Achieve Synergy in Group Decision Making PDF

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/271670184 How to achieve synergy in group decision making: Lessons to be learned from the hidden proﬁle paradigm Article in European Review of Social Psychology · March 2012 DOI: 10.1080/10463283.2012.744440 CITATIONS READS 67 10,699 2 authors: Stefan Schulz-Hardt Andreas Mojzisch Georg-August-Universität Göttingen Universität Hildesheim 96 PUBLICATIONS 5,799 CITATIONS 118 PUBLICATIONS 4,482 CITATIONS SEE PROFILE SEE PROFILE All content following this page was uploaded by Andreas Mojzisch on 13 February 2015. The user has requested enhancement of the downloaded file. EUROPEAN REVIEW OF SOCIAL PSYCHOLOGY 2012, 23, 305–343 How to achieve synergy in group decision making: Lessons to be learned from the hidden proﬁle paradigm Stefan Schulz-Hardt1 & Andreas Mojzisch2 Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 1 Georg-Elias-Müller-Institut für Psychologie, Georg-August-University Göttingen, Göttingen, Germany 2 Institute of Psychology, University of Hildesheim, Hildesheim, Germany Based on over 25 years of research on hidden proﬁles and information sharing in groups, and particularly our own work in this area, we outline a general model of how groups can achieve better decisions in a hidden proﬁle situation than their individual members would have been capable of (i.e., synergy). At its core the model deﬁnes intensity and bias as the two key parameters that have to be optimised with regard to both the discussion of information and the processing of information in order to ensure synergy in group decision making. We review the empirical literature on information sharing and group decision making in the hidden proﬁle paradigm (with a particular focus on our own studies) to illustrate how group decision quality can be enhanced by increasing intensity and decreasing bias in the discussion and processing of information. Finally we also outline why we think that the lessons learned from research using the hidden proﬁle paradigm can be generalised to group decision-making research in general, and how these lessons can stimulate studies in other ﬁelds of group decision-making and group performance research. Keywords: Group performance; Hidden proﬁle; Information pooling; Process gain; Synergy; Biased sampling. Important decisions with far-reaching consequences are often made by groups rather than individuals. For example, the managerial board of an organisation is responsible for strategic decisions aﬀecting the whole organisation, high-ranking personnel (e.g., in a university) are hired by selection committees, and police investigations in particularly important Correspondence should be addressed to Stefan Schulz-Hardt, Georg-August-Universität Göttingen, Georg-Elias-Müller-Institut für Psychologie, Wirtschafts- und Sozialpsychologie, Gosslerstrasse 14, D - 37073 Göttingen, Germany. E-mail: [email protected] Ó 2012 European Association of Social Psychology http://www.psypress.com/ersp http://dx.doi.org/10.1080/10463283.2012.744440 306 SCHULZ-HARDT AND MOJZISCH criminal cases are subjected to special commissions. One of the reasons for the popularity of groups when it comes to making decisions is the belief that using groups will pay oﬀ with regard to decision quality. Taking up the spirit of Aristotle’s famous words that ‘‘the whole is more than the sum of its parts’’, groups are often expected to achieve ‘‘synergy’’; that is, to yield a surplus in decision quality that individual decision makers would not be capable of (e.g., McGrath, 1984). However, the reality of group decision making often falls short of these expectations. The aim of this paper is to Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 shed some light on why this may be the case and, hence, how this deﬁcit might be overcome. NECESSARY PRECONDITIONS FOR SYNERGY IN GROUP DECISION MAKING The above-mentioned question of whether and under what conditions working in groups yields a group-speciﬁc surplus in performance is at the heart of social psychological group performance research (for an overview, see Schulz-Hardt & Brodbeck, 2012). The basic aim of this research is to determine whether interaction and social interdependence in a group facilitate or hinder performance, and to specify the particular cognitive, motivational, and coordination processes that are responsible for this facilitative or inhibiting eﬀect of group interaction. To this end, group performance is compared with a statistical baseline that indicates what performance would be expected in the absence of any functional or dysfunctional group processes. Often this baseline is determined by either introducing nominal groups (i.e., groups where no interaction takes place and all members generate their contributions to the group product individually) or by measuring the real group members’ individual contributions (e.g., their individual judgements or their individual propo- sals) prior to the real group interaction. If group performance exceeds this baseline, then a group-speciﬁc performance surplus has been identiﬁed (and vice versa for group performance below the baseline). Over time, such group-speciﬁc performance surplus has been discussed and investigated under diﬀerent labels such as, for example, ‘‘assembly bonus eﬀect’’ (Collins & Guetzkow, 1964) or ‘‘process gain’’ (Hackman & Morris, 1975). Recently Larson (2010) extended and integrated these earlier approaches into a general framework of synergy in groups. Whereas so-called weak synergy is said to occur if group performance exceeds the average performance in a nominal group, strong