Polarization Medium*Task Interaction Contrast Hypothesis: Hypothesis __| Df_| Contrast $3 | Mean Square | F Value | Pr>F | Supported? Persuasive Arguments. _ __4| Hala: GSS.Int < FTFInt 553.07 553.07 315 | 008 ves | Hae: GSS-Jud FTF-Jud | 1 303.54 303.54 103 _[ 0.32 No Polarization _ _ Hab: GSS-int=FTF-int_ | 1 | 052 | 0518 07 | 079 yes Had: GSS-Jud FTF No Polarization Hib: GSS < FTF Yes, Task Pers. Arg Ha: intellective > Judgmental | Yes Polarization Hb: intellective > Judgmental_ | No _| Group Comp Pers. Arg H3: AD > FG No Polarization HS: AD > FC No Medium’ Task Pers. Arg, Ha: GSSrlntellective < Yes FTF + intellective Hac: GSSiJudgmental > No . FTF +Judgmental Polarization Hab: GSSrintellective = Yes FTF intellective Had: GSS+Judgmental < Yes FTF.Judgmental Medium’Group Comp | Pers. Ard H5a:GSS+A-D FTF+AD Polarization Ha GSS+A-D FTF+A-D 182 MS Quarteryrlune 1896 Polarization & Persuasion Table 3. Results of ANCOVA Dependent Variable: Polarization | Sum of Mean ] Source DF | Squares | Square | FValue | ProF Medium 1 68.95 68.95 545 0.023 Group Composition | t 2.04 2.04 0.16 0.689 Task 1 96.18 96.18 761__| 0.008 lAK(cowariate) i 272 272 022 | 0.645 Med* Group Comp t EE 19.56 155 | 0219 Med'Task 1 85.65 35.65, 677 0.012 YES Task*Group Comp 1 2.38 2.38 0.19 0.666 NO Task*Group’Med 1 0.28 0.28 0.02 0.882 NO Table 4. Results of ANCOVA Dependent Variable: Persuasive Arguments ‘Sum of Mean Source DF | Squares | Square | FVaue | Pr>F | Significant? Medium 1 1.70 1.70 O.01 0.934 NO Group Composition | __1 79.39 79.39 032, 0574 NO Task 1 | 2or719 | 2atrae | tta0 | 007 YES lAlfcovariate) 1 1.93 1.33 0.01 0,942 NO. Med'Group Comp 1 188,74 198,74 0.76 | 0.387 NO Med'Task i 969.22 969.22 3.82 0.053 YES Task"Group Comp, 1 28.49 28.49 0.12 0.736 NO Task*Group"Med 1 | 1548 | 15.48 0.06) 0603 | NO Table 5, Descriptive Statistics Persuasive Treatment N Polarization Arguments Mean Std Err Mean Std Err Medium 685 32, o7t 0.63 31.38 3.15 FTF 33 —1.44 0.65 32.41 aay Task Sudgrenial 32 ~139 063 25.36 3.36 intellective 33, ore 065 saat | 2.94 Group Composition Flexible-cohering 32 0.93. 0.69 size | 355 ‘Assertive-directing 33 i 3224 274 MIS Quartestyésune 1998 181 Polarization & Persuastorr Results Using a factorial analysis of covariance (ANCOVA) with planned comparisons, the hypotheses developed in the previous sections were tesied. Planned contrasts were used to test the a priori comparisons thal were spaci- ‘ied in hypotheses 4 (a, b, c, and d) and 5 (a and b). ANCOVA was conducted for the dependent variables: group polarization and persuasive argumonis. Tables 1 and 2 present the significance of the overall model fit, The results indicate a good model fit for both group polarization (F = 2.89, p< 0.0082) and persua- sive arguments (F = 2.03, p< 0.08). In addition, ‘examination of the results of ANCOVA (Tables 3.and 4) indicate that IAL had no significant effect on either the process or outcomes of ‘group decision making in this study. Main effects The results of the analysis of covariance are presented in Tables 3 (group polarization} and 4 (persuasive arguments}. The analysis indi- cates a significant main effect for medium {Table 3} in relation to polarization (F = 5.45, pe 0.023) but not for persuasive arguments (Table 4}, The results do not suppor hypothe- sis 1a, 1e,, GSS groups do nat exchange more persuasive arguments when compared with FTF groups. The results, however, do support hypothesis 1b that GSS groups observe signi icantly lower levels of polarization than FTF groups (Table 5: means = 0.71, -1.44). The ANCOVA analysis shows a signiticant main effect for task characteristics both from the perspective of group polarization and per- suasive arguments (Tables 3 and 4), On closer analysis (mean comparisons, Table §), the sludy finds thal, as predicted in hypothesis 2a, groups exchanged a significantly higher num- ber of persuasive arguments (F = 11.80, p < 0,001) when attempting intellective tasks (39.41) than when attempting judgmental tasks (25.35). In terms of polarization, the results Indicate (hat groups undertaking intellective tasks experience a significantly different level of polarization than those attempting judgmen- tal tasks (F = 7.61, p < 0.008). Table 5, howev- ar, suggests the relationship ig not in the direc- tion predicted by hypothesis 2b. Contrary to the relationship hypothesized, the results show that groups performing judgmental tasks expe- rience a higher level of polarization (1.30) than groups engaged in intellective tasks (0.72) Table 1. Overall Model Fit Dependent Variable: Polarization Sum of Source DF | Squares | Square | FVae | PraF | Significant? Model a 201.89 2.89 | _0,0082 YES Error 56 708,02, Corrected Total 84, 999.90 | l Table 2. Overall Model Fit Dependent Variable: Persuasive Arguments Sum of Source OF | Squares | Square | FValue | Pr>F | Significant? [oder 8 | a01s69 | so195 | 2.09 0.08 YES Error 49 |_tarnn7e | 247.18 ConectedToial | $7 | 16127.38 180 MIS Quarterly(June 1998 vvides support for the homogeneity of variance actos groups. Scatter plots of the dependent variables and the covariate within each group indicate a linear relationship. In addition, when the effects of all other terms in the model are controlled for, the within-group slopes terms in the model do not achieve significance at the 0.05 level,"® allowing the conclusion that the ‘within-group slopes are equal (homogeneity of regression). Having met the assumptions of independence, nermality, homogeneity of vari- ance. linearity, and homogeneity of slopes (Neter et al. 1990}, the study proceeded with the ANCOVA technique. Data analysis—protocol analysis The purpose of the prateccl analysis was to document similarities and differences in the decision process using the theory of persua- sive arguments. The components of the deci- sion process were captured by protocol ana- |yzing the transcripts for both the FTF and the GSS sessions, Protocol analysis is a research technique that allows examination of a group's predecisional behavior, It is asserted that pro- tocol analysis is @ promising method for open- ing the “black box" af decision processes in group-oriented research in information sys+ tems (Todd and Benbasat 1987). Although rel- atively undenutiized, this method supports the traditional experimental studies by providing