Abstract
Members of 43 three-person groups completed a deductive logic task. Each member was provided with unique clues and information indicating the utility of each clue. Objective validity varied because some clues were essential to problem solution and others were not useful. Social validity was manipulated by informing participants that some clues would likely be useful and that other clues were not likely to be useful. Clues were discussed more frequently if either objective validity or social validity was high, but the effects were not additive. Clues were discussed less frequently in the low objective validity–low social validity condition than in all other conditions. Alignment of both types of validity facilitated the group’s ability to focus on relevant information.
Groups are increasingly used to perform important work (Devine, Clayton, Philips, Dunford, & Melner, 1999; Kozlowski & Ilgen, 2006) and much of that work involves the integration and utilization of knowledge. Consistent with this trend, the view of groups as information-processing units has emerged (Hinsz, Tindale, & Vollrath, 1997; Larson & Christensen, 1993). Because a group has greater knowledge and cognitive capacity than an individual member, groups have greater resources to solve problems (Hackman, 1987; Larson, 2010; Steiner, 1972). However, a number of social processes interfere with the optimal utilization of member resources and groups often fail to fully reach potential on intellective tasks (Laughlin, 1980; McGrath, 1984). Failure to utilize available information can lead to negative consequences. For example, Larson and colleagues (Larson, Christensen, Franz, & Abbott, 1998) found that the failure of medical professionals to share critical information resulted in incorrect diagnoses. Likewise, Foushee (1984) indicated that failure to communicate relevant information was a major factor in airline accidents. Failure to discuss relevant information has been linked to ineffective and even disastrous governmental policy decisions (Janis, 1982) and to lower levels of organizational performance (Campion, Papper, & Medsker, 1996; Cummings, 2004).
The present study examines an issue that affects the group’s utilization of informational resources, the member’s choice of information to discuss. Specifically, it examines two factors that may affect what information is discussed: social validity (social endorsement) and objective validity (the utility of the information for problem solution).
Participation Patterns
Group problem solving is an interdependent activity, but it involves the use of resources that are held by individual members. A group cannot utilize unique member knowledge unless that member shares it with the group (Larson & Christensen, 1993). There are a number of ways in which member information can be excluded from the group discussion. First, participation patterns are typically skewed with a number of members participating to a limited degree (Stephan & Mishler, 1952). Those who dominate the discussion may not be the most informed because the extent of talking is sometimes unrelated to actual expertise (Bottger, 1984; Littlepage, Schmidt, Whisler, & Frost, 1995). In addition, those who talk generally express their ideas and opinions and do little to question or draw out quiet members (Bales, 1955). Furthermore, members who hold unique information only discuss a fraction of that information with the group (Stasser & Titus, 1985, 1987) and this inhibits effective group problem solving (Mesmer-Magnus & DeChurch, 2009).
Talkative members exert greater influence on group decisions (Littlepage et al., 1995), and the group usually chooses the alternative that receives the most support during the discussion (Hoffman, 1979). Such findings underscore the importance of focusing group discussion on the most relevant, important, and reliable information that is held by the members. However, the pattern of information utilization commonly observed in groups suggests that the group discussion frequently fails to incorporate the full range of member knowledge and ideas. Thus, it is important to understand factors that determine which pieces of information a member chooses to discuss with the group.
A number of factors have been linked to participation in group discussions. These include member attributes such as ascendancy, gender, and communication apprehension; group characteristics such as status and group size (Berger & Zelditch, 1977; Bonito & Hollingshead, 1997); and the flow of communication among group members (Bonito, 2007; Parker, 1988). Familiarity with a topic and perceived expertise are also related to participation. Members with a larger number of topic-relevant thoughts make more substantive contributions (Bonito, 2001, 2004). Likewise, being perceived as competent can lead to higher levels of participation (Bonito, 2006).
