Key theories of foreign policymaking include: the rational actor model, prospect theory, poliheuristic theory, cybernetic theory, bureaucratic politics, and organizational politics; and, at the group level, groupthink, polythink, and con-div. These theories are based on unique decision rules, including maximizing, satisficing, elimination by aspect, lexicographic, etc. A new, two-group model of foreign policy decision-making includes a decision design group and a decision approval group.
Leaders make decisions at the individual, group, and coalition levels (Hermann, 2001). Studies have found that the way they process information, and the decision rules they employ, affect their choice (Mintz & Geva, 1997). The following is a review of key theories that explain and predict foreign policy decision-making processes and choice. We also introduce a new model of decision-making consisting of two groups that explains decisions on national security and foreign policy.
Allison (1971, p. 30) defines rationality as “consistent, value-maximizing choice within specific constrains.” The rational decision-maker chooses the alternative that provides the consequence that is most preferred (Allison, 1971).
Cashman (1993, pp. 77–78) provides a set of steps in the rational model (see also Maoz, 1990, pp. 157–160; Rosenberg, 1995, p. 111):
Identify and rank goals
Analyze alternatives by considering the costs and benefits of each alternative and probabilities associated with success
Select an alternative that maximizes chance of selecting the best alternative
Monitor and evaluate
Rational actors are assumed to employ purposive action, to display consistent preferences, and to maximize utility (MacDonald, 2003, p. 552). Scholars distinguish between “thin” rationality, the strategic pursuit of stable and ordered preferences, and “thick” rationality, which assumes, in contrast, that actors have specific preferences. In politics, the preference is typically perpetuation in office (Ferejohn, 1991).
The expected utility model of rational decision-making assumes that the decision-maker “attempts to maximize expected utility in his choice between risky options by weighing the utilities of individual outcomes by their probabilities and chooses the option with the higher weighted sum” (Levy, 1992, pp. 172–173). In this model, the utility scores associated with each outcome are multiplied by the probability of that outcome manifesting. The cells are summed for each alternative, and the alternative with the highest sum is selected to maximize utility (Mintz & DeRouen, 2010, p. 61).
Bueno de Mesquita’s expected utility model ( 1981 ) posits that states will not go to war if the expected gains are smaller than expected losses. He tests the theory using data from the 19th and 20th centuries , and his basic hypotheses are strongly supported. Rational, compensatory behavior processes are at the heart of the expected utility model of decision-making. Most game theoretic models of foreign policy decision-making are likewise based on rational choice assumptions.
Herbert Simon (1985) claimed that cognitive models assume that “decision makers have limited information processing capabilities. Instead of objectively searching all information for the best outcome, decision makers will select an alternative that is acceptable” or “good enough.” The cybernetic theory model minimizes uncertainty through the use of information feedback loops. The basic argument is that individuals face processing and other cognitive constraints that limit their computational capabilities, their memory, and recall abilities. Because of these constraints, individuals develop decision procedures that enable them to deal more effectively and decisively with both their own cognitive limitations as well as with the demands imposed by the decision environment (March, 1986; Simon, 1985, 1957). Simon (1957) used the term “satisficing” to denote these decision procedures. “Satisficing” implies that decision-makers stop searching for information once they have found a satisfactory alternative; moreover, this alternative need not to be the optimal one, merely one that satisfies some minimum threshold (Monroe, 1991; Zey, 1992).
The cybernetic paradigm “precludes the need to calculate optimal procedures and alternatives on the basis of preferred outcomes by eliminating alternatives and ignoring the environment” (Mintz & DeRouen, 2010, p. 69). The decision-maker filters out extraneous information and therefore is only focused on a narrow range of incoming information. Maoz (1990, p. 161) observed that the decision-maker chooses one alternative at a time, calculates it, and, if it is eliminated, “chooses another alternative out of few options available until he finds the one that is most satisfying.”