valuable insight into the process that causes the eventual outcome being studied (Jarvenpaa et al. 1988) In conducting the protocol analysis, this study Caretully adhered to the guidelines spectlied by Ericsson and Simon (1993), Specifically, the protocols for the study ware collected concur rently and non-obtrusively, coding schemes were identified a priori, two coders were used (as recommended, neither of the coders was the principal investigator), and finally inter-rater reliability was calculated using Cohen's Kappa castlicient (Van Someren et al. 194). "SA-bedium (Fovabue = 001, p= 9282) lad Task (FVaive = 0.19. p= 6605) APGRP (FMaive © 122, p= 2796) Polarization & Porsuasion The concurrent collection of protocols was accomplished by capturing the communication among the group members while they were engaged in the problem solving task. For the FTF sessions, this was done by video recort ing the groups. For the GSS sessions, the complete session transcripts were printed, This method of concurrent data collection is considered less obtrusive and provides a more exact depiction of the decision process than the allemative of capturing the group process: retrospectively, i.2., by interviewing the particl- pants at the end ol the session. A coding scheme was developed 4 pricri to document level of persuasiveness using the theory of persuasive arguments as the basis. The process for coding the transcripts included defining “arguments” as the unit of analysis, classifying the arguments into different types of statements, and labeling the arguments by attributes of persuasiveness. In the GSS ses- sions, each separate remark was defined as an argument and used ag a unit of analysis, In the face-to-lace settings. an argument was defined as each uninterrupted statement’ remark by a participant (Ericsson and Simon 1999), ‘The rules for the classification of arguments based on persuasive arguments theory and selective excerpts from transcripts are present- ed in Appendix D. For both the GSS and the FTE situations, the entire session was content coded using two coders. The coders were ade- ‘quately trained prior te their attempting this activity, Both coders were given identical instructions and used definitions in Appendix D to code the arguments in the sessions for their persuasive content. Inter-rater reliability was calculated to determine if tha rules for coding were applied consistently by the coders. A conservative measure of inter-rater reliability Kappa-Goetticient"’ score was calculated to be 78%, whieh is well within the acceptable range (Van Someren et al. 1994) ‘Kappa = \proparvon af corresponding codes between Ihe ‘oro coders - tote expacted propastion for crrespanding ‘codes)!(t total expected proportion tor corresponding est MWS Quartenlptlune 1998 179 Polarization & Persuasion valid: namely, the perception by individuals that an argument is truthful and plausible, the extant to which this argument fits the current discussion thread, whether the argument fot lows {rom accepted facts or previously expressed views considered by the group, and if it contributes to the final decision at hand. Novelty of an argument similarly was depicted as a function of the argument being creative and nonredundant. In classifying an argument as novel, it had te present a new form of orga- nizing the information, or it had to suggest an idea that was not previously considered by the group The level of persuasiveness in a group setting was determined by the level of validity and novelty of arguments presented in a group set- ing {roler to Appendix O for details related to coding of validity and novelty). The variable persuasive arguments was deemed as inter nally reliable yielding a Cronbach a value of .893 when aking into consideration the sub: components of both validity and novelty. Initial Agreement Index An initial agreement index (IAl) was devised to ablain the level of initial agreement within the group. For groups with identifiable leaders, the IAI indicates the level of agreement in the {group when compared with the leader's intial stance.® For flexible-cohering groups. compari- ‘son is made with the average decision of all members cl the group. The IAl was used as a covariate when analyzing the dala using ANCOVA. III can range from 0 to 1. A score of 0 for Al indicates complete initial consensus, with the leader, whereas a score of 7 indicates ‘no initial consensus with the leader. ‘1a «20 Range-otthe dacision scale, Wihere X = ¥ divided by numoer of group members (ther than te Teac) YY = 2 iIngividual Dectsion «Leader Oeelsionjl—ehen there is slenoer OR ‘Y = Eanctvidual Deezson - Group's Avarage Detision}i— Woe thore is no leader ange of the cecssion scale is 6 (ram 7 7} 178 MIS Quartenyllune 1995 For groups with no leaders, individual dec sions are compared with the group average, ‘The IAI indicates the level of agreement in the group. A low IAl score indicates greater agree ment within the group than higher scores. Extreme scores (0 of 1) are not possible for ‘groups with no leaders (assertive members), since individual decisions are compared to the ‘group's initial average decision Data analysis—ANCOVA The experiment was designed as a 2x2x? {task x medium x group) factorial. ANCOVA was used to model relationships between the dependant and the independent variables. Linear contrasts were used to test hypotheses. ‘The dependent variables were polarization and persuasiveness. The independent variables ‘were communication medium (FTF or GSS): task characteristics (intellective or judgmental); and graup composition (assertive-directing or Hexible-cohering}. The covariate for the ANCOVA analysis was IAI, which indicated the level of initial agreement in the group. Preliminary analyses determined that the assumptions associated with ANCOVA have been mel. ANCOVA assumes that the covari= ate and the treatments are independent Measuring the covariate before administration of the treatments together with random assign- ment of subjects to groups leads to the asser- lion that the ageumption of independence has been met (Shavelson 1988). The Wstatistic?is very close to ? and thus the null hypothesis that the input data values ware normally dis- tributed could not be rejected. In cases where sample sizes are equal, the assumption of