synergy requires the group to perform better than even the best individual member of a nominal group does.1 1 So far, ‘‘process gains’’ is the term most frequently used in the literature to denote a performance surplus that is due to group-speciﬁc processes. However, the use of this term SYNERGY IN GROUP DECISION MAKING 307 With regard to group decision making, this approach implies that groups realise synergy if (and only if) they make decisions that have a higher quality than either the average or the best of their group members’ individual decisions prior to group interaction—the former would constitute weak synergy, whereas the latter would be a case of strong synergy. This is an important implication, because it indicates that in many decision-making situations groups will not have the potential to realise such synergy. For example, let us assume that the group has to decide among several Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 alternatives, and each alternative is characterised by certain advantages and disadvantages. For the sake of simplicity, let us further assume that all advantages and disadvantages have equal strength. (We know that this is often not the case, but all of the following arguments could also be made with advantages and disadvantages of diﬀerent strength—only with the reasoning becoming more complex.) Obviously then, the best alternative is the one with the most advantages and the least disadvantages, the second best alternative scores second best on advantages and disadvantages, and so on. Typically, not all group members know about all advantages and disadvantages of the various alternatives individually. Whereas some pieces of information will be known to all members of the group (such pieces of information are called shared information), other pieces of information (so-called unshared information) will be unique knowledge of a single group member. There is, of course, also information that is known to some but not all members of a group—once again, for the sake of simplicity, we will leave this partially shared information aside. However, it would not change the nature of our arguments if this type of information were taken into account. involves some problems. Similar to Larson’s concept of synergy, the conceptual idea is to determine a plausible baseline of performance that occurs in the absence of (functional or dysfunctional) group processes and then measure whether group performance exceeds this baseline (which would indicate process gains) or falls below this baseline (showing process losses). The baseline used to determine process losses and process gains is the so-called group potential, a concept developed by Steiner (1972) in his inﬂuential book on productivity in groups. The problem is that Steiner did not foresee the possibility of process gains; his analysis only dealt with process losses. Consequently his idea of group potential was that it should denote the maximum productivity possible when taking task demands and group member resources into account. This implies that researchers who are (also) interested in the detection of process gains either have to work with an artiﬁcially enhanced baseline, which makes the detection of such process gains rather unlikely, or have to apply a more pragmatic group potential, which (at least partially) departs from the original deﬁnition and meaning of the concept. In turn, what might be called ‘‘process loss’’ in one analysis might be a ‘‘process gain’’ in another one. To avoid such problems we decided to apply the Larson (2010) synergy concept throughout this paper. 308 SCHULZ-HARDT AND MOJZISCH It is the unshared information that makes group decision making an attractive choice for someone who wants to increase decision quality: Because diﬀerent members bring diﬀerent knowledge to the table, having a group discuss and decide a decision problem should lead to a broader knowledge base for the decision, and this should increase decision quality (e.g., Hollenbeck et al., 1995). Now, unless systematic asymmetries are present that make the advantages (or disadvantages) of certain alternatives particularly unlikely Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 to be shared among members, it is very likely that each group member has individual information prior to discussion that is more or less representa- tive of the overall information. In other words, not everybody knows everything, but most group members should have individual information that makes the best alternative look good (or even superior to all others) right from the beginning, and the worst alternative should also appear to be relatively weak from the start onwards. Such a situation is called a manifest proﬁle, because the rank-order of alternatives is evident from the beginning (Lavery, Franz, Winquist, & Larson, 1999). The important point is: Whenever information is distributed in this way, groups can hardly realise synergy in group decision making. Discussing the issue in the group might enlarge each member’s individual knowledge base, but this will not have substantial impact on the ﬁnal decision: Because it was evident from the beginning which alternative was best (and, hence, at least the large majority of group members should have preferred it from the beginning), having a group discuss and decide the issue hardly has any potential to increase decision quality (for a related point, see Hastie & Kameda, 2005). This reasoning directly implies that, if we expect group decision making to lead to synergy, and if we want to check whether this really is the case (or how such synergy can be facilitated), we