Steinbruner (1974) provides several examples of cybernetic decision-making in the everyday world. His well-known example features a tennis player striking the ball without consciously making hundreds of mental calculations each time a shot is made. There is no need to make complicated calculations each time a shot is made because the player relies on information stored in feedback loops (Steinbruner, 1974).
This theory holds that people are risk-averse with respect to gains and risk-acceptant with respect to losses (Kahneman & Tversky, 1979). Prospect theory is posited as an alternative to expected utility theory and other rational models that are based on levels and assets. Prospect theory postulates that individuals evaluate outcomes as a function of deviations from a reference point. They also overweigh losses relative to comparable gains (Kahneman & Tversky, 1979; Levy, 1992, 1997). The theory consists of two phases: “In the Editing phase, the decision is presented, options are identified, the outcomes and their associated probabilities are also ascertained. Framing effects occur in this phase. Because prospect theory asserts that the way the decision information is presented can affect the choice. The second phase is the Evaluation phase, a choice is made based on the reference point and the value of utility function” (Levy, 1992; McDermott, 2004a).
Prospect theory with its emphasis on loss aversion has broad implications for foreign policy decision-making and international relations. For example, DeRouen (1995) utilized prospect theory to account for uses of military force when presidents are in political trouble. McDermott (1992) used prospect theory to describe the events surrounding President Jimmy Carter’s attempt to rescue Iranian hostages in 1980 . The rescue operation was risky, but Carter was in a domain of political loss leading up to the 1980 election.
The term poliheuristic can be broken down into “the roots poly (many) and heuristic (shortcuts), which alludes to the cognitive mechanisms used by decision-makers to simplify complex foreign policy decisions” (Mintz et al., 1997, p. 554). According to this theory, “policy makers employ a two stage decision calculus consisting of: (1) rejecting policies that are unacceptable to the policy maker on a critical dimension or dimensions, and (2) selecting an alternative from the subset of remaining alternatives while maximizing benefits and minimizing risks” (Mintz & DeRouen, 2010, p. 79). Mintz and Geva (1997, p. 84) assert that the “political dimension is important in foreign policy decisions not so much because politicians are driven by public support, but because they are averse to loss and would therefore reject alternatives that may hurt them politically.” Because it is based on cognitive, simplifying heuristics and other rules of thumb that leaders use in the first phase of the decision, poliheuristic theory can also explain complicated foreign policy decisions (Mintz, 2004).
Redd (2005, p. 129) examined President Bill Clinton’s decision-making in the Kosovo crisis and found that Clinton’s decision was influenced by “non-compensatory domestic political calculations and strong influence of his Secretary of State, Madeleine K. Albright.”
Decisions made by organizations are often based on standard operating procedures (SOPs). Based on the organizational politics model, decisions might be considered through the organizational lens. Often governmental decisions involve little uncertainty, are not crisis decisions, and are made on the basis of some a priori guideline or administrative rule (Dougherty & Pfaltzgraff, 1990, p. 487).
Braybrooke and Lindblom (1963) describe a standard operating procedure known as “incrementalism” (see also Wildavsky, 1964). Incrementalism leads to decisional inertia because the same alternatives are chosen again and again and are accepted over and over (Mandel, 1986, p. 259). Because there is no large deviation from past choices, there is little chance of catastrophic failure resulting from one decision. However, although involving low risk, if left unmonitored, incrementalism can get out of control. Because incremental decisions only make for small changes in the status quo, they rarely completely solve problems, but rather provide temporary solutions. Braybrooke and Lindblom (1963, p. 71) assert that this is the typical kind of problem solving encountered in everyday politics.
In making decisions, leaders use decision rules. The key rules are described here.
According to the traditional rational choice maximizing rule, policymakers will select the alternative that provides the highest net gain—in other words, the alternative with the most total benefits minus total costs. According to this rule, low scores on one dimension should not affect the ultimate choice (except, of course, in that it brings down the total score). However, the weight (or importance level) of each decision criterion is rarely equal to other criteria. Thus, the score for each dimension will be multiplied by the weight of the relevant dimension in order to create the final score for each alternative (Mintz & DeRouen, 2010, p. 34).