homogeneity of variance causes litle problem (Kirk 1962), In this case, the cell sizes, although not exactly equal. were within 1 of each other. Given the assurance of normally, the siudy proceeded to use Hartley's simple test statistic (Shavelson 1988). which yielded values of 3.09 for polarization and 2.98 for PAT that are not significant. This F,,,, test pro- OW-statstc:Palanzat arguments = 099 gar iT = 0,087, MIS Quasterlyidune 1998 175 Polarization & Persuasion Medium-Group Interaction In explaining group polarization, the rhetoric- of-risk theory holds that shifts occur when peo- ple are permitted to argue for their position and exchange information using the combination of both verbal and non-verbal communication alforded by face-to-face discussion (Kelley and, Thibaut 1969). Researchers have reported higher levels of polarizatian when subjects in an experiment watched and listened to a group discussion; however, anly a partial shift was observed when they meraly listened to the dis- cussion (Lamm 1867). Polarization researchers (Pruitt 19714) explained these data by asserting that in the listening only con- dition, the possibilty for non-verbal communi- cation was eliminated, This reduced the mar ner in which assertive group members could demonstrate their confidence, which dimin- ished the degree of shift in the decision. Rhetoric-ol-risk theory has significance in cases where technology is introduced as a means of communication. One of the charac- teristics of GSS is that it provides anonymity during group discussion and eliminates any non-verbal communication that Is possible in face-to-face meeting settings {Connolly at al. 41990). It is our intention to examine whether such GSS characteristics wil have effects that are in line with thetoric-of-isk theory in group polarization. GSS researchers have also sug gested that the potential effect of assertive members will only be felt when their presence is salient in a particular group interaction. In addition, in other studies where the interast is in frst advocacy effects, it was found that first advocates are persuasive in FTF discussions and lass likely to influence others in computer- mediated settings (Weisband 1992). In an anonymous GSS meeting, it is predicted that assertive members will have a diminished effect on the meeting process and outcomes. Flexible-cohering groups that do not have assertive members traditionally converge toward the majority view in FTF settings. The low degree of polarization in flexible-cohering groups is altributed to the desire of such indi- viduals to be viewed as flexible team players. Placed in an anonymous GSS situation, how- 174 MIS Quarterlyilune 1898 ever, these individuats may experience a lower need to conform and a lavr desire to be viewed as team players. Without the pressure of con- forming and fear of being viewed as non- cooperative group members. flexible-cohering groups may focus more on the task at hand rather than their desire to be viewed favorably. AS a result, it is expected that individual group members will ee! liberated to contribute to the group discussion more novel and creative arguments leading to an inctease in the level ‘of persuasive arguments generated. This can lead to greater resistance to reaching a com- ‘mon group decision by adhering or insisting on ‘heir individual premaating opinions. Thus, it is predicted that, for flexible-cohering groups in GSS settings, the increase in informational influences and decrease in normative influ- ences will cause an increase in the level of persuasiveness and the degree of polarization. In sum, using rhetoric-of-risk and the results from GSS studies as a guideline, itis predicted thal assertive-chrecting groups in GSS settings ‘will experience lower levels of persuasiveness and polarization than assertive-directing groups in FTF settings. For flexible-cohering ‘groups that tend to converge toward the group average in an FTF setting, itis anticipated that the GSS setting will allow them to experience @ higher level of persuasiveness as well as ‘express more extreme positions and thus. greater polarization HSa: Assertive-directing GSS groups will ‘experience lower levels of persu: siveness and polarization than A-D FTF groups. Hsb: Flexible-cohering GSS groups will experiance higher levels of persua- siveness and polarization than F-C FIF groups. Research Design and Methods This study uses multiple data collection and analyses techniques to investigate the out- comes and process of group decision mak- H3: Agsertive-directing groups will demonstrate a higher degree of polarization and greater persuasive- ness when compared with flexible- cohering groups. Medium-Task Interaction When groups undertake tasks in a GSS set- ting, the features of the medium may interact with the characteristies of the task to affect group processes and outcomes and result in seemingly contradictory outcomes. In GSS settings, individuals, who may otherwise have reservations about voicing their opinion regardless of its accuracy, feel empowered to do so given the anonymity provided, in the case of intellective tasks, by reducing the nor. mative pressures present in a FTF setting, GSS provides the opportunity for group mem. bers) with the correct solution to present their position to the group. The correct answer therefore surlaces without interference from position, status, or individual domineering behavior. This presentation of the solution to the group for such tasks would forge consen- sus and eliminate the need for an extended discussion. The impact of anonymity in the ‘context of intellective tasks can be dramatic a5. it would permit individuals with the “correct answer’ to reveal the same to the group with. ‘out any evaluation apprehension. Given the characteristics of intellective tasks, itis pradict- ed that ance such a "correct answer” is revealed, the group will rally around such a solution and come to consensus. For intellec- tive tasks, no differences in polarization are anticipated since, in both FTF and GSS situa- tions, individuals are expected to converge toward the correct solution. However, it is anticipated that, for the same level of polariza- tien, groups in GSS settings Undertaking intel- lective tasks will observe a less engaged and less prolonged discussion about