have to look at situations where the correct or best choice is not evident from the beginning. Such situations are called hidden proﬁles (Stasser, 1988). In the next section we will brieﬂy describe these situations, and we will highlight the, more or less, disappointing performance of groups when dealing with hidden proﬁles. We will then outline a general model categorising the processes that are necessary for solving hidden proﬁles and, hence, for the realisation of synergy (and even strong synergy!) in group decision making. In addition we will map the empirical hidden proﬁle literature onto these categories, with a particular focus on our own studies during the last 10 years. In the ﬁnal section we will draw implications from the model and will discuss whether and to what extent these considerations can be generalised to other types of group tasks like, for example, problem solving. SYNERGY IN GROUP DECISION MAKING 309 HIDDEN PROFILES: THE (OFTEN MISSED) OPPORTUNITY TO REALISE SYNERGY IN GROUP DECISION MAKING The hidden profile paradigm As previously outlined, systematic synergy in group decision making is only possible if the correct choice (i.e., the best alternative) cannot be detected by group members based on the individual information that they privately have prior to discussing the issue in the group. This is exactly the case in the Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 hidden proﬁle paradigm, introduced by Stasser and Titus (1985). In a hidden proﬁle one of the decision alternatives is superior to the others when the full information set (i.e., the total of the information available at the group level) is taken into account. However, this superiority only becomes evident when the group members’ unshared (i.e., uniquely held) information is pooled and integrated into a revised appraisal of the alternatives. In the case of a hidden proﬁle, much of the information supporting the best alternative (i.e., advantages of this alternative and disadvantages of other alternatives) is unshared, whereas the shared information either supports one or more suboptimal alternatives or no alternative at all. Hence hidden proﬁles can be solved only if group members exchange and integrate their unshared pieces of information and thereby are able to detect the decisional implication of the full information set. An example of such a hidden proﬁle information distribution is given in Table 1; we have used this material in some of our own studies on group decision making (e.g., Schulz-Hardt, Brodbeck, Mojzisch, Kerschreiter, & Frey, 2006). The participants play the role of personnel managers in an airline company that is looking to hire a new pilot for long-distance ﬂights. The material is constructed for use in three-person groups (although it can easily be modiﬁed for other group sizes). When taking all information into account, it is clear that Candidate C is vastly superior to Candidates A, B, and D: Whereas Candidate C has seven advantages and only three disadvantages, Candidates A, B, and D each have only four advantages, but six disadvantages. As we have mentioned above, we operate with the simplifying assumption that the diﬀerent advantages and disadvantages are (almost) equal in strength. Therefore in empirical hidden proﬁle studies extensive pretests are necessary to make sure that this assumption is met and that participants do, on average, perceive similarity in importance and reliability between the diﬀerent pieces of information. However, whereas all of the advantages of Candidates A, B, and D as well as all the disadvantages of Candidate C are shared, all disadvantages of Candidates A, B, and D as well as most advantages of Candidate C are unshared (and each group member has the same amount of unshared information). As a consequence each group member individually knows the 310 SCHULZ-HARDT AND MOJZISCH four advantages of A, B, and D, but only two of the disadvantages of each of these candidates. In contrast, each group member is aware of all three disadvantages of Candidate C prior to discussion, but knows only three of this candidate’s seven advantages (the one shared advantage plus two of the six unshared advantages). Hence for each group member Candidate C should appear as the weakest candidate prior to discussion (and, hence, group members should favour A, B, or D), although this candidate is, in fact, the best choice based on all information. As outlined earlier, the correct Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 choice cannot be identiﬁed prior to discussion. And, as we can see from Table 1, this information distribution bears a large potential for synergy, because by choosing Candidate C the groups can make a vastly superior decision compared to what should be the outcome without information exchange in the group (namely a decision for Candidate A, B, or D). The failure of groups to solve hidden profiles Since Stasser and Titus’ seminal study published in 1985, many empirical investigations of group decision making in hidden proﬁles situations have been conducted. So far, these empirical investigations have been —with one very recent exception (Stasser, Abele, & Vaughan Parsons, 2012)— exclusively laboratory experiments. This focus does not neglect the importance of hidden proﬁles in the ﬁeld, but for practical reasons investigating hidden proﬁles in ﬁeld settings is almost impossible: Detecting them would require that the researcher knows the whole set of decision- relevant cognitions of all group members prior to the ﬁrst group meeting, and it is hard to see how that could be realised in the ﬁeld. In contrast, in a laboratory setting