6Social Cognition and Information Processing

The main factor contributing to the increased emphasis on collective information processing is the importance currently ascribed to socially shared meaning (Cannon-Bowers et al., 1993; Echterhoff et al., 2009; Klimoski and Mohammed, 1994; Levine et al., 1993; Mohammed et al., 2010; Thompson and Fine, 1999; Tindale et al., 2001).60 In other words, the renewed attention given to “social cognition challenges the assumption that cognition is exclusively and individual act, distinguishable from external social processes that may influence it” (Thompson and Fine, 1999: 281). Social cognition is the result of social interaction, which allows groups of individuals to construct, share, and distribute information and knowledge. Therefore, social interactions “generate shared perceptions, behaviors, and products, including memories, norms, belief systems, and interpretations of shared events and activities” (Thompson and Fine, 1999: 281).

The ways social cognition influences group decision-making will be developed with greater depth and precision. In the meantime, it is worth noting that, considering the role that small groups play in foreign policy decision-making, a social psychological approach seems more appropriate than theories focused on the individual (Stein, 2005; Tetlock, 1998; ‘t Hart et al., 1997).

Particularly relevant is social psychology’s contribution in explaining how groups develop shared representations of a particular decision problem, that is, problem representation/definition of the situation. While traditional foreign policy analysis (FPA) approaches, founded on the concepts and methods of cognitive psychology, have endorsed many studies focusing on problem representations, these have traditionally involved individual decision makers or treated political entities (such as states) as rational unitary actors. Constructivist-inspired research, for its part, has drawn nearer to social psychology by subscribing to issues of shared identity. Constructivists, in accordance with Stein (2005: 303), “have expanded the repertoire of psychological explanations of international relations – that traditionally focused on beliefs, images, and judgment of leaders – to include the collective or shared beliefs that constitute a common identity, and processes of norm creation and norm observance.” Weldes’ (1996) exploration of the construction of US national interest adheres to this view. Rather than assuming the existence of some “real” and “objective” national interest like most realist theorists, Weldes’ constructivist stance privileges the act of social interpretation. For that reason, she argues, a constructivist approach “allows us to examine the intersubjectively constituted identities and interests of states and the intersubjective meanings out of which they are produced” (Weldes, 1996: 280).

The significance of the social outlook for geographic mental maps in FPA should not be underestimated. As previously mentioned, research on geographic mental maps has centered predominantly on individual decision makers. Some studies have explored the mental maps of particular policy-making groups (Mitter, 2008; Read, 2008), but any reference to the social dynamics involved in the construction of the shared geographic representations has been absent. In fact, though recognizing an individual’s cognitive limitations, Henrikson’s (1980a: 502 emphases added) initial conceptualization reinforced the unique quality of the mental map by asserting that “No single mind – which, strictly speaking, is the only unit of consciousness to which a mental map can attach – can encompass all that is humanly known of the global environment.”

The notion of collective mental maps is not, however, unusual. Behavioral geographers have long envisioned the existence of collective geographic representations. In their influential study on mental maps, Gould and White (1974) pointed to the need to consider how we could construct a single representative map assembled from many individual mental maps. While admitting that the notion was not yet clearly defined and conceptualized, this process of “homomorphic mapping” would allow for the construction of a single mental map revealing a group’s geographic preferences by combining the various individual mental maps. Nevertheless, the problem with this approach is analogous to that in much of the work conducted on the role of beliefs and ideas on group decision-making, that is, the collective quality of beliefs or ideas is obtained only through the process of aggregation. This is in line with the conventional argument, which states that only an individual can construct a problem representation (Axelrod, 1976b; Beasley, 1998). In fact, Axelrod (1976b: 239) is categorical in asserting that “collectivities do not think and have internal cognitive processes as individuals do.” However, since there is the possibility for a certain degree of correspondence in the ideas and belief systems of the different decision makers in a group, the various individuals’ problem representation affects decision-making through the process of aggregation (Beasley, 1998; Rosati, 1991). One such process consists of regarding a collectivity as “an artificial aggregate of its members, with ‘beliefs’ that are than simply aggregated from known beliefs of its members” (Axelrod, 1976b: 239). In fact, in the past, most research centered on measuring group cognition has applied and adapted individual measurement methodologies to groups (Cooke et al., 2007).