various eom- peting solutions, This will lead to less argu- mentation. Therefore, the prediciton is: Haa: GSS groups attempting intellective tasks will exchange lower levels of persuasive arguments than FTF groups attempting intellective tasks. Potarization & Persuasion Hab: GSS groups attempting intellective tasks will experience similar levels of polarization as FTF groups attempt- ing intellective tasks. Judgmental tasks involve situations where there is no demonstrably correct answer, necessitating a group pracess whereby mem- bers present their individyal altitudes and opin- jons. share information, and recencile differ ences to reach consensus (Hollingshead et al 1999; Laughlin and Earley 1982). Previous ‘work in the area suggeste that, in judgmental tasks, shifts occur due to more of a normative rather than an informational process {Kaplan and Miller 1987: Laughlin and Earley 1982), These studies Nave reveaiad low levels of per: suasion for judgmental tasks due to the salience of social influences. This paper pre- dicts that the features of GSS will allow not only allow a reduction in normative influences (Tan et al. 1993) but also a greater focus on the arguments exchanged, Judgmental tasks will thus benefit from an increased focus on the content of the arguments rather than the initiator of these arguments. The expectation is that an increase in the persuasive arguments aol for judgmental tasks in the GSS setting will be observed when compared with the FTF setting. Due to the democratization that GSS facilitates, however, the expectation is that these arguments will be representative of every member of the group and thus result in a convergence toward the majority view. In terms of polarization, lawer levels in GSS than in FTE situations is expected. As such, it is hypothesized that: Hc: GSS groups attempting judgmental tasks will exchange higher levels of persuasive arguments than FTF ‘sroups attempting judgmental tasks, Had: GSS groups attempting judgmental tasks will experience lower levels of polarization than FTF groups attempting the same tasks. MIS Quartery.sune 1998 173 Polarization & Persuasion suasive arguments) predominates for intellec- tive tasks and normative influence predomi- nates when the Issue is judgrnental and there is no correct answer. The expectation, there- tore, is that there will be a larger pool of per suasive arguments, both novel and valid, in intellective tasks where there is clearly a cor- ect answer and where facts and relevant information are important in establishing the correctness of one’s position. For judgmental tasks, however, where there is no right answer, “facts and data are less important and decisions are supported by appeal to social norms and consensus of preference” (Kaplan and Miller 1987), the expectation is that there will be significantly fewer persuasive, novel, and valid arguments, H2a: Groups undertaking intellective tasks will generate a larger pool of persuasive arguments when com- pared with graups undertaking judg- mental tasks. One study reports that a greater number of novel and valid arguments within the argument poo! lead to greater polarization (Ebbesen and Bowers 1974). Some researchers have found that mere exposure to others’ positions (i.e. normative influences} could also lead ta. grour: polarization. While both informational and nor mative influences lead to polarization, there is evidence suggesting that informational intlu ence (persuasive arguments} alone can pro: vide sufficient explanation for the degree of polarization. In addition, while supporting the notion that informational and normative intlu ences are interrelated, it has been suggested that the effects of persuasive argumentation ‘on group polarization is strenger than that of social camparison alone (Isenberg 1986). In some eases, normative influence has been shown to have @ mediating effect on group polarization by causing an increase in generat- ed persuasive arguments. Based on these Findings, the belie! put forward ia this paper Is that in judgmental situations, which are more susceptible to normative influences, there will be lower levels of polarization than in situa- tions that are mostly dominated by infarmation- al influence and persuasive arguments. 172 MIS Quarteriyitune 1896 H2b: Groups undertaking intellective tasks will exparience a higher level of polarization when compared with groups attempting judgmental tasks. Group Composition The role of group composition has been emphasized in the GSS and group polarization literature. In explaining group polarization, leadership-confidence theory (Pruitt 1971a} suggests that individuals demonstrating greater assertiveness in a group setting are more persuasive in group discussions and thereby cause other members of the group to shift toward their initial stance. In the current study, the interest ig in how the presence or absence of assertive members in a group set- ting affects the process and autcomes of group decision making, When a group has an assertive-directing member wha Is competing for authority, responsibility, and leadership, and exercising persuasion in an attompt to exart influence and steer group members in the direction of his or her choice, the tendency of the group is to shift toward this individual's stance on the issues. Additionally, such individuals have been identified (Isenborg 1986: Pruitt 1871a} as having more extreme pesitions.