Choice of Information to Discuss
Whereas previous work has focused on differences between the contribution levels of group members, the current study examines a different issue: member decisions about which items of information to discuss. That is, much previous work has focused on which members contribute, whereas this study focuses on what information is contributed. As indicated by Stasser and by Larson and their colleagues (e.g., Larson & Christensen, 1993; Stasser & Titus, 1985) group members frequently possess more information than they discuss with the group. Consequently, members must select those items to discuss from a larger set of information. Current theory emphasizes the essential nature of discussion of critical information in problem-solving groups. Stasser and Taylor (1991) suggested, “Speaking is the mediating social process that transforms members’ information resources into a group outcome” (p. 675). Larson and Christensen stated, “Introducing problem-relevant information into a group discussion is an element of ‘social’ cognition that can be essential for group problem solving” (p. 6). Therefore, the choice of information to share in a group discussion has important implications for group performance. This study deals with factors that influence the choice of items to discuss.
Initial preferences and prediscussion distribution of information are factors known to affect the extent to which items of information are discussed. Members are more likely to discuss preference-consistent information than preference-inconsistent information (Dennis, 1996). Numerous studies (e.g., Stasser & Titus, 1985, 1987) have demonstrated that information that was available to all members prior to discussion (shared information) is discussed more frequently than information that was available to only one member. In addition, shared information is repeated more during discussion (Larson, Foster-Fishman, & Keys, 1994; Winquist & Larson, 1998). At least three factors account for the greater discussion of shared information. Multiple members can mention an item of shared information while only one person can mention an unshared item, shared information can be socially validated, and in many studies the information distribution leads to member preferences consistent with the shared information (Larson, 2010). The current study seeks to extend this work on the choice of information to discuss by examining effects of the objective validity and of social validation of information. Much of the research on information sharing utilizes the hidden-profile paradigm in which information initially available to all group members supports a suboptimal option and information initially available to only one member supports an optimal solution. In the current study, we do not use the hidden-profile paradigm, but rather we include only unshared information, since it is the discussion of initially unshared information rather than the discussion of shared information that is related to decision quality (Winquist & Larson, 1998). This also allows for a manipulation of social validity that is not confounded with the number of persons having prediscussion access to the information.
Sharing and promoting ideas in a group involves social risks; thus, persons are expected to be more likely to share information they believe to be valid and useful to the group (Stasser, 1999; Wittenbaum & Stasser, 1996). Hinsz (1990, 2004) identified two factors that may make persons feel that information is valid and useful to the group. He distinguished between social validity and objective validity and utilized this distinction in work on group memory performance. He used social validity to refer to the extent that others share one’s memory and objective validity to refer to the accuracy of the memory. Hinsz found that both types of validity were related to group responses in a recognition task. Whereas Hinsz studied group memory, the current study examines group problem solving. In the context of the present study, social validity refers to the extent that another person confirms or endorses information; objective validity refers to the extent that the information is actually useful for solving the problem.
A few studies provide evidence about the effects of objective validity and social validity on discussion. Relevant to objective validity, Steinel and colleagues (Steinel, Utz, & Koning, 2010) suggest that discussion of important information is critical to effective group decisions. A study by Kelly and Karau (1999) suggested that objective validity may affect discussion content; they found that positive and negative facts were discussed more frequently than neutral facts. Based on these findings, we hypothesize as follows:
Hypothesis 1: There will be more discussion of information with high rather than low levels of objective validity. That is, useful information (information that promotes problem solution) will be discussed more frequently than information that is not useful.
Because group members may be uncertain about the validity or utility of information, they may seek social validation, that is, reassurance that the information is valid and useful. Information can be socially validated by other members (Wittenbaum & Bowman, 2004; Wittenbaum, Hubbell, & Zuckerman, 1999), by expert endorsement (Stewart & Stasser, 1995), or by endorsement by the experimenter (Steinel et al., 2010). Information that has been validated by other group members is judged to be more important, relevant, and accurate (Wittenbaum & Bowman, 2004; Wittenbaum et al., 1999) and can have more impact on group decisions (Mojzisch, Schulz-Hardt, Kerschreiter, Brodbeck, & Frey, 2008; Parks & Cowlin, 1996). Additional evidence comes from a series of studies of effects of social motivation on individual choices of information to contribute to an anticipated discussion (Steinel et al., 2010). With cooperative goals (like those in most group decision-making studies), individuals tended to contribute socially validated information. Although tests of this effect were not reported, the pattern of means suggests that clues that had been designated by the experimenter as important were selected for discussion more frequently than clues designated as less important. Together, findings from these studies suggest that information that is socially validated will be discussed more frequently.