the researcher has perfect control over the decision- relevant information available to group members and, thus, has the ability to induce any information distribution of interest. The typical procedure in these laboratory experiments is to invite participants to the lab and provide each of them with general instructions as well as individual information about a decision case. Usually university students serve as participants in these experiments, but there are also studies with a practitioner sample such as, for example, medical teams working on medical disease diagnosis (e.g., Larson, Christensen, Franz, & Abbott, 1998). There are typically two (Winquist & Larson, 1998), three (e.g., Stasser & Titus, 1985), or four (e.g., Schulz-Hardt et al., 2006) decision alternatives. After participants have studied their individual materials and have evaluated the diﬀerent decision alternatives based on this information, the information materials are collected by the experimenter, which means that participants discuss the decision case from memory. Although this might, at ﬁrst glance, appear to be an artiﬁcial constraint, a closer look reveals that this procedure is not that ecologically invalid, because group members hardly ever have a SYNERGY IN GROUP DECISION MAKING 311 written excerpt of all their decision-relevant knowledge with them when discussing and deciding an issue in a group. Even in cases where written (or electronically stored) materials about the decision case exist (e.g., the candidate information ﬁles in an appointment committee), and hence group members can ﬂip through these materials during discussion, it is often the private (unshared) knowledge of the group members that has a particular impact on the ﬁnal decision (e.g., ‘‘I know that XY just bought a house in Aston, so it is highly unlikely that he will move over to London’’). Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 Prior to group discussion, participants usually receive further instruc- tions. In particular participants are made aware of the fact that the individual information sets that they have received are not identical: Whereas some pieces of information are shared among them, other parts of the individual information diﬀer among members. Furthermore, it is typically emphasised that the decision case features a superior alternative that can be found on the basis of all the information that is available to group members as a whole, and that the participants’ task is to identify this superior alternative. At least in our own experiments, it is also standard to have a ﬁnancial incentive for groups to detect this superior alternative. Hence, unlike in decision-making groups in practice that often lack the above-mentioned features, preconditions for the realisation of synergy in group decision making are almost optimal in these experiments: There is no deception involved, full transparency about the requirements of the situation is given, and there are no side-bets that might motivate group members to favour suboptimal alternatives (e.g., they have no external incentive for ‘‘pushing through’’ their own candidate in a personnel selection task)—the only external incentives given are in favour of cooperatively trying to ﬁnd the best alternative. In spite of these almost ideal conditions, more than 25 years of research on hidden proﬁles have shown that groups actually fare pretty badly at detecting and solving hidden proﬁles. For example, in the seminal Stasser and Titus (1985) study, 83% of the groups chose the best alternative in the manifest proﬁle condition2, but only 18% did so in the hidden proﬁle conditions. Similarly, in the Schulz-Hardt et al. (2006) study that used the case material from our example in Table 1, all groups in the manifest proﬁle condition made the correct decision, whereas only 35% of the groups in the hidden proﬁle conditions picked the best alternative; although, as we have seen, this alternative is vastly superior to any of its competitors. These two studies are no outliers: As systematic reviews of the hidden proﬁle literature (e.g., Brodbeck, Kerschreiter, Mojzisch, & Schulz-Hardt, 2007) as well as a 2 In this condition participants were given the full information set right from the beginning. However, we still label such a condition ‘‘manifest proﬁle’’, because the deﬁning feature (initial information set is representative of the overall information set) is necessarily true in such a case. 312 SCHULZ-HARDT AND MOJZISCH TABLE 1 An example of a hidden profile task Candidate A Candidate B – can anticipate dangerous situations – keeps calm in a crisis – is able to see complex connections – known to be 100% reliable – has excellent spatial vision – good at assessing weather conditions – has very good leadership qualities – has excellent computer skills – is sometimes not good at taking – can be grumpy criticism – can be uncooperative Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 – can be unorganised – has a relatively weak memory for – is regarded as a show-oﬀ numbers – is regarded as being not open to new – makes nasty remarks about his ideas colleagues – is unfriendly – is regarded as pretentious – eats unhealthily – sometimes adopts the wrong tone when communicating Candidate C Candidate D – can make correct decisions quickly – responds to unexpected events – handles stress very well adequately – creates a positive atmosphere with his – can concentrate very well crew – solves problems extremely well – is very conscientious – takes responsibility seriously – understands complicated technology – is regarded as arrogant – puts concern for others above – has