Beliefs, Identities, and Images

Beliefs are persistent ideas an actor holds describing the way the world works (causal beliefs), as well as what constitutes appropriate or legitimate conduct (normative beliefs; Jervis, 2006, pp. 642–643). Along with affect and behavior, they are a component of attitudes, or evaluations of a particular entity with some degree of favor or disfavor (Eagly & Chaiken, 1993, p. 1). They are not as broad as values, which act as general guides that are less context-specific than beliefs or attitudes (Rokeach, 1973). Belief systems, or the associated term “cognitive maps,” are cognitive structures that detail the causal and normative relations between different concepts that constitute beliefs, with those of primary importance to a decision making context being invoked more frequently than connected but peripheral beliefs (Axelrod, 1976; Larson, 1994, p. 20).

A prominent area of FPA is research on leaders’ beliefs that constitute operational codes (Leites, 1951; George, 1969). Stressing policymakers’ cognitive limitations, George hypothesized that leaders regularly drew from a set of five instrumental and five philosophical beliefs in order to simplify the decision making process, establish policy options, and choose among them. Instrumental beliefs specify ends-means relationships, or the causal mechanisms a policymaker thinks link decisions, actions, and outcomes. Alternatively, philosophical beliefs establish a decision maker’s views on the fundamental nature of politics, such as whether it is primarily characterized by conflict or cooperation (George, 1969, pp. 199–200). Both types of beliefs are stable, though they can change as leaders learn from experience (George, 1969, pp. 216–220; Renshon, 2008).

Operational code theory does not hold that the limited set of beliefs George detailed constitute the whole of policymakers’ ideas about the political realm. Indeed, among many possible examples, other scholars have detailed beliefs about whether threats typically arise from other countries’ foreign policies or domestic institutions (Saunders, 2011); whether leaders’ beliefs about domestic policy are associated with those concerning foreign policy (Rathbun, 2007); and about the effects of military experience on leaders’ beliefs and policies regarding the use of force (Horowitz & Stam, 2014). Furthermore, there is considerable variation across different leaders in terms of their instrumental and philosophical beliefs, and in that regard operational code theory recognizes a degree of actor-specificity. However, operational code research also assumes that the same ten instrumental and philosophical categories are central and applicable to numerous leaders’ belief systems (Walker, Schafer, & Young, 1998, p. 176), even though beliefs within these categories vary from person to person.

Computer-assisted content analysis of policymakers’ speeches and other communications enables researchers to analyze a large amount of data and maximize inter-coder reliability and content validity (Schafer, 2000; Walker, Schafer, & Young, 1998). Quantitative measurements of different beliefs can be used as variables in statistical models to estimate the effects of his or her beliefs on the propensity to use force, apply sanctions, and so forth (Schafer & Walker, 2006). This approach does not account for contextual specificity, or the possibility that the same words may have different meanings across time and space (Adcock & Collier, 2001, pp. 534–536). Even though Leites (1951) developed the operational code to analyze the beliefs of Soviet leaders, its developmental trajectory leaves its validity beyond the Western context in a state of greater uncertainty.

Closely related to beliefs are individuals’ images of those outside their own political communities and how these affect people’s conceptions of themselves and their social groups. Images are cognitive schemata, which like belief systems consist of a person’s interrelated knowledge about a concept or stimulus, and are structured hierarchically from general to more specific (Larson, 1994, pp. 19–20). For example, one may possess an image of dictators characterized by linked characteristics (authoritarian, aggressive, etc.) into which specific leaders, such as Adolf Hitler, fit. As with beliefs, images are simplified versions of their referents. Racial stereotypes are a well-known type of prejudicial image that can also affect foreign policy decisions (Hemmer & Katzenstein, 2002).