Hypothesis 2: There will be more discussion of socially validated information than of information that has not been socially validated.
We expect socially validated information to be discussed more frequently. This implies that the alignment of objective and social validities will lead to discussion focused on objectively valid information:
Hypothesis 3: Discussion will be more focused on truly useful information when objectively valid information has higher social validity than does less objectively valid information.
Discussion of critical information is associated with group decision accuracy (Stasser & Stewart, 1992; Stasser, Stewart, & Wittenbaum, 1995). On the basis of these studies and the expected discussion pattern predicted in Hypothesis 3, we further hypothesize as follows:
Hypothesis 4: Alignment of objective and social validities will result in more effective group decisions.
As a research question, we examined the potential presence and nature of an interaction between objective validity and social validity. Specifically, we explored whether the two types of validity combined in an additive fashion (such that clues were discussed most frequently when both types of validity were high) or in a disjunctive fashion (such that either type of validity would increase discussion, but when one type was high, the level of the other type of validity would be irrelevant).
Research Question 1: Are the effects of objective validity and social validity additive or disjunctive?
Method
Overview
The present study examined the effects of social validity and objective validity on the discussion of information in a problem-solving context. We did this by providing participants with a series of clues and with probability ratings purported to indicate the probability that each clue would be useful for solution of the problem. The level of the clue probabilities can be regarded as information about what an expert (the experimenter) believes about the utility of various clues. Wittenbaum and Stasser (1996) and Steinel et al. (2010) suggested that information from an expert can serve as a source of social validation. Thus, the manipulation of the probability values assigned to clues was designed to create high and low social validity conditions. In reality, some of the clues are useful and others are not. The two levels of actual clue utility provide a manipulation of objective validity. The manipulation of social and objective validity created conditions under which clue probability ratings were accurate (useful clues with high probability ratings and useless clues with low probability ratings) and conditions under which clue probability ratings were inaccurate (useful clues with low probability ratings and useless clues with high probability ratings).
Participants
A total of 129 undergraduates (85 female and 44 males) participated in partial fulfillment of course requirements for introductory psychology classes. Participants were assigned to 43 three-person groups. Three-person groups are frequently used in research on information sharing because they meet all criteria for a group while minimizing recruitment issues. With the task used in the current study, a three-person group allows an equal amount of information to be assigned to each participant.
Materials
The Skiing Vacations Task (Whipkey, 1982) that had been used in previous research (Littlepage, 1991) was adapted for this study. In this task, groups were informed that five married couples vacationed at five ski resorts and that one and only one couple went to each resort. Based on clues that were provided, the group was to correctly match the first and last names of each woman and man, and the destination. From the 9 (useful) clues that were provided with the original skiing vacations problem, it was possible to make 15 deductions and solve the problem. Although none of the participants had the information necessary to solve the problem, through sharing and analysis of this information each group could correctly solve the problem (correctly make all 15 deductions). Examples of useful clues include the following: “Valerie and her husband did not need passports to reach their vacation site,” and “Neither Tammy nor Sue vacationed in Innsbruck.” In addition, we developed nine clues that were of no use in solving the problem. Examples of the useless clues were “Sue fell off the ski lift and broke her new skis,” and “Mr. and Mrs. Norris had wonderful weather on their vacation.” The useful and useless clues provided the operationalization of high and low levels of objective validity. The nine useful and nine useless clues were organized into three complementary sets of clues (A, B, C). Within each group, one member had Set A, another Set B, and the third had Set C. The three sets were constructed such that (a) each set contained three useful and three useless clues, (b) each clue was in one and only one set, and (c) as a unit, the three sets contained all clues.