relatively weak leadership skills everything – is regarded as a ‘‘know-it-all’’ – has excellent attention skills – has a hot temper – has diﬃculty communicating ideas – is considered moody – is regarded as egocentric – is regarded as a loner – is not very willing to further his education The decision case deals with an airline looking to ﬁll the position of a pilot for long-distance ﬂights; the four Candidates A, B, C, and D are characterised by the attributes listed; shared information is given in bold. recent meta-analysis of hidden proﬁle studies (Lu, Yuan, & McLeod, 2012) consistently show, groups typically fail to solve hidden proﬁles. To illustrate this failure let us take a closer look at the results from one of our own hidden proﬁle studies (Greitemeyer, Schulz-Hardt, Brodbeck, & Frey, 2006). In this study three-person groups worked on four subsequent decision cases. Three of these cases had a hidden proﬁle distribution of information, whereas one of them was a manifest proﬁle. Each of the cases comprised three decision alternatives, forming a clear rank-order in decision quality. Whereas in the manifest proﬁle case this rank-order was already present in the group members’ individual information sets (i.e., the alternative that was best based on all information also had the best relation SYNERGY IN GROUP DECISION MAKING 313 of advantages and disadvantages in each group member’s individual information set), the members’ individual information sets in the hidden proﬁle cases reversed this rank-order. In other words, in the hidden proﬁle cases the particular alternative that was best based on all information initially appeared to be the least suited to each group member (vice versa for the worst alternative). Now, as the results of this study show, groups had no diﬃculty in detecting this rank-order when dealing with a manifest proﬁle distribution of Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 information: The vast majority of groups (89%) chose the best alternative, whereas only a small proportion of groups chose the second best (7%) or the worst alternative (4%). The opposite happened in the hidden proﬁle cases. Most of the groups chose the worst alternative (87%)—that is, the one that appeared to be superior in the group members’ initial individual materials— and only very few groups decided for the second best (6%) or the best alternative (7%). In other words: Groups are able to realise a high decision quality in situations where their individual members would already realise a high decision quality (manifest proﬁle). Ironically, however, they fail to do so in the very situation where they have the opportunity to achieve synergy by realising a surplus in decision quality compared to what would have been possible without group interaction. To further illustrate this point we have reanalysed the data published in our Greitemeyer et al. (2006) study. For these purposes we have calculated the baselines for the tests of weak and strong synergy, and we have done that separately for the manifest proﬁle case and for the hidden proﬁle cases. As already outlined, strong synergy occurs only if the group decision is better than the best individual decision that any of the group members has initially made. With regard to this, all groups contained at least one member who favoured the best alternative prior to discussion. As a consequence there was no room for strong synergy (ceiling eﬀect). This picture was almost the same for weak synergy: For weak synergy to occur, the group decisions have to be better than the average individual decisions. In the manifest proﬁle case 90% of the group members already chose the best alternative prior to discussion, 5% chose the second best, and another 5% chose the worst. Thus, once again we can see that in a manifest proﬁle hardly anything can be gained from having a group decide the issue—and hardly anything was gained, because the real group decisions were almost identical with this latter baseline (as illustrated in Figure 1a). This is completely diﬀerent in the case of a hidden proﬁle. As can be seen in Figure 1b, here a very high potential for weak synergy was present: Prior to discussion, 88% of the group members favoured the worst option, 4% favoured the second best, and 8% favoured the best. Hence average individual decision quality was low, as would be expected given the insuﬃcient and misleading individual information. Even for strong synergy 314 SCHULZ-HARDT AND MOJZISCH Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 Figure 1 (a) Individual decision quality prior to discussion vs group decision quality in the manifest proﬁle case (Greitemeyer, Schulz-Hardt, Brodbeck, & Frey, 2006). (b) Individual decision quality prior to discussion vs group decision quality in the hidden proﬁle cases (Greitemeyer, Schulz-Hardt, Brodbeck & Frey, 2006). the potential was high: In about three out of four cases all three members initially favoured the worst option, which means that strong synergy would occur if the second best or best alternative is chosen by these groups. Furthermore, in the 7% of the cases where the best individual decision was for the second best option, a group choice of the best alternative would also have yielded strong synergy (and only for the 17% of the groups containing one member who initially preferred the best option, no strong synergy was possible). However, as the comparison with the real group choices in Figure 1b shows, both types of synergy failed to materialise, because group decision quality was more or less at the level of the average individual SYNERGY IN GROUP DECISION MAKING 315 decisions—although