* The pros- ence of such individuals in a graup can have a profound effect an the degree of group: palar- ization. Alternatively, for groups that are flexi= ble-cohering and without assertive-directing individuals, the focus is on being open-minded and willing 1 adapt to change. Such groups are typically composed of individuals who like to be known as flexible team players (Porter and Maloney 1989). Given that such teams lack a dominant member, the tendency of such groups is to canverge toward the group average, resulting in both a lawer level af shift and less persuasive argumentation since no one individual is marketing a particular point of view. Based on the above, the following Is predicted: “in ths e1uoy, asconive-directing individuals also tanded to Demora oxrame: A.0-= 6; FsC #38 lose a train of thought, typically increases the total number of arguments generated, GSS. researchers have similarly suggested that indi- viduals tend ta not only create @ greater num- ber of total arguments but algo ba more cre- alive and novel in their problem solving {Nunamaker et al. 1991}, In addition, the anonymity of GSS causes reduction in confor mane pressure and a diminished fear of neg- ative evaluation, which would cause group. members to withhold their ideas or comments. ‘The expectation, therefore, is thal for groups communicating via GSS there will be an increase in the total number of arguments as ‘well as the level of novel solutions or alterna- tives generated. ‘The feature of anonymity may, however, result in a reduction of the validity of the argument pool. The reduction in validity may result from individuals not having te be accountable for the arguments they contribute. & reduction may also eecur if individuals perceive their input as unneeded and rely on others to accomplish the group's goals (free-riding). Ironically, the same process of reduction in evaluation apprehension that results in a high- er level of creativity and novelty and in a greater number of argurhents may also result In a reduced level of validity. The net result is that the role of GSS on per suasiveness of arguments and polarization is, ‘complex. It is predicted that a net increase in Persuasiveness will be observed. While a decrease in the validity of arguments is expected, the increase in navelty will offset this reduction and result in a net increase in per- suasiveness. Novelty in group polarization studies has been shown to play a central role in causing more extreme graup decisions. lsenberg 1986; Vinokur and Bumstein 1978) In terms of choice shift, however, the expecta- tion is thet for GSS settings the equality of par- ticipation aftorded wil lead to the consideration of a more balanced set of persuasive argu- ments that represent the positions of all group members rather than one specific extreme position. The result will be consensus toward a ‘majority view rather than an extreme view and thus less polarization. Polarization & Persuasion In sum, the functionality of the GSS medium may increase persuasive arguments and infor mational influence and yet praent a mare bal: anced set of arguments causing the prediction that GSS supported groups will demonstrate higher levels of persuasive argumentation and yet lower levels of polarization than groups meeting in a face-to-face setting. Hi: GSS groups will exhibit higher levels of persuasive argumentation when compared with FTF groups. Hib: GSS groups will demonstrate a lower degree of polarization when com- pared with FTF groups. Task Characteristics The importance of task characteristics on group processes and decision outcomes is weil documented. Studies have shown that tasks can account for 50% of the variance in group performance (Poole et al. 1985). The ‘elationship between task and group polariza tion was first alluded to in the proposal of a group-task continuum (Laughlin 1980). At one end are judgmental tasks, which entail achiev- ing group consensus on tasks that invelve ‘judgmental behavioral, ethical or esthetic judgments for which there are no demonstr bly correct answers” (Kaplan and Miller 1987, , 907). At the other end are inlelfective tasks, which involve a demonstrably correct solution. For intellective issues, the group's task is to uncover the correct answer, For judgmental issues, the group's task is to decide on the moral, ethieal, or preterred position (MeGralh 1984), While itis important to make such a dis- tinction based an task characteristics, it should be noted that these are not exclusive classifi cations, i.e., judgmental tasks have intellective components and vice versa (Kaplan and Miler 1987; MeGrath 1984) Group polarization researchers (Isenberg 1986; Kaplan and Miller 1987; Laughlin and Earley 1992) report that the lype of task under- taken affects the mode of influence that the members adopt. Infarmational influence {per- MIS Quarteryilune 1988171 Polarization & Persuasion standing of the polarization phenomenon only when the mediating effect of communication medium is taken inte consideration. This model, while remaining faithful to explanations of group polarization in face-to-face settings, also introduces communicatian medium into the equation by hypothesizing that the effect of task characteristics and group compasition on group process and outcome will be moderated by the communication medium employed to Support the graup decision making pracess. Communication Medium While the sociopsychological literature has extensively studied group polarization in the context of face-to-face meetings {Isenberg 1986; Myers 1982), the interest related to the impact of computer mediation on group polar- ization is relatively new. The two communica: tion media that are considered here are FTF meetings and GSS-supported meetings. GSS research has reported differences in group decision outcomes when comparing GSS meetings with FTF meetings (Jessup et al 1990). & premise of such GSS research is that the difference in the outcomes when compar- ing GSS with FTF meetings results from the significant differences in the process of dev sion making provided by the GSS environment when compared with the FTF environment Two features of GSS that can cause differ ences in group process and outcomes are anonymity and parallel communication. The introduction of anonymity in a group process can have contradictory elfecis. On the ‘one hand, anonymity decreases conformance pressure (Nunamaker et al, 1981; Valacich ot al, 1993) for the participants in a group setting and allaws participation based on content of communication rather than the source that generated the communication, On the other hand, anonymity has the potential to cause trea riding whereby some members of a group. rely on athers to accomplish the group goals {Diehl and Stroebe 1987; Nunamaker et al. 1991). Anonymity and its impacts on polariza- tion have been studied in the context of com- puter-mediated communication (Kiesler et al. 170° MIS Quarterlytiune 1998 1984; Siegel et al, 1986). An immediate impact ‘of anonymity In groups is that it de-empha- sizes the status of individuals that are held in high esteem by the group members. One study suggests that this would increase peo- ple's