Design
Each clue was accompanied by a percentage value ostensibly representing the probability that the clue would aid in the solution of the skiing vacations task. Within each clue set, three of the clues were identified with a high probability estimate (80%-90%), and the remaining three clues were identified with a low probability estimate (10%-20%). The high and low clue probabilities served as the manipulation of social validity. In order to counterbalance objective validity and social validity, two versions of each clue set were constructed. In one version (accurate clue probabilities), the clue probability values were consistent with clue utility; that is, social validity information matched the objective validity of the clues. In the other version (inaccurate clue probabilities), the clue probability values were inconsistent with the objective validity of the clues; in this condition, social validity information was at odds with the objective validity of the clues. Participants in both clue probability conditions had the same clues but with the probabilities reversed. Groups were randomly assigned to either high or low clue probability conditions. Table 1 provides an overview of the experimental design.
Summary of the Latin Square Design
The crossing of accurate and inaccurate clue probabilities with useful and useless clues resulted in a 2 (clue probability accuracy) X 2 (objective validity) design with repeated measures on the second factor. Within this design, two cells represent conditions of high social validity: high (accurate) probability values for useful clues, and high (inaccurate) probabilities for useless clues. For clues in these two cells, social validity is high because participants were informed that the clues were useful. Two other cells represent low social validity: low (accurate) probabilities for useless clues, and low (inaccurate) probabilities for useful clues. In these two cells, the clues were identified as nearly useless. The high and low clue probabilities provided two levels of social validity: clues endorsed or devalued by materials provided by the experimenter. This design represents a 2 X 2 replicated Latin Square design (Judd & McClelland, 1989). In this design, each group receives both levels of social validity and both levels of objective validity, but within groups these variables are not fully crossed. The counterbalancing of objective validity and social validity occurs between groups. This design allows for comparison of effects of differing levels of objective validity, differing levels of social validity, and differing levels of clue probability accuracy, but because of complexities involving the choice of appropriate error terms the interaction between objective validity and social validity is best evaluated in targeted analyses rather than a typical 2 X 2 ANOVA (analysis of variance).
Procedure
Participants were escorted to a meeting room and seated around a round table. As a preliminary task to acquaint participants with the nature of deductive logic problems, each group member was given a relatively simple deductive logic problem. Participants were given 3 minutes to individually examine the problem and make deductions. Then, the experimenter explained the solution and the logic supporting each deduction. The experimenter demonstrated the use of a grid to mark valid deductions and alternatives that logically could be ruled out. The grid is a useful aid in organizing information to make valid deductions. After this orientation, each group worked on a more complex logic problem that was the focal task of the study.
A description of the skiing vacation problem was distributed to each group member. Each member received one of the three sets of clues (each containing three objectively useful clues and three that were not useful), with clue probabilities reflecting either the group’s inclusion in the accurate or inaccurate clue probability condition. Participants were instructed not to let other group members read from their clue sheets, but they were told to discuss the problem and otherwise share information as they saw fit. Each group was given 30 minutes to solve the skiing vacations problem. A large problem grid was posted on a blackboard so that the group could better organize their thoughts. On the basis of conclusions reached by the group, the experimenter would use the grid to mark group deductions and alternatives that were ruled out. For example, if the group decided that Valerie Miles went to Aspen, then neither she nor Mr. Miles could have gone to any of the other destinations and none of the other persons could have gone to Aspen.
After 30 minutes, the group was instructed to formally list all deductions on an answer sheet. The number of accurate deductions represents one of the primary dependent variables, group performance. Postexperimental questionnaires were used to assess the effectiveness of the manipulations of social and objective validity.
Coding
Videotapes of the sessions were scored to determine the extent to which groups discussed various clues. Because of a problem with one of the videotapes, the analysis of the group discussions utilized 42 groups, and because another tape was damaged during scoring and coded by only one person, reliability was assessed using 41 groups. In order to assess reliability, two persons independently coded the videotapes. Each coder was given a list of the clues and recorded each instance that the clue was discussed. This provided two ways to assess coding reliability. One measured the extent to which coders agreed about whether each clue was either not discussed, discussed but not repeated, or repeated at least once. Aggregated across clues and groups, there was 86% agreement across coders (79% when corrected for chance agreement using an adaptation of Scott’s Pi; Kappa = .71; Potter & Levine-Donnerstein, 1999). In addition, both the total number of times useful clues were discussed and the total number of times useless clues were discussed were compared across the two coders. The coders showed acceptable levels of agreement for both useful clues, r(39) = .71, and useless clues, r(39) = .93.