Images do not arise haphazardly. They are in part produced and sustained by an individual’s cultural milieu (Bar-Tal & Teichman, 2005; Jahoda, 1999), as well as being affected by individual-level psychological mechanisms. Images are often used to evaluate another political community vis-à-vis one’s own. Multiple possible images of outgroups are possible, from enemy or barbarian to ally. In turn, these different images have separate implications for whether another group’s capabilities and culture are superior or inferior to one’s own, whether they are threatening or benign, and so forth. These perceptions lead to preferences for some foreign policy strategies over others (Cottam, Mastors, Preston, & Dietz, 2016, pp. 57–58). Given the massive stakes of international conflict in the nuclear era and the potential consequences of misperception during crises, Cold War–era researchers intensely examined Soviet and U.S. leaders’ images of one another (Herrmann, 1985; 1986; O. R. Holsti, 1962).

If there are multiple possible images of another political community, as will typically be the case, people’s need for self-esteem may lead them to adopt a less flattering image of the outgroup to highlight their own perceived superiority (Herrmann, 1985, p. 672). Attribution theory indicates that political leaders will ascribe adversaries’ and allies’ undesirable actions to the latter groups’ inherent resoluteness or weakness, respectively. This has significant implications for the importance of reputation for resolve, a subset of leaders’ overall images of other states, in international politics (Mercer, 1995). In general, members of Western societies are biased toward attributing other people’s actions to their personal traits rather than situational pressures, a predisposition called the fundamental attribution error (Ross, 1977). This can give rise to or reinforce images of other individuals’ perceived groups, as well as reinforce dangerous patterns of international relations. For instance, if state leaders are more likely to attribute one another’s decisions on military spending as signs of an innately hostile character instead of systemic pressures to defend their countries, this will exacerbate the security dilemma (Glaser, 1997, p. 182).

State leaders trying to estimate whether an opponent has defensive or aggressive intentions must evaluate whether or not the adversary’s words and actions decidedly place it in one of these two categories (Jervis, 1970). Information that allows a decision maker to differentiate opponents along these lines may be a costly signal (Schelling, 1966, p. 150) in that certain images are too costly to project unless the sender is sincere—otherwise the cost of sending the signal would outweigh the gains sought. However, variations in individuals’ prior expectations and images of an adversary will lead different actors to draw dissimilar conclusions about an action’s meaning. Even if the leaders on one side make a statement they consider cheap talk in that they do not believe it will affect their opponent’s perception of them, leaders on the other side may significantly change their estimates of the adversary (Hall & Yarhi-Milo, 2012; Mercer, 2013; Yarhi-Milo, 2014).

According to constructivist International Relations (IR) theory, images can be co-constitutive (Wendt, 1998), meaning that the structure of one’s schema of the other is fundamental for understanding oneself, and vice versa. This perspective is most prominent within cognitive FPA on role theory. Roles are relational identities. For example, it makes no sense to see one’s state as leader of the free world if one has no image of what the followers in the free world are like. K. J. Holsti (1970) argued that countries had numerous national role conceptions that described these ideas of self and other. Along with scripts for behavior, roles can also define where one stands in the international hierarchy. The behavioral scripts and status-conscious dimensions of roles combine to inform foreign policy choices. For instance, if state leaders adopt a hostile, oppositional role conception while simultaneously believing their country has not risen to its proper station in international politics, they may be more apt to pursue nuclear weapons than a comparable country in which satisfied and nonoppositional role conceptions nevertheless predominate (Hymans, 2006).

Although individuals’ cultural and psychological traits predispose them to accept some images and roles and reject others, they still may have a multitude of possible images and roles to choose from when making policy and evaluating others. For instance, K. J. Holsti (1970, p. 260) identified 17 different national role conceptions, which other researchers elaborated on and extended (Walker, 1987; Le Prestre, 1997). In cases in which a multitude of images and identities are available, outside factors must be incorporated into the theoretical framework to predict which ones will rise to the fore. Institutional context and bureaucratic politics may be especially important in determining the officials who constitute the ultimate decision making unit (M. G. Hermann & C. F. Hermann, 1989) and thus which role conceptions and images they favor. Research from the field of American politics demonstrates that threat perceptions are likely to make the negative images of outgroups particularly salient (Marcus, Sullivan, Theiss-Morse, & Wood, 1995). If people fear for their own survival, they are also likely to reject novel roles, being drawn instead to those highlighting traditional behaviors, beliefs, and values (Greenberg, Solomon, & Pyszczynski, 1997).