According to the lexicographic decision rule (Payne, Bettman, & Johnson, 1988), policymakers will select from among policy alternatives based solely on the dimension (criterion) most important to them. Low scores on other dimensions will not affect the ultimate decision, but a low score on the key dimension would automatically disqualify the alternative from consideration by that decision-maker (Mintz & DeRouen, 2010, p. 36)
According to the elimination by aspect decision rule, policymakers will eliminate alternatives sequentially based on dimensions they judge to be important in descending order. “Rather than just deciding based on which alternative has the highest score on one critical dimension, decision makers will first eliminate only those that have a relatively low score on their critical dimension” (Mintz & DeRouen, 2010, p. 36). Then, policymakers will proceed down the list of dimensions, eliminating alternatives that have low scores on secondary key dimensions.
According to the poliheuristic two-step decision rule, policymakers first eliminate alternative courses of action from consideration based on a non-compensatory rule and then select from among remaining alternatives based on a maximizing decision rule. In other words, options are first considered on one crucial, non-compromising dimension. If they are low on this dimension, they will be automatically discarded, even if, in all other respects, the choice scores very high. After this first stage, the different choices are evaluated based on a more traditional utility maximizing strategy (Mintz & DeRouen, 2010, p. 86).
Under the conjunctive decision rule, the decision-maker sets a minimum acceptable value for each dimension of the decision. To be accepted, an alternative has to be above the minimum acceptable value on all dimensions. An alternative is rejected if it fails to meet a minimum value, even if its overall sum is the highest. Conjunctive decision rule is different from the lexicographic or disjunctive rules, as it requires the alternative to be above a minimum value on all dimensions, not just the most important one (Mintz & DeRouen, 2010, p. 35).
According to the satisficing decision rule, a policymaker will select an alternative that is “good enough” but not necessarily “the best” (as assumed in all rational choice models). Bounded rationality theory and poliheuristic theory are some of the satisficing theories of decision-making.
The sections above have focused on individual decision-makers (i.e., state leaders) in making foreign policy decisions (see also Allison, 1971; Bueno de Mesquita, 1981). In the next section we discuss three models of group decision-making and offer a new, two-group model.
The group decision-making approach of foreign policy analysis challenges the monolithic view of nation states as unitary actors (Hudson & Vore, 1995). Group dynamics influences foreign policy decisions (Snyder, Bruck, & Sapin, 1954, p. 53).
Mintz and Wayne (2016) assert that groups can be placed on a continuum of decision-making dynamics (from “completely cohesive” [groupthink] to “completely fragmented” [polythink]). Polythink is essentially the opposite of groupthink. Whereas groupthink tends toward overwhelming conformity and unanimity, polythink is characterized by a large plurality of opinions, views, and perceptions among group members. This divergence of opinions, even dissent within the group often leads to a suboptimal decision or even deadlock. Polythink can be seen as a mode of thinking that results from membership in a highly disjointed group rather than a highly cohesive one.
Some of the symptoms of polythink are intra-group conflict and the existence of contradictory interests among group members (Mintz & Wayne, 2016). This may lead to a situation where it becomes virtually impossible for group members to reach a common interpretation of reality and common policy goals.
In the context of decision unit dynamics, it is important to distinguish polythink from the concept of “multiple advocacy” in which decision-makers “harness diversity of views and interests in the interest of rational policymaking” (Geroge, 1972, p. 751). Rather, the multiple advocacy model can actually be construed of as a type of structured, “pure” polythink process, in which the leader capitalizes on the already existing polythink dynamic to articulate divergent opinions of group members into a single, cohesive policy direction. Thus, whereas multiple advocacy is a type of polythink, it is important to note that most forms, structures, and variants of polythink are not multiple advocacy.
Both polythink and groupthink should be considered as “pure” types. In real-world decision-making situations, there is rarely a case of pure or extreme polythink or proupthink (Geroge, 1972). It is therefore more useful to think of these two models as extremes on a continuum in which “good” decision-making processes typically lie toward the middle, whereas defective decision-making processes lean closer to one of two extremes—the group conformity of groupthink or the group disunity of polythink (Geroge, 1972).