consideration of alternate (majority) views, causing consensus to occur in the direction of the majority view of the group rather than the most dominant individual (Kiesler et al, 1984). This would result in less extreme views and thus less polarization. Persuasive arguments theory (Isenberg 1986, Pp. 1145) contends that a group's final choice ‘on an issue is a function of the poo! of persua sive pro and con arguments exchanged Persuasiveness is in the forth of valid and novel arguments. In studying computer-medi- ated communication, one study found that the complexity, number, length, and novelty of arguments Was lass in a computer-mediated setting than in FTF discussions (McGuire et al 1987). Furthermore, in the context of group polarization, it was found that arguments were less persuasive in computer-mediated groups than in FTF groups (Kiesler et al. 1984). A more recent study of the social processes that lead to first advacate effects in group decision making explains that ‘The findings .. replicated those of previous researchers (Siegel etal. 1986; McGuire etal 1987). Electronic groups, when given as mulch te a5 they needed to work on a prot lem, exchanged the same number of com- ments but the information exchanged was qualitatively ifterent (Weisband 1992 ». 258}. Building en this foundation of work, the interest here is studying whether the functionality afforded by GSS alfects the process and out comes of group decision making when study: ing choice shifts. In addition to anonymity, GSS also provides parallel processing capabil- ities (Nunamaker et al, 1991), which can also aifect (both positively and negatively} the group discussion and the informational Influ- ence processes at work. The feature of parallel communication should prove to be a liberating experience for mast group members. The abill= ty to contribute without having to wait or take turns to present a point of view, and potentially over the past three decades, social psycholo- gists have also suggested a similar relation- ship. These researchers have suggested that task characteristics (Kaplan and Miller 1987; Laughlin and Earley 1982) and group compasi- tion (Kovacs 1982: Vidmar 1974) can alfect group polarization, The research model for the present study (Figure 1) blends explanatory constructs trom the group polarization literature with the con- temporary context afforded by the GSS tech- nology to explain group polarization. This model holds that task characteristics and group composition have a direct bearing on group decision processes and outcomes. as stated by both GSS and group polarization researchers. Furthermora, the model po: that GSS will have a direct effect on group polarization and, more importantly, that the relationship of task characteristics ar group composition with the decision process and oul- come ‘will be a function of the communication mediurn used, In this study, the focus is on the following questions: _——__, Taskers} Hf sImtellective *Judgmental Task Comeunication Mediun Pace to Face “GSS Group composition ‘*Flexible-Cohering Medium. Interaction -GSS>FIF (Persuasive Arguments) “GSS Judgmemal (Persuasive Arguments) sIntellective > Judgmental (Polarization) Polarization sAssertive-Dicecting “AD>I suasive Arguments) wa{ DeFC (Persuasive Arguments) *AD>FC (Polarization) Figure 1. The Mode! and Hypotheses MIS Quartenylune 1998 169 Polarization & Persuasion group decisions result from normative influs ences that occur due to an individual's desire to conform to the expectations of the other members. A tenet of this theory Is that group polarization occurs when individuals review other group members’ positions and then revise their decision to align themselves ‘with the preferred position of the group, Group polarization results when a significant number of group members fellow this process, causing a shill in the group decision in a common direction valued by the group, The theory of persuasive arguments (Burnstein 1982; Vinokur and Burstein 1978), the more prominent of the two theories, focus: es on information eollection (Heath and Gonzalez 1995). This theory suggests that shifts in group decisions resull from sharing relevant and factual information about the sits ation at hand, Explanations provided by this theory indicate that the group's decision is a function of the persuasivencss of the argu ments. As part of the group process, individu: als contribute and collect arguments for the various positions that are espoused. The notion of persuasiveness is pivotal to this theo- ry. Two major factors contribute to the persua- siveness of an argument: its validity and its novelty (Isenberg 1986; Burnstein 1962) Validity of an argument is described by its level of acceptability and relevance to the group and the discussion the group is engaged in Novelty, on the other hand, is described by the extent to which an argument is unlikely to have been previously considered by the group {Vinokur and Bumstein 1978) Researchers have argued about the relative merits of both theories in best explaining group polarization. There have been a number of studies conducted to evaluate the relative merit of social comparison and persuasive arguments for explaining polarization. Where the phenomenon of interest is choice shift, one study indicates that the theory of persuasive arguments provides a consistently superior explanation (Zuber et al. 1992). More recent studies however, have once again started the discussion about relative merits of social eom- parison and persuasive arguments theories (Heath and Gonzalez 1995; Paritt 1994), Botn 168 MiS Quarieriyrlune 1988 these efforts, while supporting the impor tance of persuasive arguments theory, suggest that social comparison plays a mediating role in the group polarization phenomenon. One study found that group members typically pre- sent arguments supporting decisions they pre- fet and refrain from presenting information that contradicts their preterted position (Pravitt 1984). The other study found that group mem- bers use persuasive arguments from other group members that support their position as a Means of rationale construction (Heath and Gonzalez 1995). These findings suggest that in presenting persuasive arguments, indlividu- als are also relaying their personal positions ‘on the decision