Heuristics

Beliefs, images, and roles, which are themselves simplified representations of reality, may be the products of cognitive processes that are simpler than those involved in substantively rational decision making. These processes are forms of heuristic reasoning, or mental shortcuts. Heuristics privilege certain types of information over others to make evaluations or determine a course of action, even if the excluded information is relevant to the decision in question. Tversky and Kahneman (1974) documented numerous cognitive heuristics that people use when they are faced with uncertainty. Someone using the Representativeness Heuristic, for instance, attends to stereotypical characteristics of an entity in order to judge the probability that it belongs to a certain category. Say someone is asked what the probability is that a political leader is a dictator versus the probability that the same leader is a dictator and a communist. A person who is reliant on the representativeness heuristic whose stereotypical image of a dictator is a communist like Mao instead of a right-wing dictator like Pinochet would say the latter probability is higher. This would be an incorrect answer, but because most people are unfamiliar with the rules of conditional probabilities, it is simpler to use the heuristic to arrive at a conclusion.

Heuristics are a biased form of information processing, used because a full assessment of all relevant evidence in a decision making context is often beyond a person’s cognitive capacities. Heuristics are more likely to be used if there is very little time to make a decision, or multiple problems must be addressed simultaneously (Rieskamp & Hoffrage, 2008; Sweller, 1988). Nevertheless, heuristic reasoning can approximate substantive rationality, especially for experts, who can base decisions on patterns they recognize from past experiences. Even if a chess grandmaster cannot calculate all possible moves given the position of pieces on the board, she may recall numerous games in which the configuration of pieces resembled that of the current match and accordingly outline a nonexhaustive set of effective strategies. Likewise, though political experts may not do well at predicting international events or learning from past errors (Tetlock, 2005), they do appear better able than nonexperts to use heuristics to prioritize information when addressing specific problems. This serves to streamline the advisory process and allows officials to reach decisions in a more efficient manner (Hafner-Burton et al., 2013; M. G. Hermann, 1986; Saunders, 2016).

The poliheuristic model (Mintz & Geva, 1997) tries to account for a multiplicity of mental shortcuts decision makers use, and it has been applied to a wide array of foreign policy issues (Mintz, 2004, pp. 4–6). Each option for dealing with a foreign policy problem can be evaluated along multiple dimensions: military feasibility, economic utility, the likelihood of public support, and so on. The poliheuristic model theorizes a two-step decision process that begins with an individual eliminating certain options in a noncompensatory manner. This means that, rather than engaging in a substantively rational cost-benefit analysis of each option in which a weakness along one dimension can be compensated for by strengths associated with another (e.g., option A has high military costs but will likely receive popular support), weaknesses along a particularly salient dimension of a policy option might not be offset by strengths of the option’s other dimensions no matter how great they are.² As an example, Farnham (2000) argues that president Franklin Roosevelt judged foreign policies primarily by whether he could secure public support for them. If he did not believe support would be forthcoming, he would not consider the option or its other attendant dimensions any further. In fact, the most common cognitive shortcut is to evaluate whether or not an action will damage a policymaker’s own political fortunes (Mintz, 2004, p. 7). Once a policymaker has simplified the decision making process by limiting the number of acceptable choices, the remaining ones are evaluated in a more substantively rational manner.

Poliheuristic theory has been criticized for not explaining how political actors define foreign policy issues as problems that must be addressed;³ overemphasizing the salience of political considerations compared to other policy dimensions; and neglecting the institutional context in which decisions are made (Stern, 2004, pp. 110–111). The theory has been applied to leaders in the Middle East and South Asia, as well as in FPA’s typical emphasis on the decisions of U.S. politicians (Mintz, 2004, p. 6), which somewhat deflects the critique regarding the neglect of institutional context and avoids generalizations based solely on Western states. Another criticism of the poliheuristic theory is its assumption that rational actors are risk averse. In the second stage of the model, decision makers are presumed to rationally select the policy option that maximizes benefits and minimizes risks (Mintz, 2004, pp. 6–7). However, a tendency to take big gambles when making decisions is not necessarily irrational, but may simply be considered a preference like any other. Furthermore, research employing other cognitive models has shown that an individual’s preference for risk is contingent on the situation and the type of heuristics they employ, a subject to which the chapter now turns.