groups theoretically had much more information than their individual members initially had, and although that additional information would have allowed them to detect the best option. Taken together, this reanalysis shows that groups reach a fairly high decision quality in situations where—in terms of decision quality—nothing can be gained from having groups as decision makers. In contrast, in the very situation where they have the potential to achieve synergy and thereby make group decision making pay oﬀ in terms of decision quality, Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 they typically fail to do so (for a more extensive discussion of this issue, see Mojzisch & Schulz-Hardt, 2011). As a consequence, if we want to continue using groups as a tool for increasing decision quality, we have to ﬁnd out what particular processes hinder groups from solving hidden proﬁles and, in turn, how synergy in group decision making can be facilitated. In two earlier publications (Brodbeck et al., 2007; Mojzisch & Schulz- Hardt, 2006) we have already suggested a model that tries to give answers to these questions. However, some of our recent studies have led to the necessity to substantially revise and further develop this model, particularly by shifting the focus from biases in group information processing to the intensity of group information processing. In the next section we will outline our new, revised model, and we will describe hidden proﬁle studies supporting and illustrating the claims made in this new model. HURDLES FOR SYNERGY IN GROUP IN DECISION MAKING: A GENERAL MODEL AND EMPIRICAL EVIDENCE A general categorisation scheme As outlined in the previous sections, a true potential for synergy in group decision making arises only if the exchange of information in a group changes the decisional implication of the information available to group members. From this boundary condition we derive the general proposition that the realisation of such synergy requires that decision-relevant information is (a) discussed in the group and (b) individually processed by group members. Discussion encompasses the introduction of information in the group (i.e., mentioning information for the ﬁrst time) and its repetition by the same or other group members (i.e., something that was introduced before is taken up at a later point in the discussion). By processing we mean the encoding, storage, interpretation, and evaluation of information that is discussed in the group. The necessity of both processes for the realisation of synergy is hardly surprising in itself, but since most research on hidden proﬁles has only addressed the discussion of information, without taking 316 SCHULZ-HARDT AND MOJZISCH into account how and to what extent the information discussed is processed by the group members, we felt that this point has to be emphasised. Furthermore, it can be derived that both the discussion of information and the processing of the information discussed have to be characterised by two qualities: a suﬃcient intensity and a suﬃcient lack of bias: Only if a suﬃcient amount of information is exchanged, and only if the group members process the information discussed in the group suﬃciently to realise its implications, do group members have a chance to detect the Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 superiority of an alternative that initially did not appear to be that strong. At the same time, if biases cause the group to discuss too little information in favour of this particular alternative, or cause group members to evaluate this information too negatively, the hidden proﬁle will also not be solved (what we particularly mean by ‘‘bias’’ and what would be an unbiased discussion or processing will be speciﬁed in the corresponding sections below). Once again, this proposition is rather straightforward, but due to the fact that empirical hidden proﬁle research has been far more been concerned with biases (in particular, with discussion biases) than with the intensity of the discussion and processing of information, it is important to emphasise that both qualities are necessary. The combination of the two processes discussion and processing with the two characteristics intensity and bias gives a 2 6 2 matrix which is illustrated in Figure 2 and which is the basic organising principle of our model. The processes in each of the four cells have to be positively manifested (i.e., suﬃcient intensity and lack of bias in both discussion and processing of the information discussed) in order to allow groups to solve hidden proﬁles. In contrast, a negative manifestation of any of the processes in any of the four cells (i.e., insuﬃcient intensity, insuﬃcient processing, biased discussion, biased processing) can be suﬃcient to completely hinder groups from solving hidden proﬁles. Unfortunately the ‘‘default mode’’ of both how groups exchange information and how information from group Figure 2. Classiﬁcation scheme for impairments to the solution of hidden proﬁles. SYNERGY IN GROUP DECISION MAKING 317 discussions is processed individually works against the solution of hidden proﬁles. In other words, it is rather the rule than the exception that the processes in each of the four cells of Figure 2 have negative manifestations. In the following we will illustrate this unfortunate situation for each of the four cells by referring to exemplary hidden proﬁle research, with a particular focus on our own studies. Discussion bias Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 As already outlined, hidden proﬁle research so far has prioritised group discussion at the expense of processing of