and thus exerting social com- parison forces. In order to capture the process of decision making, this paper focuses on studying the variations in the persuasive level of the argu ment pool in the different contexts considered Unlike previous experimental studies that imposed controls on either the position expo- sure or the discussion, this paper studies group polarization in the presence of both informa: tional and normative influences. By capturing and analyzing the level of persuasiveness that precedes the group outcome of polarization? the study seeks to provide additional insight into tha process of group polarization, Research model ‘A.variety of factors have been hypothesized to result in beneficial outcomes associated with the use of technology to support group meet ings. Despite some mixed findings and incan- clusive results (George et al. 1990; Sambamurthy and Poole 1992), the growing volume of research has provided support for the important role of task characteristics, group compositian, and technological factors (Benbasat and Lim 1999; Dennis et al. 199% Pinsonneault and Kraemer 1889; Reder and Conktin 1987) on the process and oulcomes of group decision making. In a different context ®in this study, me terms group polarization and choice shift are used interchangeably acteristics of the group (Burnstein 1969; Dennis et al. 1988; Lim and Benbasat 1993). Examining this phenomenen from an outcome perspective, the paper reports on the effects of communication medium (GS$ or tace-to-lace), task characteristics (intellective or judgmental), and group composition (assertive-cirecting or flexible-cohering) and their interactions on the lovel and direction of greup polarization, To gain insight into the nuances of the actual group process, the group discussion that ensues is examined and further analyzed for persuasiveness of content in each of the dimensions of interest. This study employs multiple data collection and analyses techniques to study the research questions posed. Using a lab experiment with 33 groups each composed of either five or six? individuals, the degree and direction of group polarization was tracked for different task lev- els and different interpersonal interaction styles in a face-to-ace (FTF) and a GSS set- ting, In adition, transcriptions of both the face- to-lace and GSS meetings are content coded for the persuasiveness of arguments exchanged in different group settings. Theoretical Background and Research Model Group polarization is the tendency of individu- als in a group setting to engage in more extreme decisions than their original private individual decisions (Myers and Lamm 1876) The group polarization phenomenon was cavered by Stoner (1961) when he observed that group decisions are riskier than private decisions of individuals comprising the group. He labeled his observation as “risky shift Subsequent researchers, however, observed that changes in group decisions occur in the direction of risk as well as in the cautious direction and, as such, this phenomenon has been relabeled as group polarization (Myers and Lamm 1976). Over the past 30 years, vari- ‘Groups consietes of five ncvguals: however, twas af me 33. roups had te members, Polarization & Persuasion cus researchers have studied group polariza- tion by focusing on group discussion and its effect on inducing group polarization both at the level of individual opinions—aititude shifts (Hinsz and Davis 1984)—and for group deck. sions—chaice shifts (Zuber et al. 1992). In many organizational settings, groups need ta reach a common decision despite aifferences in their individual opinions. An example of such a choice shift is evidenced when strategic plans are cratted for organizations using expertise from various functional areas: production and sales may have contradictory goals and yot must agree on a common plan that will meet both their requirements (Ackoff 1987). On the other hand, interest in attitude shift occurs when studying situations where individuals con- sult with others and yet make their own final decision, such as when buying a car, a house, or other items where individuals collect infor- mation but choose to use ar ignore the informa- ficn collected. While the development of GSS is quite independent of the debate between choice and attitude shift, it is reasonable fo suggest that a purpose of this technology is to pravide communication and decision support to allow groups to share a vision, build consen- us, and make joint decisions (Nunamaker et al, 1991). Itcan further be argued thal GSS can be used to forge joint group decisions regard- tess of whether individual personal preferences agree with this joint decision, Given tha interest in common group decisions in organizations, the more intense group dynamics associated with reaching a common group decision (Heath and Gonzalez 1995}, and the capability of GSS to provide groups with the means for reaching such joint decisions, this paper specifically focuses on choice shift as a measure of group polarization. Several theories have been proposed as explanations for group polarization (see Lamm and Myers 1978 and Pruitt 197 1a for extensive reviews). A meta-analytic study of this pha- nomenon indicated that there are two viable explanations—the social comparison theory and the theory of persuasive arguments—that need further examination (Isenberg 1986). The theory of social comparison (Baron and Roper 1976; Brown 1965) suggeste that changes in MIS Guarteryiune 1998 167 Poiarization & Persuasion departments or by relatively isolated individ als within organizations (Ancona and Nadler 1989; Drucker 1988). While teams have tradi tionally been used in the lower echelons of the organization, only recently has teamwork per- meated the top layers of the organization. in many companies, team-based organization designs are replacing more traditional exec- itive structures (Ancona and Nadler 1989), This trend of increased teamwork and group decision making is Supported and inspired by a host of neve information technologies that pro- vide a networked environment to aid “interper: sonal computing’ and group work. With the increasing aeceplance of computer networks in the workplace, and their use for individual and group productivity, these technologies are fast becoming the