Risk, Loss, and Overconfidence

Risk is defined as the variance of outcome values in a probability distribution, such that policy options become riskier as the range of their possible outcomes increases. The common understanding of the term simply associates risk with the potential for bad outcomes. What is more, there will often be a difference between the real risks involved in a policy scenario and the risks the decision makers perceive, which will be a function not only of the variance and magnitude of potential outcomes but also the amount of confidence individuals have in the information on which they have based their decisions (Vertzberger, 1998). Despite findings that indicate political actors would normally be risk averse (Baumeister, Bratslavsky, Finkenauer, & Vohs, 2001), instances in which decision makers have been overconfident have also been thoroughly studied by FPA scholars. Heuristics can be a source of unjustified optimism.

Prospect theory details how individuals become more willing to take risks when in the domain of loss, and more cautious when in the domain of gains (Kahneman & Tversky, 1979b; McDermott, 1998; Mercer, 2005). This is true even when logically equivalent scenarios are framed differently to make losses more salient than gains, or vice versa. The best-known example of this is Kahneman and Tversky’s (1984) experiment in which participants were given a choice of two policies with which to address a hypothetical outbreak of disease. Some participants were told that if they chose Policy A, 400 out of 600 sick people would live; whereas if they chose Policy B there was a one-third chance that all 600 people would be saved and a two-thirds chance that no one would be saved. Other participants were told Policy A would kill 200 out of 600 people; the probabilities attached to the different outcomes possible under Policy B remained the same. Policies A and B have the same expected results, but Policy B is the riskier option because probability comes into play.⁴ Risky Policy B was preferred by a majority of people when Policy A was framed as producing a loss (killing 200 people instead of saving 400), but the certainty of Policy A was preferred when it was framed as a gain (saving 400 instead of killing 200).

Prospect theory has numerous implications for FPA scholars. Among these, leaders of declining powers may be expected to be more risk-acceptant in their decision making than would leaders of rising states. Deterrence should also be easier than forced compliance: the former requires challengers to forgo a gain, whereas the latter requires states to accept a loss. However, one difficulty in applying prospect theory to foreign policy is that whether a political actor is in a domain of gain or loss depends on its reference point (Levy, 1992, pp. 176–177). Consider the hypothetical leader of a state that lost a piece of territory years ago. If the leader’s reference point is the time period prior to this loss, prospect theory would expect him to be risk-acceptant and take aggressive action to retake the former possession. However, if the leader’s reference point is the present, it is more plausible that he would be risk-averse, as acquiring the territory would represent a gain over the status quo.⁵ Whereas psychologists can establish participants’ reference point during an experiment, the same cannot be said for FPA scholars studying historical cases of political decision making.

A related challenge for prospect theory is that the theory acknowledges that one’s reference point may lie in the present, past, or future, but does not take into account how that point’s temporal distance from the present matters (Krebs & Rapport, 2012). Studies have shown that people do not discount losses or gains at a constant rate. Instead, both—but especially gains—lose their subjective value more rapidly than standard discounted utility theory would expect as they become further removed from the present. Conversely, as losses and gains become quite temporally distant, they virtually cease declining in subjective value. This hyperbolic form of discounting, where values decline rapidly in the near-term and then plateau in the long-term, can lead to irrational reversals of policy preferences and make cooperation between states harder to achieve than rational-institutionalist models of IR would suggest (Streich & Levy, 2007).

