群体决策过程中的信息取样偏差

一般认为,决策群体的优势是能综合各个成员拥有的不完整信息,形成对所有决策备择方案的无偏差的认识,从而作出最佳选择。但“群体讨论中的有偏差信息取样模型’认为,群体在决策中往往表现出倾向于讨论两种信息：(1)各成员讨论前都拥有的信息;(2)支持成员在讨论前所偏好的信息。该实验通过模拟人员选拔决策形式的实验室实验,对这一模型的假设进行验证。研究结果表明,大多数群体选择了一开始受到大多数成员支持的候选人,而非实际的最佳人选;群体讨论非但未纠正,反而加强了成员对候选人原有的歪曲印象。     

The Collective Preference for Shared Information

Decision-making groups prefer to discuss shared information that all members know instead of unshared information that a single member knows. This bias toward discussing shared information can lead groups to make suboptimal decisions when unshared information is critical for good decision making. This preference for discussing shared information may stem from group members' positive evaluations of each other's task capabilities when shared information is communicated. Members who already are perceived as capable (i.e., those high in status, experts, and leaders) need not bolster their image by communicating shared information. Instead, they discuss unshared information more than members perceived as less capable. As members low in status gain respect by communicating shared information, they may risk mentioning unshared information later during discussion. Assigning group leaders, informing members of their expert roles, and allowing ample time for discussion may increase groups' discussion of unshared information.     

自由讨论条件下群体决策质量的影响因素

通过实验室实验考察了自由讨论条件下群体决策质量的影响因素,并对Stasser所提出的信息取样模型进行了验证,结果发现：(1)部分证实Stasser的信息取样模型。在信息不分享的条件下,如果讨论前群体成员的偏好比较一致时,群体的确倾向于讨论分享信息和群体所偏好的候选人的信息;但如果讨论前的偏好不一致或任务难度较低时,这一结论难以成立。(2)自由讨论条件下,群体规模的增加会增加分享信息的讨论量,而对非分享信息的讨论程度则无显著影响。而在任务难度方面,只有任务难度较大的情况下才有分享信息的讨论优势。     

Information sampling and group decision making: the effects of an advocacy decision procedure and task experience

Group discussions tend to focus on information that was previously known by all members (shared information) rather than information known by only 1 member (unshared information). If the shared information implies a suboptimal alternative, this sampling bias is associated with inaccurate group decisions. The present study examines the impact of 2 factors on information exchange and decision quality: (a) an advocacy group decision procedure versus unstructured discussion and (b) task experience. Results show that advocacy groups discussed both more shared and unshared information than free-discussion groups. Further, with increasing experience, more unshared information was mentioned in advocacy groups. In contrast, there was no such increase in unstructured discussions. Yet advocacy groups did not significantly improve their decision quality with experience.     

A social validation explanation for mutual enhancement

Group members experience mutual enhancement when they evaluate one another’s task capabilities more positively because they are discussing shared as compared to unshared information. Two experiments investigated a social validation explanation for mutual enhancement. In Experiment 1, members of dyads read and discussed either shared or unshared information and collectively recalled this information under conditions where accuracy was important or not. Mutual enhancement occurred when need for accuracy was higher, but not when it was lower. In Experiment 2, members of dyads read validating or non-validating information from an outsider after they collectively recalled shared or unshared information. Members who discussed unshared information and received no validating communication from the outsider gave one another the lowest task capability ratings compared to members who discussed shared information and/or received validating communication from the outsider. All of these results are consistent with a social validation explanation for mutual enhancement.     

Children's collaborative recall of shared and unshared information

This study examined age differences in collaborative inhibition and the role of inter‐subjectivity, collective information sampling (CIS) and collaborative inhibition for the collaborative recall of shared and unshared information in groups of 7‐ and 9‐year‐old children. Three‐hundred and thirteen 7‐ and 9‐year‐old children recalled memorized wordlists either in real or nominal groups of three. All group members either recalled the same items, or each group member was given some unique items. Nine‐year‐olds, but not 7‐year‐olds, recalled significantly more items in nominal than real groups, a phenomenon called collaborative inhibition. Groups whose interactions were characterized by higher numbers of inter‐subjective exchanges recalled fewer words than groups low in inter‐subjectivity. In both age groups, a higher proportion of shared compared with unshared information was recalled consistent with processes of CIS. However, 7‐year‐olds recalled more unshared items than predicted, suggesting that collaborative inhibition additionally contributes to the recall of shared and unshared items.     

The effect of information distribution on collaborative inhibition

Two experiments examined collaborative memory for information that was studied by all group members (shared items) and information that was studied by only a subset of group members (unshared items). In both experiments significant collaborative inhibition (reduced output of the collaborative groups relative to the pooled output of individuals) was obtained for both shared and unshared information. In Experiment 1 the magnitude of collaborative inhibition was larger for unshared items than for shared items, possibly because unshared items were less likely to be acknowledged and thus incorporated into the groups’ recall. In Experiment 2 the magnitude of collaborative inhibition for shared and unshared information was equivalent once all participants were provided with the category name associated with the shared and unshared items. The results of the experiments are discussed in relation to the retrieval strategy disruption hypothesis of collaborative inhibition and the role of social process variables, such as acknowledgement, in influencing collaborative inhibition across situations involving memory of shared and unshared information.     