the information discussed, and it has been far more concerned with biases than with the general intensity of group discussion and, to a lesser extent, individual information processing. Hence most previous hidden proﬁle studies have addressed biases in group discussions as a reason for the failure of groups to solve hidden proﬁles. Generally, we speak of a discussion bias if the mentioning and/or repetition of information is not representative of the overall information set available to the group members. For example, if in a personnel selection case 60% of the available pieces of information are positive (i.e., advantages of particular candidates) and 40% are negative (i.e., disadvantages), and if then 70% of the pieces of information that are introduced into discussion are negative, we would call this a bias in favour of negative information. At least for the ﬁrst 10 years since the initial Stasser and Titus (1985) study, research on information pooling and hidden proﬁles in groups almost exclusively dealt with one particular discussion bias: the bias in favour of shared information. As we have seen in our exemplary decision case from Table 1, solving a hidden proﬁle requires the group to exchange and integrate their members’ unshared information, because in a hidden proﬁle a large portion of the information in favour of the best alternative is—by deﬁnition—unshared. Now, the most robust and well-replicated ﬁnding in the group information pooling literature is that groups discuss more shared than unshared information (for a summary, see the meta-analysis by Lu et al., 2012). Speciﬁcally, this discussion bias means that more shared than unshared information is introduced into group discussion and, once introduced, shared information is also repeated more often than unshared information, with the latter being true both for self-repetitions (i.e., the same member who introduced the piece of information also repeats it) as well as for repetitions by other group members (e.g., Larson, Foster-Fishman, & Keys, 1994). To put it crudely, group discussions focus on what everybody has already known prior to discussion. This bias stems from stochastic rather than psychological processes: The group will only fail to discuss a shared piece of information if all group members independently forget to mention or are unwilling to mention this 318 SCHULZ-HARDT AND MOJZISCH particular piece of information. In contrast, if an unshared piece of information is not mentioned by the one group member who holds it, there is nobody who can compensate for this neglect. Thus, as long as the probability of mentioning any given piece of information is not 0 or 1, a stochastic bias in favour of shared information is inevitable (Stasser & Titus, 1987).3 Importantly, the critical information for solving a hidden proﬁle (i.e., advantages of the best and disadvantages of the other alternatives) is not Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 only predominantly unshared, but is also predominantly inconsistent with the group members’ initial preferences. Because in a hidden proﬁle the best alternative is, by deﬁnition, not evident from the group members’ individual information, most or all members will favour other alternatives prior to discussion. Thus the advantages of the best alternative will be inconsistent with the members’ initial decision preferences, as will be the disadvantages of the particular alternative that a group member initially favours. In our example from the previous section (Table 1), group members can be expected to favour either A, B, or D prior to discussion. Thus most of the pieces of information that are critical for choosing C (the best alternative)— the advantages of C as well as the disadvantages of A, B, and D—will be preference-inconsistent. Interestingly, in their seminal study Stasser and Titus (1985) had already predicted that group discussions would be biased not only in favour of shared information, but also in favour of preference-consistent information. However, it took more than a decade before the ﬁrst experimental test of this idea was published (Dennis, 1996). In his experiment six-person groups that either worked face-to-face or used computer-mediated communication discussed which of three students should be given admission to the university. Information was distributed as a hidden proﬁle. Results showed the already well-known bias in favour of shared information, but when analysing the unshared information separately Dennis showed that, independent of the communication mode, members mentioned more information supporting their initial preferences than neutral information 3 For the repetition of shared versus unshared information this bias seems to be less inevitable, because once an unshared piece of information is introduced, all group members have the possibility of repeating it. Some researchers have thus tried to explain this part of the discussion bias in favour of shared information with psychological concepts, particularly referring to social validation: Because the veracity of shared information can be conﬁrmed by other group members, it appears more credible and, in consequence, gets repeated more often (e.g., Wittenbaum et al., 1999). We will not go into the details of this reasoning because, according to some critical replications that we conducted, the evidence for this approach is based on a methodological shortcoming of the original studies and when controlling for this shortcoming no evidence for a social validation explanation of biased repetitions favouring shared information can be found (Mojzisch, Kerschreiter, Faulmüller, Vogelgesang, & Schulz-Hardt, 