When future losses and gains are probabilistic rather than certain, as is usually the case in international politics, individuals tend to be overconfident in their plans to secure goals in the more distant future in comparison to goals meant to accomplish near-term tasks (Johnson, 2004, pp. 41, 237). To effectively assess one’s future plans, it is necessary to evaluate both the probability of a plan’s success and the magnitude of the benefits sought or losses avoided. However, the further in the future one expects to achieve a goal and execute part or all of one’s plan, the harder it is to envision the contextual features that will affect the probability of its success. One might think this would lead to pessimism, but Kahneman and Tversky’s (1979a) study of the planning fallacy found the opposite: temporal distance usually led people to underestimate the costs of attaining their goals or completing a task, even if they had experienced difficulty with similar tasks in the past. According to construal level theory (Trope & Liberman, 2003), this is because people employ a heuristic when evaluating long-term plans. Instead of exerting cognitive energy assessing the plan’s feasibility, they simplify their assessments by concentrating on the desirability of their goal, which is much less dependent on future contextual details. They thus construe their actions in terms of why they are pursuing their high-level goals, rather than how they are going to formulate and execute the concrete steps of their plan. This abstract, high-level focus leads people to neglect or ignore how unexpected events and other factors that generate risk could affect their plan’s feasibility. This has implications for how state leaders think about addressing future threats or seizing distant opportunities (Krebs & Rapport, 2012), as well as how they assess combat versus post-conflict tasks in war (Rapport, 2015).

Learning and Foreign Policy

Learning may be the most complex of all cognitive activities. It involves the categorization of an event as a puzzle or problem; identification of the fundamental features of the problem based on existing beliefs and schemas; retrieval of information regarding past events; the organization of that information in such a way that it can be applied to the current situation; and observation of the effects of one’s own or another’s actions that are then used to update causal and/or normative beliefs. Learning is often done on the fly through trial-and-error as an individual executes strategies to solve a puzzle and then adjusts based on the feedback he receives, sometimes redefining the problem at hand (Levy, 1994; Stein, 1994; C. F. Hermann, 1990).

Foreign policy is an especially difficult area in which to learn and apply lessons. Most political problems are ill-structured in that they have no single solution (C. F. Hermann, 1990, p. 10; Levy, 1994, p. 292; Stein, 1994, pp. 172–173). Not only do policymakers typically lack a great deal of relevant information, but they may not even be aware of what pieces of information they are lacking (those infamous unknown unknowns). The very definition of learning is conceptually fraught in that scholars disagree about whether or not it necessarily involves an increase in the accuracy and precision of beliefs (Stein, 2002), or simply requires cognitive change in response to some external stimuli (Levy, 1994; Reiter, 1996).

A prominent area of the FPA literature studies how policymakers use historical analogies to learn and make decisions. This is a form of both causal and diagnostic learning (Levy, 1994, p. 285): past experiences inform present estimates of what the outcomes of different policy actions will be, and are used to identify key features of the current situation. Leaders typically overgeneralize from past events, obscuring differences between historical cases and the ones at hand (Khong, 1992; Neustadt & May, 1986; Vertzberger, 1986). Khong (1992, p. 7) models analogous reasoning as AX:BX::AY:BY, meaning that if event A had attribute X and event B also has X, a policymaker infers that since A had attribute Y then B also has Y. For example, Khong argues that U.S. officials who subscribed to the Dien Bien Phu analogy before the Vietnam War reasoned that because French military efforts against the Viet Minh had resulted in prolonged fighting and defeat, ongoing U.S. military efforts would also fail. Alternatively, policymakers using the Munich analogy (referring to the deal struck with Hitler in 1938) had learned that appeasing hostile forces leads to further aggression, with countries falling like dominoes as the adversary presses its military campaign forward.⁶ They were thus much more in favor of U.S. military action in Vietnam.

Overgeneralization is just one impediment to accurate and efficient learning. Instead of neutrally applying evidence to learn about existing theories, people use numerous heuristics and rationalizations to sort information (Tetlock, 1999). Their beliefs thus change more slowly than a substantively rational model would expect.⁷ Foremost among these heuristics is the confirmation bias: people search for and attend to evidence that supports their existing worldview rather than that which would challenge it (Nickerson, 1998). Another inhibitor is the hindsight bias (Fischhoff, 1975). Once people know the outcome of a case, such as whether a given foreign policy succeeded in attaining its goals, they retroactively interpret the evidence in such a way that the outcome seems virtually inevitable. The hindsight bias can lead those who make correct predictions overconfident in the theory underlying their estimates, since they neglect to consider how uncertain they were prior to the outcome being observed. Scholars of covert intelligence keenly note how political leaders use hindsight to blame surprise attacks or disappointing policy outcomes on intelligence failures even if there was no realistic way such events could have been predicted (Pillar, 2011, p. 228).