The theoretical assumptions underlying this perspective can be traced back to the early social psychological research on group decision-making, which solely emphasized the individual group members’ preferences as the only reasonable contribution for aggregation. In particular, Davis’ Social Decision Scheme (SDS) model set the theoretical foundations by determining that “small group interaction can be seen as a combinational process wherein preferences for decision alternatives across group membership must be combined in such a way as to allow the group to reach consensus on a single group choice” (Kameda et al., 2003: 461).

However, recent developments in several different disciplines focusing on group research have contributed to alter this perspective. As referred to above, cognition is increasingly being treated as a fundamentally social activity (Tindale et al., 2001). Levine et al. (1993) have challenged the conventional wisdom that cognition is exclusively an individual act and identified several ways in which social factors influence individuals’ cognitive contents and processes. At the most basic level, the mere physical presence of other people can affect an individual’s cognition (Hare, 1981). Numerous studies conducted throughout the years have demonstrated that either the active or passive presence of other individuals can generate episodes of social facilitation and social loafing. Equally, the presence of others can contribute to the sensation of crowding, which can lead to situations of insecurity, hesitation, and cognitive overload. Another way in which the mere presence of individuals can affect individual cognition is through group composition. The social organization of a group can affect individual cognition namely by shaping judgments, projections, and identities. As Moreland and Levine’s (1992) work on group problem-solving has suggested, the mere presence of the other individuals can facilitate or impede the process of problem identification.

Another way in which social features influence individuals is by providing social roles, positions, and identities. Research in social psychology has confirmed that “cognitive activity is strongly affected by how people construe the social situation in which they find themselves” (Levine et al., 1993: 591). Social roles for their part confine individual behavior by establishing a set of predictable expectations in a position or particular organization. For instance, institutions generally “distribute roles that mutually constrain actions and that increase the probability of a strong correspondence between expectations and outcomes; that is, once state actors adopt a particular role they usually limit their behaviour in a continuous and predictable manner” (Barnett, 1993: 272). Social positions, characterized as a socially accepted category of actor, play a comparable function. According to Levine et al. (1993: 592), “when a positional category is assigned to a person, the individual is expected to possess particular attributes and is responded to on the assumption that he or she has these attributes.” This is the case, for example, of designated experts in foreign policy groups (c. f., George, 1980). The role of the US National Security Advisor is illustrative of the need to conform to a certain socially established behavior, namely by maintaining an honest and/or impartial role in counseling the President and managing the information process (Burke, 2005).

In a similar fashion, social identities also determine individual cognition and behavior. By subscribing to a group’s identity, an individual assumes some actions as legitimate and intelligible and others as not. In this sense, the social group is deemed to work as a source of positive social identity for its members, specifically by comparing and distinguishing itself from other groups (Tajfel, 1982). Such is the case of ethnic interest groups or ethnic lobbies involved in influencing foreign policy decision-making (c. f., Ambrosio, 2002).61

Even when not in the presence of others, the mental representations of others can influence an individual’s cognition given that individuals may possess or expect to acquire knowledge about others behavior. Role taking – that is, when an individual looks at his own performance from the perspective of another – is one case in point. By assuming to place oneself in another’s position, we are inferring their reaction to the situation. This logic is quite analogous to that of social comparison. By comparing oneself to others, an individual uses his cognitive capacities to construct and distort information to achieve his objectives. Also, a person can be influenced by the presumed reaction of a reference individual. While physically absent, the idea of being judged by another leads to self-evaluation, which can “involve a reflected appraisal process whereby self is assessed according to how significant others would likely respond” (Baldwin et al., 1990: 436). Moreover, anticipated interactions with others also encourage anticipatory cognitive activity. The multiplicity of possible anticipatory interactions with others – for example, receive, transmit, or receive and transmit information – weighs heavily on the amount and type of cognitive activity that occurs.

A typical example of this type of phenomenon is the way in which a foreign policy decision maker acts due to his perceived accountability to his constituency (Tetlock, 1998). The pressure to perform in accordance with a specific constituency’s believed desire has led decision makers to behave contrary to their own inclinations on many occasions. For instance, despite various forewarnings of the need to assure all the required operational conditions in Afghanistan, George W. Bush pressed Cabinet members for action. In his view, the American public would not understand the delay in answering the terrorist attacks of 11 September (Woodward, 2005).