Under groupthink, the group making the decision “seeks consensus at the expense of exploring a variety of alternatives. Conformity to the group’s views is an overreaching concern for all members, so dissent is stifled and in some instances even punished. The group exhibits self-censorship and feelings of invulnerability and does not tolerate contrary viewpoints as it seeks to consolidate its unanimity (Janis, 1982).
According to Janis (1982), the symptoms of groupthink are: possessing an illusion of invulnerability; belief in the group’s inherent morality; the use of collective rationalizations; stereotyping the out-group; self-censorship; illusions of unanimity; directly pressuring dissenters; and the presence of self-appointed mind guards.
The symptoms of defective decision-making include: gross omissions in surveying objectives; gross omissions in the survey of alternatives; conducting a poor information search; processing information in a biased manner; failing to consider rejected alternatives; failing to examine the costs and risks of the preferred choice; and failing to work out detailed implementation, monitoring, and contingency plans (Schafer & Crichlow, 2010, pp. 63–64).
Polythink is a group dynamic whereby different members in a decision-making unit espouse a plurality of opinions and offer divergent policy prescriptions, even dissent, which can result in intra-group conflict and a fragmented, disjointed decision-making process. Members of the polythink decision-making unit, by virtue of their disparate worldviews, institutional affiliations, and decision-making styles, typically have deep disagreements over the same decision problem (Mintz & Wayne, 2016). Consequently, members of groups characterized by the polythink dynamic will often be unable to understand or appreciate the perspectives of other group members and thus will fail to benefit from the consideration of various viewpoints (Mintz, Mishal, & Morag, 2005). As such, polythink is no less problematic or prevalent than Groupthink.
The term polythink implies many ways of perceiving the same decision problem, goals, or solutions (Mintz et al., 2005). It can be contrasted with the homogeneous, uniform, monolithic viewpoint of groups characterized by groupthink. The sheer level of dissention in a polythink group may create a situation where it becomes virtually impossible for group members to reach a common interpretation of reality and common policy goals. As a result, it can often lead to suboptimal decisions (Mintz et al., 2005).
Polythink characterizes many decision units. How such units are composed shapes policy options and ultimately decisions (Hermann & Hermann, 1989). It is very likely that a decision unit that is handicapped by polythink will reach a different decision than a decision unit plagued by groupthink.
There are a number of important consequences of polythink, some of which (confusion, leaks, and framing) are counterintuitive.
Greater likelihood for intra-group conflict and leaks: As group members have different, sometimes even opposing views of the situation and of potential solutions, there is greater likelihood for group conflict due to polythink. this is in comparison to groupthink, where group members share more common views. Group conflict may impede not only short-term decisions but also long-term planning and implementation. Moreover, since group members do not hold uniform views of the situation under polythink, they are more likely to leak information in order to undermine positions that they oppose than in a groupthink situation.
Confusion and lack of communication: Exacerbated by the often unwieldy size of the federal national security and diplomatic infrastructures, polythink will increase confusion through willful lack of communication, mixed messages sent from different members of the decision-making unit, and inadvertent failures to effectively communicate between the large and diverse decision-making structure. Since group members express different or opposing views of the situation and of potential solutions, there is less likelihood for the group to speak in one voice under polythink. This is likely to confuse the decision-maker (along with the decision-maker’s constituency) and make him or her less clear on the optimal course of action.
More likelihood for framing effects: Polythink may cause advisors to rely on biased or selective information to effectively make their point, and make it convincing, in order to rise above the crowd of other advisors’ opinions. Under polythink, some members will use this selective information to frame offers, proposals, counterproposals, and even disagreements in different ways: some may give it a positive spin, others a negative spin. The likelihood of members of the group framing it in opposite directions when there is a group consensus, as in groupthink, is more limited.