Scholars have also explored personality factors that promote or inhibit learning. In his study of U.S. foreign policy toward Latin America, Etheredge (1985) argued that policymakers failed to learn effectively because they typically had a hardball personality type. Harkening back to psychoanalytic theories, Etheredge contends that hardball politicians’ insecurities and their need to demonstrate social dominance led them to resist compromise and favor coercive actions designed to undermine leftist governments, regardless of the past outcomes of such policies. Twenty years later, Tetlock (2005) highlighted a different aspect of policy experts’ personalities, drawing from Isaiah Berlin’s metaphor of foxes and hedgehogs. Foxes are open to multiple explanations for why any given foreign policy outcome occurred, and are apt to treat each foreign policy problem as unique. Conversely, hedgehogs deduce likely foreign policy outcomes and events from an overarching theory; they know one big thing, which they use to understand the world. Tetlock’s multiyear study of foreign policy experts’ predictions about international events revealed that foxes were considerably better at the task than were hedgehogs, and were more effective at updating their beliefs after mistaken predictions. These findings support related studies on the association between reliance on general theory, cognitive complexity, and learning (Herrmann & Choi, 2007; Koopman, Snyder, & Jervis, 1990; Suedfeld & Tetlock, 1977).

FPA scholars have also emphasized how contextual factors affect the learning process. Drawing from research in neuroscience, Holmes (2013) argues that face-to-face meetings between diplomats and foreign leaders enable greater learning than less direct forms of communication. Directly observing another’s expressions during talks activates mirror neurons in the brain that simulate the mental processes of one’s conversation partner. In other words, face-to-face meetings better allow people to understand what their interlocutor is thinking. Communicative expressions that are automatic and involuntary act as indices, which unlike signals, are determinative indicators of another’s beliefs because they are not consciously manipulated (Jervis, 1970).

Aside from applications of the life sciences to FPA work, social features such as accountability have been argued to promote learning. If individuals believe they will face significant consequences depending on how well they analyze a problem, they are likely to expend more effort searching for relevant information and to be more critical when evaluating it (Tetlock, 1985). Stern (1997) asserts that foreign policy crises probably induce greater accountability, as a large portion of political leaders’ constituencies will be attentive to how policymakers perform when the stakes are high. At the same time, Stern (1997, p. 77) correctly notes that too much stress can inhibit learning, a point that has also been made elsewhere (Hammond, 2000; Lebovic, 1995, p. 845). Given this caveat, it is important to distinguish between learning during and after a crisis. During a crisis, decision makers’ responses to stress may partially cancel out the positive effect accountability has on learning. Afterward, leaders may have the time to more thoroughly reflect on what lessons they should take away but, depending on the crisis outcome, may feel less accountable to public scrutiny and unmotivated to evaluate their decisions.

Accountability may not be an unalloyed good. If leaders expect their decisions to be heavily scrutinized, they may think it is safer for them to agree with whatever consensus appears to emerge within a decision making group. Such a strategy will diffuse responsibility across numerous decision makers instead of focusing it on an individual whose position stood out from the rest, diminishing creativity and willingness to challenge the majority view in the process (Meade & Stasavage, 2008). Diffusion of responsibility can in turn lead decision making groups to adopt more extreme positions, since the blame for failure will not fall on any particular individual. This will especially be the case in democracies governed by coalitions, as shared responsibility will diffuse harm at the ballot box rather than concentrate it on a single party (Beasley & Kaarbo, 2014; Kaarbo, 2008, pp. 61–62).