Measures
Because both initial mentioning of clues and repetition of clues may differentially influence decisions (Larson et al., 1998) and may reflect different social and cognitive processes (Larson, 2010; Wittenbaum et al., 1999), we utilized multiple discussion measures. At each level of objective validity and social validity, three measures of discussion were computed: clues discussed, clues repeated, and times discussed. The number of clues discussed reflects the number of clues that were discussed at least once. The number of clues repeated represents the number of clues that were discussed at least twice. The number of times clues were discussed represents the total number of instances that clues were discussed. The number of correct deductions served as an additional dependent variable.
Results
Manipulation Checks
In order to assess the social validity manipulation, each participant was presented with a listing of the clues he or she had been given and was asked to recall the probabilities originally assigned to each clue. Participant’s recall of the probabilities assigned to each clue was assessed to verify that subjects attended to the probability values. Recall was classified as correct if it was within 25% of the value that had been provided. The mean number of clue probabilities correctly recalled by group members was 14.28 (or 79.33% of 18 clues, SD = 2.29). Correct recall of probability values was somewhat higher in the accurate clue probability condition (M = 15.23, SD = 1.85) than the inaccurate clue probability condition (M = 13.29, SD = 2.33), F(1, 41) = 9.22, p < .01, η2 = .18. However, in general, participants in both conditions were attentive to clue probabilities. This provides evidence that the manipulation of social validity was effective (that participants knew which clues had been endorsed and which had been devalued).
Two types of information suggest that the manipulation of objective validity was successful. First, we examined postdiscussion judgments of the accuracy of clue weights. Using a 9-point scale anchored by responses strongly disagree (1) and strongly agree (9), each group member indicated his or her agreement that the clue probability values accurately reflected clue utility. The three-member ratings were summed to yield a group rating. Groups in the accurate clue probability condition rated the accuracy of the assigned clue probabilities more highly (M = 18.36, SD = 4.04) than groups in the inaccurate clue probability condition (M = 7.71, SD = 4.24), F(1, 41) = 71.03, p < .01, η2 = .63. Thus, persons receiving accurate clue probabilities tended to agree that they were accurate, and persons receiving inaccurate probabilities tended to believe that they were inaccurate. This suggests that participants were sensitive to the objective validity of the clues.
As a second means of assessing objective validity, data from a separate sample indicated that participants could easily distinguish useful from useless clues. In that study, each of the 28 participants was provided with the logic problem and one of the sets of clues used in this study (but without probabilities). Each participant estimated the chances that each of his or her clues would be useful for the solution of the problem. The useful clues were judged as more likely to be useful (M = 77.36%, SD = 19.26) than were useless clues (M = 8.56%, SD = 12.41), F(1, 26) = 371.91, p < .001, η2 = .82. This provides additional evidence that the participants could perceive differences in objective validity.
Level of Analysis
Each dependent variable was manipulated at the group level, and there was a high degree of group-level variance (intraclass correlation coefficients [ICC] for times clues were discussed, times useful clues were discussed, and times useless clues were discussed ranged from .55 to.59). Thus, hypotheses were examined at the group level (Kozlowski & Klein, 2000). 1
Tests of Hypotheses
In order to evaluate Hypothesis 1 (more discussion of clues with high rather than low levels of objective validity), each of the three measures of discussion (clues discussed, clues repeated, and times clues were discussed) was analyzed using a 2 (accurate vs. inaccurate clue probabilities) X 2 (high vs. low objective validity) ANOVA with repeated measures on the second factor. Although the pattern of means was consistent with predictions, objective validity did not reliably affect the number of clues that were mentioned, F(1, 40) = 2.58, ns, or the number of clues that were repeated, F(1, 40) = 1.04, ns. However, the total number of times clues were discussed was higher when objective validity was high (M = 19.62, SD = 5.96) rather than low (M = 18.05, SD = 7.41), F(1, 40) = 4.45, p = .04, η2 = .10. These findings provide limited support for Hypothesis 1; objectively valid clues were not more likely to be discussed or to be repeated at least once, but they were discussed more frequently.