The Effect of Individual Difference Variables on Information Sharing in Decision‐Making Groups

We examined need for cognition, social desirability, and communication apprehension for their influence on the mention and repetition of shared and unshared information in 8‐person decision‐making groups. Both need for cognition and social desirability influenced the discussion of shared and unshared information in decision‐making groups. The findings indicate that increasing motivation to participate in group discussions may not help overcome the bias favoring shared over unshared information. Additionally, there are indications that social desirability increases the repetition of shared information. This finding is consistent with the idea of mutual enhancement (i.e., the idea that group members discuss shared information because it enhances their position with other group members).     

The effect of information distribution on collaborative inhibition

Two experiments examined collaborative memory for information that was studied by all group members (shared items) and information that was studied by only a subset of group members (unshared items). In both experiments significant collaborative inhibition (reduced output of the collaborative groups relative to the pooled output of individuals) was obtained for both shared and unshared information. In Experiment 1 the magnitude of collaborative inhibition was larger for unshared items than for shared items, possibly because unshared items were less likely to be acknowledged and thus incorporated into the groups' recall. In Experiment 2 the magnitude of collaborative inhibition for shared and unshared information was equivalent once all participants were provided with the category name associated with the shared and unshared items. The results of the experiments are discussed in relation to the retrieval strategy disruption hypothesis of collaborative inhibition and the role of social process variables, such as acknowledgement, in influencing collaborative inhibition across situations involving memory of shared and unshared information.     

The dissemination of critical,unshared information in decision‐making groups: the effects of pre‐discussion dissent

Previous research in group decision making has found that in situations of a hidden profile (i.e. the best choice alternative is hidden from individual members as they consider their pre‐discussion information), unshared information is disproportionately neglected and sub‐optimal group choices are highly likely. In an experimental study, three‐person groups decided which of three candidates to select for a professorial appointment. We hypothesised that minority dissent in pre‐discussion preferences improves the consideration of unshared information in groups and increases the discovery rate of hidden profiles. As predicted, consideration of unshared information increased with minority dissent. The expectation of an improvement of group decision quality was partially supported. In diversity groups (i.e. each member prefers a different alternative) consideration of unshared information and group decision quality was significantly higher than in simple minority groups. Results are discussed in the light of theories of minority influence. The benefits of using the hidden profile paradigm with minority and diversity groups for theory development in the area of group decision making are highlighted. Copyright © 2002 John Wiley & Sons, Ltd.     

The sampling of critical,unshared information in decision-making groups: the role of an informed minority

A collective information sampling model and observations of discussion content suggest that decision-making groups often fail to disseminate unshared information. This paper examines the role that a fully-informed minority may play in facilitating the sampling and consideration of unshared information. University students read a mystery and then met in four-person groups to discuss the case. When critical clues were unshared among three members before discussion, a fully informed fourth member (informed minority) promoted the discussion of these critical clues when participants thought the mystery had a demonstrably correct answer (solve set) but not when they thought the clue may have been insufficient to solve definitively the case (judge set). None the less, under both solve and judge sets, the informed minority increased the likelihood that the group would identify the correct suspect. Social combination, information sampling, and minority influence interpretations of the results are discussed. © 1998 John Wiley & Sons, Ltd.     

Motivations, Goals, Information Search, and Memory about Political Candidates

This study investigated the ways in which motivations and goals affect patterns of political information-seeking and the consequent structure of memory about candidates. Undergraduate participants used a computerized system that displayed different layers of information about fictional political candidates; the system recorded the strategies they used to search through this information. Results showed that motivations to engage in effortful processing produced tendencies to engage in within-candidate searches, better recall, and memory structures clustered by candidate. The goal of forming impressions of the candidates, which was expected to lead to within-candidate searching, was in fact modestly associated with weaker tendencies to do so, once effort was taken into account. Impression-formation goals, however, were associated with less attribute-based memory structures. The findings confirm that the manner in which people acquire candidate information has important consequences for the way they store that information in memory, and that these processes vary according to individual motivations and goals.     

Information Processing and Exchange in Mediated Groups

This study examined whether features of mediated group discussion were related to participator judgments. Prior to discussion, participants (N = 138) were asked to memorize a list of qualifications for each of 3 hypothetical candidates for a faculty position. Each list contained shared (given to all members) and unique (given to just 1 member) information. Participants, working in 3‐person groups, then interacted via 1 of 2 computer text environments to choose the best candidate. Following discussion, participants rated each other on a set of participator assessment items. Discussions were coded for units that contained shared or unique information. The analysis revealed that shared contributions were related only to self‐assessment of participation. Unique contributions had a complex relation with self‐assessments—they were positively associated with such judgments only when partners infrequently provided unique information. Technology affected in a variety of ways the relation between the discussion variables and participator assessments. Discussion focuses on the relation between information pooling and interaction, as well as on the role of technology in group process.     