Key Moderators of Cognition: Group Context and Emotion

These last points highlight the important effects group-level variables and emotions have on cognitive processes. Regarding the former, political decisions are rarely made by a single person. Two of the most influential works to promote the study of psychology in the context of political groups were Janis’s (1972) research on groupthink and Allison’s (1971) conceptualization of bureaucratic politics. Both authors focused on episodes of U.S. foreign policy crises. Janis argues that ideologically homogenous groups often prized cohesion and concurrence above all else when addressing a problem. This is especially be the case under stressful conditions, such as foreign policy crises, since the perception of unanimity enhances each decision makers’ confidence in their group. As a corollary, groupthink discourages dissent and leads to poor decision making: truncated information search and limited articulation of policy alternatives; reluctance to revisit earlier decisions or reconsider definitions of the problem; adoption of simplistic, hostile images of outgroups; and so forth. Alternatively, the fundamental cognitive premise of Allison’s model is that where you stand is where you sit. In other words, the bureaucratic organizations individuals belong to significantly shape their preferences and definitions of the national interest. According to Allison, the U.S. government was not unitary during Cuban missile crisis, but rather a collection of competing bureaucratic actors bargaining and compromising with one another in order to arrive at a strategy.

Though still widely cited, both Janis’s and Allison’s models have been thoroughly criticized. Although there has been some empirical support for groupthink’s hypothesized effects on foreign policy decisions (Herek, Janis, & Huth, 1987), most of the evidence has been mixed or contrary to the model’s expectations (Fuller & Aldag, 1997; Kramer, 1998; Schafer & Crichlow, 2010; Tetlock, Peterson, McGuire, Chang, & Feld, 1992). It is difficult to say how frequently groupthink occurs, but it likely constitutes a small subset of foreign policymaking behavior at the group level. Policymaking groups take on a variety of forms and carry out numerous different functions, and are further embedded in a wide array of institutional and political contexts that affect their performance. Janis focused on how groups process information and act as spaces that help policymakers cope with stress. Groups also function as command centers that monitor subordinates and use the levers of government to implement policies; sorters that delegate and prioritize among various tasks and problems; ideologues that reflect wider societal values; and smokescreens that legitimate the policymaking process by providing the appearance democratic and thorough problem-solving, even when the most crucial decisions are made without transparency in an ad hoc and unrepresentative fashion (’t Hart, Stern, & Sundelius, 1997).

Allison, meanwhile, was not wrong to assert that bureaucracies affect individual preferences and the way people mentally represent problems. Organizations and political groups influence what information individuals attend to by framing policy issues, systematically highlighting certain features of typical problems while obscuring others (Garrison, 2001; Maoz, 1990; Yarhi-Milo, 2014). By defining what counts as success, bureaucracies also affect what information their members monitor when determining whether or not a policy is effective, as well as the adjustments they implement when goals are not being met (Gartner, 1999; Steinbruner, 1974). Nevertheless, Allison’s model of bureaucratic politics has been criticized for its murky definitions of key concepts and for drawing conclusions that do not logically follow from its assumptions (Bendor & Hammond, 1992; Welch, 1992).

Despite their flaws, these early works provided ground on which other researchers could build. Consistent with groupthink, decision making groups were observed to engage in risky shifts, adopting policies that were riskier than any individual group members preferred by themselves (Vertzberger, 1984, p. 73). However, continuing research found that groups in fact exhibit polarization, meaning they adopt policies that are either riskier or more cautious than any individual member’s preferences (Kaarbo, 2008, p. 60). The extremity of outcomes in different directions indicates the structure and composition of a group significantly moderates policymaking. It might be tempting to assert that foreign policy groups within democracies would be prone to shifts in a more cautious direction, as the multiple institutional checks and constituencies to which leaders are responsible would constrain behavior. This hypothesis, however, neglects the role that leaders play in group decisions. M. G. Hermann and Kegley (1995) posit that moderate leaders in democracies will act pragmatically, showing sensitivity to institutional constraints and flexibility in response to incoming information. Conversely, ideologically driven leaders will aggressively try to circumvent constraints and be less apt to modify goals in the light of new evidence.⁸ Schafer and Crichlow (2010) provide qualitative and quantitative evidence showing that distrustful leaders are most likely to structure their decision making groups so they are closed off from outside sources of information, often leading to higher levels of conflict with other states.