Hypothesis 2 predicted there would be more discussion of clues with high rather than low social validity. To evaluate this hypothesis, each of the three measures of discussion was evaluated using a 2 (accurate vs. inaccurate clue probabilities) X 2 (high vs. low social validity) ANOVA with repeated measures on the second factor. A nonsignificant tendency was observed for more clues to be mentioned under high social validity (M = 8.62, SD = 0.76) than under low social validity (M = 8.14, SD = 1.41), F(1, 40) = 4.04, p = .051, η2 = .09. Social validity affected both the number of clues repeated and the number of times clues were discussed. Clues were repeated more frequently under high social validity (M = 6.33, SD = 1.86) than under low social validity (M = 5.24, SD = 2.52), F(1, 40) = 15.25, p < .001, η2 = .28. Likewise, the number of times clues were discussed was greater under high social validity (M = 20.14, SD = 5.92) than under low social validity (M = 17.52, SD = 7.02), F(1, 40) = 12.36, p = .001, η2 = .24. These findings support Hypothesis 2 and suggest that discussion was enhanced by high social validity.
Although these analyses provide evidence concerning main effects of objective validity and social validity, more focused comparisons are needed to provide evidence concerning the research question (the nature of the interaction between objective validity and social validity). Thus, we conducted a series of one-way between-groups ANOVAs and least significant difference (LSD) tests. For each of the three discussion variables, comparisons were made between the four conditions: low objective validity and low social validity, low objective validity and high social validity, high objective validity and low social validity, and high objective validity and high social validity. A significant condition effect was found for number of clues mentioned, F(3, 83) = 4.11, p = .009, η2 = .13; number of clues repeated, F(3, 83) = 5.22, p = .002, η2 = .16; and number of times clues were discussed, F(3, 83) = 4.71, p = .004, η2 = .15. Descriptive statistics for these analyses are presented in Table 2. 2 For each of these measures, LSD tests indicated that discussion was lower in the low objective validity–low social validity condition than in all other conditions. No other reliable differences between conditions were observed. This pattern of results indicates that information is discussed more frequently when it is supported by either social validity or by objective validity and that when one type of validity is high the other type of validity does not reliably enhance discussion.
Mean Clue Discussion in Objective Validity and Social Validity Conditions
Note: 95% CI = 95% confidence interval; N = 21 groups per cell. Means within the same column sharing a common subscript are not statistically different at p < .05 according to LSD (least significant difference) test.
Next, we examined the effects of clue probability accuracy. The clue probability by objective validity ANOVA (also used to test Hypothesis 1) indicated that clues were discussed more times, F(1, 40) = 5.31, p = .03, η2 = .12, when clue probabilities were inaccurate rather than accurate indicating that misalignment of objective and social validities prompted greater discussion. A significant interaction, F(1, 40) = 12.36, p < .01, η2 = .24, indicated that this effect varied across levels of objective validity. Follow-up tests indicated that the accuracy of clue probabilities did not affect the number of times useful clues were presented, t(40) = < 1, but did affect the number of times useless clues were discussed, t(40) = 3.39, p < .01. Useless clues were discussed more frequently when they had been socially validated. Hypothesis 3 predicted that accurate clue probabilities would lead to discussion that was more focused on useful clues than would inaccurate clue probabilities. To evaluate this hypothesis, the ratio of times useful clues were discussed relative to the total number of times clues were discussed (useful + useless clues) was computed and contrasted across the two levels of clue probability accuracy. For groups with accurate clue probabilities, 58% of the discussion involved useful clues whereas for groups with inaccurate clue probabilities 49% of the discussion involved useful clues, F(1, 40) = 12.50, p < .01, η2 = .24. Thus, consistent with Hypothesis 3, these findings suggest that groups tend to focus on useful information when they have been correctly informed about the utility of various pieces of information (when social validity is aligned with objective validity). In the accurate clue probability condition, discussion was more focused on useful clues because useless clues were less likely to be discussed.