preferred method for communication and collaboration among team members. Computer support for such coordination mech- anisms and their usage for group activities is alse reterred to as Groupware. Groupware is defined in general terms by Elis et al. (1991 p. 40) as “computer-based systems that sup- Port groups af people engaged in a common task (or goal) and that provide an interface to a shared environment.” Using this definition, groupware can take on various forms including elactronic mail, electronic bulletin boards, tele: conferencing, group wating, and group support systems (GSS). For the purposes of this study, attention focuses on GSS settings. GSS pro- vide an environment ta facilitate collaborative wark by providing the tools for coordinating, consensus reaching, and decision making nec: essary in day-to-day organizational lite. GSSs are concerned with enhancing the quali ty of group communication (DeSanctis and Gallupe 1987; Nunamaker ¢t al, 1989}, The impact of these systems on both the outcomes from graup discussion (Nunamaker et al. 1987, 1989) and its effect an the process of commu- nication {Poole et al. 1991; Zigurs ot al. 1988) is coming under close scrutiny. This paper studies the characteristics of group decision making from an outcome as well as a process- based perspective, From an outcome-based perspective, the focus is on a phenomenon 165 MIS Quartenyilune 1998 referred to as group polarization. Group polar- n is sald to occur when group choices are different from initial individual choices or when group's initial tendency is enhanced follow. ing group discussion (Lamm 1988}. From a process-based perspective, the focus is on the discussions that precede the group decision Of particular interest is the level and nature af persuasiveness of arguments exchanged in a group setting. Persuasive arguments theory (PAT) holds that a group's decision is a tune- tion of the persuasiveness of arguments that individuals are exposed to. before formulating their final decision (Lamm and Myers 1978; Prt 1971, 19716). ‘The implication of the phenomenon of group polarization is that group dynamics have the potential of changing the group's final decision Without necessarily changing any of the under- lying facts that lead to the decision. Group polarization can be viewed as having many potential benefits as well as detrimental effects. A number of historic fiascos have been attributed to group polarization including deci sions made by President Nixon's inner circle regarding the handling of the Watergate cover up (Green and Conolley 1974) and the ‘Challenger space shuttle disaster. On the posi- tive side, group polarization is beneficial in group counseling situations such as quitting smoking, diet programs, and alcoholics anony- mous, as welll as for charitable giving situa- tions where there is a need for group consen- sus in an extreme position (Muehieman et al 1976; Toch, 1965) The purpose of this study is to examine group polarization in a contemporary context. As criti cal decisions in organizations are made in task-focused teams rather than by isolated individuals, and as group technologies with their potential for altering group interaction {Sia et al. 1996) become more accepted in organi- zations, the phenomenon of group polarization takes on additional importance. This paper synthesizes and draws upon frameworks from both the group polarization and GSS literatura. Three variables appear to be critical determi- nants of group process and outcomes: tho communication medium used, the characteris- ties of the task being addressed, and the char- Polarization and Persuasive Argumentation: A Study of Decision Making in Group Settings’ By: Maha EFShinnawy Department of Information and Operations Management Mays College and Graduate Schoo! of Business Administration Texas A&M University College Station, TX 77843-4217 USA. Er-Shinnawy@tamu.edu Ajay S. Vinze Sehool of Accountancy and Information Management College of Business Arizona State University Box 873606 Tempe, AZ 85267-3606 Abstract This research focuses on group decision make ing from beth an outcome and a process- based perspective. This study draws from the well-established literature of group polariza- tion, as well as the growing body of GSS Itera- ture, 10 develop a model fo study group polar- "Robart 2muc-was he accepting senior edeot fo thi pape Polavization & Persuasion zation in a contemporary communication can. text, The proposed mode! focuses on commu. nication medium, task characteristics, group ‘composition, and their interaction as exptana: ffons for the outcome of group polarization and the process that precedes it. An experimental research method is used to test the refation- ships suggested by the model. In this study, group polarization is racordad by comparing decisions at the individual level, face-to-face group settings, and GSS mediated settings. The initial agreement index indicates the divev- sity of individual stances within the group. This index is used as a covariate to enhance under- standing of the extent of group polarization. The group process is decumented by protoco? anatyzing transcriptions of the FTF and GSS group sessions fer persuasive content. A -2x2x2 factorial design was used 10 analyze the resuits. The analysis indicates that, for both process and outcomes, the medium of com- munieation and task characteristics interact with one another to provide the dominant explanation. Surprisingly, group composition had no impact on either polarization or persua- sive arguments. The findings reported In this study are of importance to organizations that increasingly rely on groups as units of decision making. The results also provide insight to researehers of group deeision making and to future developers and users of group support systems. Keywords: Group polanzatien, choice shilts, group support systems ISAL Categories: AAO803, AI0105, GBO1, HAO301, HA1101 Introduction In today’s organization, teams are essential management tocls (Gersick 1988), Organizations by and large consist of a combi- nation of both permanent and temporary ‘groups, There is widespread consensus that in the coming years work will, in general, be per- formed in task-focused teams, not in traditional AMS Quasteryicune 1998 165

Vous aimerez peut-être aussi