Charles Hermann and colleagues (C. F. Hermann, Stein, Sundelius, & Walker, 2001) further specify how a leader’s orientation and the composition of the decision making group interact. If members primarily identify with the decision making group, as opposed to their bureaucratic agencies or other constituencies, and are led by an ideologue, they are likely to collectively exhibit many symptoms of groupthink. A pragmatic leader who encourages norms of dissent, however, can preclude such decision making pathologies even within a tightly knit group. If members do not primarily identify with the group and require unanimity to arrive at a decision, the outcome may be deadlock or compromise depending on whether the leader is able to act as a broker. If a mere plurality of group members is required to adopt a policy, then the degree to which the decision reflects the views of the group will depend in part on how much encouragement the leader gives to minority preferences. This framework appears to provide explanatory leverage beyond the WEIRD domain (Beasley, Kaarbo, Hermann, & Hermann, 2001).

A commonly prescribed method of avoiding groupthink and polarization is multiple advocacy within decision making groups (George, 1972). Multiple advocacy requires that decision making groups be composed of members with diverse preferences and areas of expertise so that no one perspective has a pronounced advantage over the others. Ideally, all members are equally resourced and familiar with the foreign policy problem at hand, and the leader encourages a balanced debate within the group (George & Stern, 2002, pp. 491–493). By pooling expertise and affording each member’s view equal representation, multiple advocacy is meant to approximate a substantively rational decision making process. There are trade-offs involved in such an approach, however. Multiple advocacy can take a greater-than-average amount of time to effectively execute compared to other processes. Another potential challenge is that if the group leader is not mindful of encouraging group members to offer distinct perspectives, individuals may disproportionately share common information—that which was already known by most of the group members prior to discussion. Not only is common information often more prevalent than individuals’ unique pieces of information, but people are perceived as more competent and credible when they put forward information that others already know (Stasser & Titus, 2003). During crises, when time is short, people try to find consensus quickly, which further encourages the discussion of common information (Karau & Kelly, 1992).

The discussion so far has included observations about the role emotional stress can play in cognition. A wealth of new findings on affect and cognition have cropped up since the 1990s, and the political science and IR disciplines have increasingly made emotion central to their agendas (Hall & Ross, 2015; McDermott, 2004; Mercer, 2010).⁹ Individuals’ emotional states do not always influence their cognitive processing, but they are most likely to do so when people are attending to problems that are complex and atypical (Forgas, 1995), which is characteristic of many foreign policy problems and crises.

One of the primary findings from research on emotion and cognition is that events and beliefs generally have an emotional tinge attached to them. Triggering an emotion attached to an event will make it more accessible in one’s memory (Herrmann & Choi, 2007, pp. 142–143), and if the emotional charge of the remembered event is very strong it will be more likely to dominate an individual’s decision making (Stein, 1994, p. 169). Not all emotions affect cognition in the same way, however. People in positive moods are prone to act like hedgehogs, that is, rely on general knowledge structures to make inferences. Analogies, heuristics, and basic belief systems are all examples of this type of cognitive processing. Conversely, negative emotions promote foxlike thinking, or the use of inductive, case-specific reasoning. This may be because people in good moods are more confident in relying on familiar general knowledge structures, while people in bad moods infer there is something not right in their environment and are thus more attentive to situation-specific information (Bless, 2000). Relatedly, positive moods facilitate the assimilation of incoming information into existing knowledge structures. Negative moods, on the other hand, lead people to see situations as more unique, motivating them to avoid judgment errors and retain as much incoming information from the situation as possible (Fiedler, 2000).

Although this pattern makes it appear that negative emotions are more conducive to learning, it is important to remember that those with foreign policy expertise may be better able to utilize general knowledge structures to address problems than novices. Treating each situation as unique may not be particularly efficient. Furthermore, negative emotions like anxiety can either help or inhibit cognition depending on how strongly they are felt. For instance, people generally are more attentive to feasibility considerations when they are evaluating plans that will be executed in the near future, which guards against overconfidence. However, if individuals are feeling a great deal of stress because a calamitous event like a war appears imminent and unavoidable, they may cope with their emotions by entering an implemental mindset that spurs them to execute plans confidently rather than consider the costs and risks of doing so (Johnson & Tierney, 2011).