2012). SYNERGY IN GROUP DECISION MAKING 319 or information contradicting their initial preferences. The shared informa- tion had to be left out of this analysis, because in a hidden proﬁle sharedness and preference-consistency of information are, at least partially, confounded (i.e., a large part of the shared pieces of information is also preference- consistent). To avoid this problem and hence to demonstrate that there is a general discussion bias in favour of preference-consistent information (independent of whether information is shared or unshared), in a very recent study Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 (Faulmüller, Mojzisch, Kerschreiter, & Schulz-Hardt, 2012) we used a non- hidden proﬁle information distribution (all alternatives were designed to be equally attractive). The second aim of this study was to show that—in contrast to the discussion bias in favour of shared information—the discussion bias in favour of preference-consistent information has psycho- logical roots. In two experiments participants exchanged information with a partner in order to prepare a collective personnel decision. Because the partner was either a bogus partner in a written information exchange (participants received standardised written messages prepared by the experimenter) or a confederate of the experimenter in a face-to-face discussion, we had full experimental control over the partner’s communica- tion behaviour. In each of the two experiments we manipulated whether the partner favoured the same or a diﬀerent job candidate than the participant, and whether or not the partner indicated that s/he understood why the participant favoured her/his preferred candidate. To illustrate, if a participant in the experimental condition with ‘‘congruent preferences and understanding’’ had a preference for candidate A, the bogus partner said, ‘‘I, too, believe A is the better candidate, and, like you, I would choose him. And, given the information you just passed on to me, I also understand why you prefer him.’’ In contrast, in the experimental condition with ‘‘congruent preferences and non-understanding’’, the bogus partner said, ‘‘I, too, believe A is the better candidate. But, given the information you just passed on to me, I don’t understand why you prefer him.’’ Analyses of the written messages transferred to the partner (Experiment 1) or the videotapes of the discussions (Experiment 2) allowed us to measure to what extent the participant mentioned preference-consistent versus preference-inconsistent information about the job candidates. Across both experiments there was a clear predominance of preference-consistent information being mentioned. Furthermore, whereas the partner’s preference (similar or dissimilar to the participant’s preference) hardly aﬀected this predominance, preference- consistent information sharing was particularly strong when the partner said that she did not understand why the participant held her or his preference, as compared to when the partner stated that s/he understood this preference. Figure 3 illustrates this pattern for the results of Experiment 2 (the face-to- face discussions). 320 SCHULZ-HARDT AND MOJZISCH Downloaded by [T&F Internal Users], [Sarah Scoffield] at 05:57 23 January 2013 Figure 3. Preference-consistent information sharing (number of preference-consistent pieces of information minus number of preference-inconsistent pieces of information communicated by the participants) dependent on preference congruity and preference understanding (Faulmüller, Mojzisch, Kerschreiter & Schulz-Hardt, Expt. 2, 2012). These ﬁndings inform us about possible motivational underpinnings of the discussion bias in favour of preference-consistent information. Whereas a motivation to convince the discussion partner does not seem to be the driving force here (in this case we would have expected a particularly strong bias if the partner’s preference is incongruent with the participant’s preference), a motivation to be understood by the partner could be an underlying motive of this bias: As Stasser and Titus (1985) had already assumed, group members might implicitly take an ‘‘advocacy role’’ during discussion; that is, their conversation behaviour is driven by the attempt to explain to the other group members why they hold a particular preference— and explaining a preference means mentioning information that is consistent with this preference. In addition, one of our other recent experiments (Mojzisch, Grouneva, & Schulz-Hardt, 2010) hints at the possibility that cognitive factors might also be responsible for the preference-consistent discussion bias: As we will outline below in the section on biased information processing, people have a general tendency to attribute a higher quality to preference-consistent as compared to preference-inconsistent information. Hence, if people simply want to communicate the best information (in terms of credibility, reliability, and information strength), they should predominantly mention preference-consistent information. In other words, group members might SYNERGY IN GROUP DECISION MAKING 321 discuss in a preference-consistent manner without intending to do so. The Mojzisch et al. (2010) study provides some data in accordance with this interpretation. In this study participants ﬁrst received information about two decision alternatives and were asked to indicate their preference. Next they were asked to evaluate the quality of each piece of information. Thereafter partici

Achieve Synergy in Group Decision Making PDF

Document Details

Tags

Related

Summary

Full Transcript

Upgrade to continue