Although groups in the accurate clue probability condition made slightly more correct deductions (M = 6.05, SD = 3.40) than groups in the incorrect condition (M = 5.57, SD = 2.42), this difference was not statistically reliable, F(1, 41) < 1. Therefore, contrary to Hypothesis 4, accuracy of clue probabilities did not affect group problem-solving performance.
Discussion
Hypothesis 1 predicted that clues high in objective validity would be discussed more frequently than clues with low objective validity. Objective validity did enhance the total number of times clues were discussed but did not affect the number of clues mentioned or repeated. However, these conclusions are qualified by more specific analyses indicating that objective validity led to greater discussion of useful clues when social validity was low but not when social validity was high. Although this pattern of results provide limited support for Hypothesis 1, the more important message is that objective validity increases discussion when social support is lacking but not when social support is strong. This pattern of findings suggest that group members can effectively distinguish between clues that are useful and those that are not but also that social support can affect discussion.
Although group members in the present study were able to effectively distinguish between useful and useless clues, this may be task specific. For example, the relationship between confidence and accuracy is moderately high on some tasks (e.g., general knowledge items and algebra problems) but low on other tasks (e.g., numerical estimation tasks and insight problems; Matlin, 1989; Perfect, Watson, & Wagstaff, 1993; Sniezek & Henry, 1989). In the current study, the ease of determining the actual utility of the clues probably reflects the eureka-type nature of the ski vacation problem. Although making correct deductions on this task is relatively difficult, once a correct deduction is made its validity can be demonstrated by ensuring that it is consistent with the other pieces of evidence. On eureka tasks, groups tend to recognize the correct answer once it has been suggested (Laughlin, 1980). Therefore, the ease of solution demonstrability inherent in this task may explain why participants could so readily judge the utility of the clues.
Initial analyses seemed to provide support for Hypothesis 2, that clues with high social validity were discussed more than clues with low social validity. When social validity was high (high clue probabilities), more clues were mentioned, more were repeated, and clues were discussed more times. However, more specific analyses indicated that social validity only enhanced discussion when objective validity was low; when objective validity was high, social validity did not affect any of the discussion measures. This pattern of findings provides qualified support for Hypothesis 2 but, more important, shows that social validity enhances discussion only under limited conditions.
Studies contrasting discussion of shared and unshared information suggest that shared information has greater social validity. Shared information is more likely to be mentioned and repeated (Larson et al., 1994) and is rated as more important, accurate, and relevant (Wittenbaum et al., 1999). Although the present study used only unshared information (each clue was distributed to only one group member), it also shows social validity effects. When objective validity was low, clues that were endorsed by the experimenter (given high probability ratings) were discussed more frequently. Therefore, effects of social validity observed in this study extend previous research to conditions involving only unshared information. Current findings also extend previous research by suggesting that social validation results in greater discussion of information only when the objective validity of that information is not apparent. The finding that level of social validity affects discussion of clues only when objective validity is low is consistent with social comparison theory. Social comparison theory suggests that social validation is critical primarily when objective validity is lacking (Festinger, 1954).
Information that is supported by either social validity or objective validity leads to increased discussion. There is no additive effect. Hence, the effects of social validity and objective validity are very similar; either is sufficient to lead to increased discussion of information. This suggests that either type of validity may motivate participation. Confidence that a piece of information is accurate or useful may represent a mechanism by which the two types of validity can motivate participation. Objective validity may lead to confidence because the item of information is perceived to be useful for task accomplishment (e.g., making deductions). Social validity may lead to confidence through the operation of a consensus implies correctness heuristic (Wittenbaum et al., 1999). Theory and research on metacognition (Hinsz, 1990; Nelson & Narens, 1992) and group problem solving (Hastie, Penrod, & Pennington, 1983; Littlepage et al., 1995) suggest that confidence enhances discussion. Because talkative members have greater influence (Bottger, 1984; Littlepage et al., 1995) and the extent of support for an alternative predicts its acceptance (Hoffman, 1979), the ideas in which members have confidence are likely to be accepted as group decisions. Generalizing from these previous studies, one can speculate that social and objective validity may affect member confidence in specific ideas and that confidence in an idea affects a member’s discussion in support of that idea. Although the sequence outlined above is plausible, we did not measure member confidence in the utility of each clue. Therefore, the present study does not provide data to test this causal sequence. Studies that measure member confidence in each item of information are needed to determine whether confidence plays the mediating role suggested here.