While the aforementioned phenomena relate to the way in which individual cognition is affected by diverse social factors, Levine et al. (1993: 599) also suggest the fusion of the social and the cognitive by envisioning cognition as collaboration, wherein “each person’s ability to function successfully depends upon coordinated interactions with others, and the cognitive ‘products’ that emerge from these interactions cannot be attributed to single individuals.” Rather than focusing on the individual or simply reducing group processes to individual cognition, the group is now judged as the primary unit of analysis (Brauner and Scholl, 2000).

This line of inquiry has undergone considerable development in relation to group information processing. Groups are currently viewed as information processors that are capable of encoding, storing, and processing sizable amounts of information (Brauner and Scholl, 2000; Hinsz et al., 1997; Kerr and Tindale, 2004; Tindale and Kameda, 2000). More specifically, group information processing entails “the degree to which information, ideas, or cognitive processes are shared, among the group members and how this sharing of information affects both individual- and group-level outcomes” (Hinsz et al., 1997: 43). Subsequently, as previously mentioned, “social sharedness” is the fundamental concept for understanding group information processing. At the most basic level, “the concepts ‘shared’ and/or ‘sharing’ are what make group information processing possible, and distinguish it from individual-level information processing” (Tindale and Kameda, 2000: 124). The quintessential belief underlying this theoretical perspective is that “things that are shared to a greater degree within groups will have greater influence on the relevant group outcomes/responses than those things shared to a lesser extent” (Tindale and Kameda, 2000: 124). In other words, by approaching information processing through the concept of social sharing, we can gain a superior understanding of what separates effective from ineffective decision-making groups because it is assumed that members of an effective decision-making group possess similar or compatible knowledge that they can use to guide their actions (Cannon-Bowers and Salas, 2001; Mohammed et al., 2010).62

Surely enumerating all the things that group members can share is an impracticable feat.63 Nor is it central to the objective of this study. Rather than providing a detailed review of all the current theoretical models, I will put forward a general account of some of the models, which most contribute to our understanding of how social sharedness influences information processing – that is, shared preferences, shared information, shared identity, shared metacognition, and shared task representations.64

The initial research on group decision-making focused essentially on the preferences of members. As mentioned above, the SDS model was the dominant framework for aggregating individual preferences. While SDS models have spawned a great amount of research and empirical results, the most consistent findings suggest that in group decision processes the majorities/pluralities generally triumph (Tindale et al., 2001). In particular, when groups cannot provide an “optimal” or “correct” alternative during the discussion, the “correct” alternative is defined by the group consensus, which is established by the dominant factions (Tindale and Kameda, 2000). SDS models have been criticized for being constrained to decision situations with discrete decision alternatives. Recently, several models have been developed that consider preference aggregation for continuous responses. For example, the Social Judgment Scheme (SJS) put forward by Davis looks to determine how groups reach consensus on a continuous response scale. The SJS model is based on the discrepancies of the position– that is, distance among preferences – along a response continuum among the group members (Kameda et al., 2003). Like the original SDS model, SJS and other recently developed models (c. f., Kameda et al., 2003; Kerr and Tindale, 2004; Tindale and Kameda, 2000) reveal the influence of social sharedness at the preference level by demonstrating that the members who share a particular preference can impose that preference on the group.

However, the models mentioned above fail to calculate how individual-level preferences will perform after a consensus in the group has been reached. Other studies have revealed that after a group choice has been made, group members generally subscribe to, or move closer to, the consensus position (Tindale et al., 2001).

The sharing of information among group members is also important to information processing and decision-making. Information sharing in groups should be understood by two distinct approaches – the common knowledge effect and the cognitive centrality of group members. In the first case, the work of Stasser and Titus (1985) opened the field for appreciating how shared information affects group decision-making. Contrary to prior theories, which postulated that unshared or unique information was determinant to decision-making, their research confirmed that “unshared information will tend to be omitted from discussion and, therefore, will have little effect on members’ preferences during group discussion” (Stasser and Titus, 1985: 1476).65

In a subsequent study, Stasser and Titus (1987) developed an information-sampling model that confirmed that the probability of a particular piece of information being recalled by the group during discussion is a function of the number of individuals possessing that same information. In this case, in group discussions, shared information is much more likely to be recalled than unshared information, with obvious consequences for the decision-making process. In other words, the research carried out by Stasser and Titus (1987: 92) found that in group discussions “information that is exchanged is seemingly biased toward confirming members’ prior preferences and does not give members a more adequate and representative picture of the decision alternatives.”