Adoption of positions with the lowest common denominator: Polythink may create decision situations in which the lowest common denominator becomes the dominant product of the group. This is the case because each member of the group needs to make concessions in his or her normative worldview, as well as organizational and political agendas, in order to reach an accommodation with other members of the group.
Decision paralysis: In situations of conflict, the decision paralysis triggered by this dynamic can result in a complete failure to act to stem or prevent violent conflict or the adoption of suboptimal, satisficing policies that are often shortsighted and inhibit the long-term planning required in war and conflict. Due to the many divergent viewpoints within the decision-making unit, consensus and clarity can be incredibly difficult. This can cause leaders to freeze up, as they are both unsure of whether their choice is correct and unclear as to whether their choice could even be accepted by the rest of their decision-making unit and their public. Some of the consequences of polythink are similar to those of groupthink. This is the case not because the group is thinking alike or sharing the same views but “because the group is failing to carry out any significant collective thinking” (Mintz et al., 2005, p. 8).
Limited review of policy options: Paradoxically, polythink can lead to a limited review of policy options even though each advisor has distinct policy preferences. This is because there are too many options to fully consider given the limited time available to foreign policy decision-makers to devote to each specific issue. Thus, decision-makers will often quickly exclude some options from consideration in order to be presented with a more manageable choice set for more thorough consideration.
No room for reappraisal of previously rejected policy options: Compared with groupthink, under polythink the group is less likely to revise its policies if and when other policy options are brought back up for discussion, as any updating of policies is less likely to result in a consensus and may result in the time-consuming rehashing of previous disagreements between group members. Thus, as is the case with groupthink, polythink is likely to lead to defective, suboptimal decisions; however, the mechanism for these flawed decision-making processes is the group’s disunity and diversity rather than the group’s unity and conformity. Mintz et al. (2005) claim that constituencies, parties, bureaucracies, worldviews, group leaders, and expertise place significant constraints on the freedom of action of members of the decision-making unit and the “psychological presence” of these audiences act to curtail the cognitive processes and “information search” of each member. Collective considerations have to compete in the mind of each member of the team with other interest-based considerations (institutional, domestic-political, and personal) (Mintz et al., 2005).
There is a mid-range on the groupthink–polythink continuum that can be viewed as a balanced group dynamic in which neither groupthink nor polythink dominates. Mintz and Wayne (2016) call this area on the continuum the con-div group dynamic (i.e., convergence and divergence of group members’ viewpoints). In this situation, all group members do not share the same viewpoints, yet neither do they possess highly diverse or divisive opinions. They are more or less in equilibrium. In this scenario, the group is most likely to benefit from thorough yet productive decision-making processes that consider a multitude of options but ultimately reach some sort of consensus or agreement and execute well-formulated policies and actions. This is very different from either groupthink or polythink, which possess more extreme cohesion or dissent. Everything being equal, con-div should produce optimal decisions and group leader benefits due to the diverse input that is assessed in an overall atmosphere of consensus over key goals.
Per Mintz and Wayne (2016), the symptoms of con-div are:
Clearer policy direction than in polythink, with little to no confusion over direction: Since group members in con-div share the same goals and have no major disagreements over general policy, there will be less confusion than in polythink on action items, but probably more than in groupthink, where the group has consensus on its specific recommendations.
Fewer group information processing biases than in groupthink: Con-div is likely to lead to fewer group cognitive biases, such as “shooting from the hip,” “plunging in,” or even the “preference over preference bias” (Mintz & DeRouen, 2010), as there is some diversity in the group (though less than in polythink). For example, as group members in con-div can have different opinions on specific policy recommendations, the group is less likely to “plunge in” and make a “wrong” decision.
Less likelihood of ignoring critical information than in groupthink: Due to the plurality of opinions in the con-div group (although not quite as pronounced as in polythink), critical information that may affect the decision is less likely to be ignored than in groupthink.
Operating in one voice: Because group members share the same goals and vision for the organization, the group is likely to speak with one voice, especially if its leader sets a clear direction. Of course, rivalries among group members may still exist and lead to some off-message comments.