Information Exchange in Context: Effects of Gender Composition of Group,Accountability, and Intergroup Perception on Group Decision Making1

In the present study, the dissemination of shared and unshared information (cf. Stasser & Titus, 1985) in unstructured group discussion is analyzed. In previous experiments, several manipulations of the task itself were introduced, in search of conditions that could facilitate the exchange of unshared information. As such, these studies neglected the potential impact of context variables on group performance. In the present study, the effects of accountability, intergroup perception, and gender composition of group on information exchange are investigated. The results indicate that heterogeneous groups exchanged more information when a second group was present. Information exchange was not promoted by the presence of an outgroup for homogeneous groups. Groups in the accountability condition displayed more difficulties to reach agreement, but this did not lead to the mentioning of more information.     

Facilitating information exchange in small decision-making groups

The present study examined the effects of several conditions on the information flow during unstructured discussion in small groups. We build on the experimental work of Stasser and his colleagues on information sampling. The results of their research suggest that groups often fail to exchange information effectively. Three experiments with a 2×2 factorial design were set up to look for conditions that could facilitate the exchange of unshared information in particular. In a first study, we replicated the experiment of Stasser and Titus (1987). The main results of the original study were confirmed; additional discussion content analysis of video-recorded material made further differentiation possible. Apart from the mere mentioning of items during discussion, the reactions to and the repetition of the items were analysed. In the second experiment we introduced partially shared information and we made group members aware of the unique information that they can contribute. In the third experiment the nature of the task was manipulated and an attempt was made to prevent an information bias in favour of decision-supporting items. Conditions that significantly enhance the dissemination of unshared information were found. An empirical basis was gathered so that the guidelines for effective information sampling as they were laid down by Stasser (1992) could be reformulated.     

Effects of Consensus Information and Task Demonstrability on Preference-Consistent Information Evaluation and Decision Quality in Group Decision Making

Groups often fail to solve hidden profiles even when all information is exchanged. This is partly due to biased evaluation of information. We examined the effects of consensus information and task demonstrability on preference-consistent information evaluation and decision quality. The results showed that the evaluation of unshared but not shared information was moderated by consensus information and task demonstrability. For unshared information, majority members exhibited a higher evaluation bias favoring preference-consistent information than minority members. Task demonstrability reduced the evaluation bias only when group members received no information about the other members' preferences. Finally, majority members were less likely to solve the hidden profile than minority members, and this was partially mediated by the evaluation bias favoring preference-consistent unshared information.     

MAU程序和自由讨论的群体决策质量比较

采用实验室研究方法,比较了在信启、不分享条件下多特征效用模型(MAU)决策程序和自由讨论对群体决策质量的影响,并对影响MAU决策质量的主要因素进行了初步探讨。结果发现：(1)MAU程序可以较好地克服群体动力学方面的一些消极影响,表现出挖掘信息(尤其是非分享信息)的充分性。但MAU程序在被试的态度和社会交互作用等心理效标上并未显示出优势;(2)在MAU条件下,群体规模的增大对分享和非分享信息的讨论量无显著影响。而任务难度效应要和群体规模一起才能起到作用。     

不同沟通方式下群体决策信息利用分析

以180名被试组成45个四人群体．完成模拟人事选拔的问题解决型任务,通过对群体决策过程的信息提取数量和提取次序的分析,考察不同的沟通方式和信息分享程度对群体决策信息利用的影响。结果表明。（1）在不同沟通方式下,决策成员的讨论前偏好对最终决策的影响很大。（2）在不同的沟通方式下,分享程度不同的信息在决策过程中的提取具有差异,体现了不同的规律。     

How does background information improve memory for text content?

In two experiments, we investigated whether reading background information benefits memory for text content by influencing the amount of content encoded or the organization of the encoded content. In Experiment 1, half of the participants read background information about the issues to be discussed in the text material, whereas half did not. All the participants were then tested for free recall and cued recall of text content. Free recall was greater for individuals who read issue information than for those who did not. The groups did not differ on cued recall, suggesting that background information did not facilitate the encoding of more text content. Measures of representational organization indicated that increased recall in the issue information group resulted from better organization of content in memory. Experiment 2 extended these findings, using background information about text sources, demonstrated that the efficacy of background information depends on the semantic relationship between that information and text content.     

Preference-consistent evaluation of information in the hidden profile paradigm: beyond group-level explanations for the dominance of shared information in group decisions

Common explanations for the failure of groups to solve so-called hidden profiles focus on group processes, namely insufficient discussion of unshared information and premature consensus on a suboptimal alternative. As 2 experiments show, even in the absence of such group processes, hidden profiles are hardly ever solved. In Experiment 1, participants first received individual information about a personnel selection task and then read a group discussion protocol containing full information exchange. If the individual information was misleading (hidden profile), most participants failed to detect the correct alternative. In Experiment 2, it was determined that this effect is due to preference-consistent evaluation of information that constitutes an individual-level process mediating the failure of group members to solve hidden profiles.