Effectively, some of the negative consequences of the common knowledge effect can be attenuated (Tindale et al., 2001). Research has revealed that unshared information becomes more accessible and widespread in group discussions over time. Also, the assignment of roles to group members also contributes to a greater pooling of unshared information (Tindale and Kameda, 2000; c. f., George, 1980).

Another way in which information sharing influences decision-making is through its distribution among group members – that is, cognitive centrality of group members. A member’s status or power in the group can be determined by the amount of information shared with the other members. Due to the importance attributed to shared knowledge in information-processing, it is argued that the members holding the greatest amount of pooled information will have a greater influence over the group decision-making process (Tindale and Kameda, 2000). Using a network framework similar to the SJS model, Kameda et al. (1997) measured the cognitive centrality of group members to determine their position in the sociocognitive network of the group. After conducting two different experiments, the study confirmed earlier findings that attested to the importance of shared knowledge in group information processing. More precisely, they found that “cognitively central members exerted more influences on group decisions, guiding consensus outcomes toward their preference to a larger extent than peripheral members” (Kameda et al., 1997: 306).

One of the reasons for the bias attributed to shared information in group decision-making may be the tendency for members to evaluate one another positively when mentioning shared information. Shared information can be validated socially, contrary to unshared information. In a series of trials, Wittenbaum et al. (1999) demonstrated that shared information is granted greater importance than unshared information because its exchange during discussion serves to validate members’ task knowledge. This process of “mutual enhancement” facilitates collective interaction by helping members relate to one another. Specifically, individuals who communicate shared information obtain more affirmative evaluations from other members.66 For their part, recipients of shared information feel better about their own knowledge when another group member reiterates that information. In sum “members who are positively reinforced (verbally or nonverbally) for communicating shared information may continue to do so because they enjoy the validation and encouragement from others” (Wittenbaum et al., 1999: 977).

In a more recent study, Wittenbaum and Bowman (2004) conducted a pair of experiments to determine if the need for social validation drives mutual enhancement and concluded that while social validation is important for information sharing other processes may also operate in conjunction with it, for example, group composition and social ties. In addition, their results suggest that the mutual enhancement effect is due less to the discussion of shared information, but rather to the discussion of unshared information. This is particularly relevant to the discussion of partially shared information. Wittenbaum and Bowman’s experiments revealed that if at least one member of the group can validate partially shared information, that information and its communicator can be evaluated in a more positive fashion than if no one can corroborate the information. The implications for the decision-making process are remarkable considering that, according to the experiments, “the most effective way to persuade a group to consider new information may be to make sure that at least one member knows and can validate it for the others” (Wittenbaum and Bowman, 2004: 182).

As mentioned above, social identity theory has also become a major focus in small group research. Its basic assumption is that individuals in a group identify themselves in a similar manner and share a definition of who they are, what attributes they have, and how they relate to and contrast with out-groups, namely by defining the features and boundaries of the group (Gergen et al., 2001; Hogg et al., 2004; Potsmes et al., 2005). The notion of social categorization is essential to understanding social identity, for people tend to divide the social world into in-groups and out-groups, which are cognitively represented as prototypes – that is, sets of attributes such as attitudes and behaviors that define one group and differentiate it from other groups (Hogg and Reid, 2006).

In a certain sense, a prototype may be understood as a cognitive representation of a group norm (Hogg et al., 2004; Hogg and Reid, 2006). Norms are embodied by group membership and define member behavior. In this sense, norms exhibit a prescriptive character and are therefore a source of social influence in groups (Hogg et al., 2004). By categorizing oneself as a member of a group, an individual implicitly accepts sharing a set of characteristics and behaviors that define the group (Bar-Tal, 1998). More significantly, a shared identity contributes to the definition of what is right and justifiable and what is wrong and illegitimate (Gergen et al., 2001).