Results provide support for Hypotheses 3 concerning effects of accuracy of clue probabilities. Experimenter-provided assessments of the utility of various clues did affect the extent to which specific clues were prominent in the group discussion. Group discussion was more closely focused on truly useful clues in the accurate clue probability condition where instructions were designed to create accurate assessments of one’s knowledge. This greater focus resulted from less discussion of useless information in the accurate clue probability condition (when objective and social validities were aligned). Useful clues were discussed regardless of the accuracy of the clue probabilities. This reflects the previously identified pattern whereby a clue tends to be discussed if it has either objective validity or social validity.
Hypothesis 4 was not supported. Accurate clue probabilities and the resulting focus on useful information did not result in reliably higher levels of group performance. We offer two possible explanations for this failure to support Hypothesis 4. One possible explanation stems from the finding that useful information was discussed regardless of the accuracy of the clue probabilities. Although discussion was more focused on truly useful clues in the accurate clue probability condition, the manipulation of clue probability accuracy did not affect the discussion of useful clues; rather, it affected the discussion of useless clues. If performance is more dependent on discussion of useful clues than on the relative mix of useful and useless clues, this may account for the lack of performance differences. A second possible explanation is based on the nature of the task. It is possible that the linkage between discussion of useful information and group performance was obscured by the complexity of the information-processing demands of the logic problem. Each deduction required the joint consideration of multiple clues. Even though the manipulation of clue weighting accuracy led to a greater focus on useful information, this is not likely to lead to improved performance unless all the information needed for a deduction was discussed. Because of the conjunctive nature of these information-processing demands, the logic problem may not be an ideal task with which to demonstrate the linkage between discussion of useful information and group performance. Nevertheless, if the importance of alignment of objective and social validity is to be fully demonstrated, it is necessary for future research to do more than demonstrate that it affects the content of group discussion. It must also show that alignment of both types of validity affects problem-solving performance.
Another area where additional research is warranted involves the investigation of effects of objective validity and social validity across task types. The high level of solution demonstrability of the ski vacations task may have heightened the effect of objective validity. It is possible that objective validity may have a smaller impact on intellective tasks that are lower in task demonstrability (Laughlin, 1980).
Social motivation is another factor that may moderate the effects of objective and social validity on discussion. The current study utilized a cooperative problem-solving task, but theory and evidence suggests that with individualistic or competitive motives individuals may strategically choose which information to discuss. For example, Steinel and colleagues (2010) found that persons with individualistic or competitive motives would withhold important information. Therefore, current findings appear to be most relevant when group members have cooperative motives.
Likewise, future research should examine the role of confidence in the choice of information to discuss with the group. A possible model suggests that objective validity and social validity affect member assessment of, and confidence in, each piece of information and that confidence affects the choice to discuss the information. Information that is promoted in the group discussion is expected to have a large impact on the group decision. If group members are accurate in their assessments of information (i.e., confidence is aligned with objective validity), groups should discuss useful and relevant information and decision quality should be enhanced.
In sum, this study extends the understanding of factors that influence the content of group discussion by showing that the information that is discussed in a group is affected by both the objective validity of the information and its social validation. Members more frequently discussed information when it was actually useful or when it had been socially endorsed as useful. It appears that either type of validity can stimulate discussion but that the effects are not additive.
Footnotes
Research Presented at the 2011 Conference of the Interdisciplinary Network for Group Research
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
The authors received no financial support for the research, authorship, and/or publication of this article.