An additional way in which social sharedness affects the information process is through the knowledge group members have of the degree of sharedness, that is, metacognition. Most research on social sharedness has centered on the degree to which group members share certain knowledge or information. However, some scholarly endeavors have also investigated members’ knowledge of what other members know and how the awareness of information distribution affects decision-making (Tindale and Kameda, 2000; van Ginkel and van Knippenberg, 2008). Particularly relevant in this field of inquiry is the concept of transactive memory developed by Wegner (1987). By adopting an individual-level cognitive template, Wegner argues that groups encode, store, and retrieve information in a manner quite similar to a single individual. According to Wegner, the possibility of group members acting as external storage locations produces a knowledge-holding system that surpasses the individual capacities of the sum of the individual group members. This allows for groups to remember much more than individuals. However, a transactive memory system requires that members know who has what information in order to access it. In other words, “a transactive system begins when individuals learn something about each others’ domains of expertise” (Wegner, 1987: 191).

Wegner’s experiments focused essentially on long-standing dyads, that is, married couples and couples in relationships. Since then, many scholars have explored how transactive memory systems work in organizational settings. Most studies have corroborated the theory. For example, Moreland (2006) and colleagues conducted a series of trials to determine the effect of individual and group training on group performance. All the experiments demonstrated that transactive memory can contribute to considerably improve a group’s work performance. However, Moreland has cautioned to the possible negative implications of member turnover on transactive memory systems. More accurately, turnover in group membership implies that the “system must be modified to reflect changes in the distribution of expertise within the group” (Moreland, 2006: 337).

This admonition is particularly important for foreign policy decision-making groups if we consider their fluid nature. Contrary to popular belief, foreign policy decision-making compromises various groups situated at the different administrative levels and small groups “often have shifting memberships caused by stratification into inner and outer circles, chronic Cabinet and staff turnover, and the ad hoc incorporation of experts into the decision unit or units” (Stern and Sundelius, 1997: 146; c. f., Hermann and Hermann, 1989).

Especially important for the argument of this study is the concept of shared task representations. Much of the research mentioned above is devoted to analyzing specific pieces or types of information and knowledge that group members can share. However, scholarly inquiries have also confirmed that group members can share a “conceptual system of ideas that allows them to realize when a proposed solution is correct within that system” (Kerr and Tindale, 2004: 638). These shared conceptual systems – that is, shared task representations – help researchers explain deviations from majority/plurality and other faction-size-related models. While majority/plurality models have demonstrated robust results in most experiments, numerous studies revealed asymmetric deviations from majority-type processes. Laughlin justified these variations by asserting that in group problem-solving tasks, small factions can influence larger factions by advocating the existence of “demonstrably correct solutions,” thus supporting “truth-wins” or “truth-supported-wins” decision schemes (Tindale et al., 2001).67 According to this perspective, demonstrability was achieved when a shared belief regarding the correct alternative existed among the members of the group (Tindale and Kameda, 2000).

Tindale et al. (1996) expanded on this work and corroborated that the existence of a shared task representation in a group allows for alternatives consistent with that representation to be more easily defended and consequently more prone to prevail as the groups’ ultimate collective choice. Tindale et al. (1996: 84) defined shared representation as “any task/situation relevant concept, norm, perspective or cognitive process that is shared by most or all of the group members.” By attributing task relevancy to the shared representation, it is inferred that it will “have some implication for the choice alternatives involved” (Tindale et al., 1996: 84). In other words, the shared task representations can influence the decision-making process as well as the final outcome.

It is generally assumed that the sharing of task representations generates beneficial effects on the group decision-making process (Klimoski and Mohammed, 1994; van Ginkel et al., 2009; van Ginkel and van Knippenberg, 2008). Particularly, by involving group members in a discussion of the group’s tasks and goals – that is, “reflexivity” – it is expected that individual members will become mindful of the eventual differences between their own and others’ representations. Once these differences are recognized and acknowledged, group members can try to reconcile them and develop more shared and task appropriate representations (van Ginkel et al., 2009). As a result, “if all the members of a group share a knowledge or belief system that lends credence to a particular alternative, that alternative becomes easier to defend in a group discussion” (Tindale et al., 1996: 86).

In a wider-ranging perspective, shared task representations can be understood in much the same way as Moscovici’s (1988, 2000, 2001) social representations. In fact, Tindale et al. (1996: 84) consider their concept of shared representations as a “subset of Moscovici’s social representations, delimited by the relevance of the representations to a specific group task.” For Moscovici (1988: 214) social representations are associated with “the contents of everyday thinking and the stock of ideas that gives coherence to our religious beliefs, political ideas and the connections we create as spontaneously as we breathe.” In this sense, the creation of social representations serves both to conventionalize objects and to prescribe human actions.