Too much harmony that may hinder real debate: Because group members in con-div share the same goals (although they may have different opinions with regard to specific policy directions), a leader’s dominant view may trigger an excess of harmony that will in turn lead to less debate over policy. In other words, the harmony associated with con-div (though less than in groupthink) may lead to recommendations that are not challenged enough by group members, especially if the views of the leader are perceived as too dominant or strong. Although in con-div there can be a debate where various opinions are presented, the ultimate decision will typically be made by the leader of the group, and group members will follow. In other words, the lack of dissent in con-div, though often constructive to the group’s deliberations, may not provide a sufficient challenge to the leader’s view. Importantly, some debate can be helpful for the decision unit, as it can illuminate logical errors in thinking or factual misjudgments.
Less likelihood of decision paralysis: Con-div is less likely to result in decision paralysis than polythink, especially if the group has a strong leader who sets a clear vision for the organization that motivates group members and other followers. However, deliberations may be prolonged en route to a decision if this dynamic exists in a group.
Greater likelihood of “good” decisions compared with decisions under groupthink or polythink: In con-div, plurality of opinions, viewpoints, and even debate are often channeled into more informed and better decisions, as group members share the same goals and vision. Thus, a plurality of opinions can provide useful input for leaders’ decisions, without the dissent and paralysis that polythink groups often experience. Moreover, unlike groupthink, in con-div a plurality of opinions serves to challenge mindguards and prevent the narrowing of vision associated with cohesive groups. Con-div may result in more balanced and superior advice to the group leader
Under con-div, group members who share the same general goals and objectives may be able to rise above smaller disagreements, if there are any. This may also lead to better decisions and more optimal outcomes.
However, con-div is not necessarily a prerequisite for optimal decisions, although this group dynamic is generally superior to groupthink and polythink because it facilitates more balanced recommendations for action. Often, con-div requires a leader with a clear vision and strong presence to bridge policy differences in pursuit of the goals of the organization.
Thus, working in a con-div environment does not necessarily guarantee a good outcome. There have been cases of government cabinets and executives working in harmony, sharing the same vision, the same key goals, and agendas, while debating policy alternatives, yet producing bad decisions.
Hudson (2014, p. 74) observed that most high-level policy decisions are made in small groups: “serious discussion of,say, a crisis situation, almost demands that the leader be able to sit around a table with a set of peers, and engage in candid and far-ranging debate of policy options.” And indeed, in a crisis situation, the critical decision unit typically consists of a leader and very few advisors. We call it the decision design group (DDG). After considering alternatives and choosing the preferred one, a larger group, consisting of cabinet members or other key executives, evaluates the situation and recommends an alternative (Sofrin, 2017). The leader has a major role in the decision-making process: he or she is heading a small group of one or two ministers and/or close advisers.
The DDG is responsible for identifying courses of action. Typically, a leader presents his or her preferred way of acting, and group members deliberate over it. Once the DDG agrees on the preferred way of action, it is presented to a larger team—a cabinet, or a government—for approval. This group is called the decision approval group (DAG). The dynamics in the DDG (i.e., groupthink, polythink) and the dynamics in the DAG and the overall dynamics between these groups shape the ultimate decision (Sofrin, 2017).
Figure 1. The two-group decision model.
When the “small group” (DDG) agrees on its preferred course of action, the leader and his or her advisors will make any effort to gain support from the “large group” (DAG). Members of the DDG will try to convince DAG members to support the DDG’s decision. They may also not expose all the information but only information that may support their preferred course of action (Sofrin, 2017).
There are several possibilities for inter-group interactions:
When the “small group” (DDG) is homogeneous and the dynamics is of groupthink and the discussion in the “large group” (DAG) is also of a groupthink style, the course of action designed in the “small group” will be adopted “as is” (e.g., the decisions of the Israeli government to enter the First and Second Lebanon Wars).
When the dynamics in the “small group” (DDG) is of groupthink, but the “large group” (DAG) is characterized by polythink, the ultimate decision will be similar to the one recommended by the small group. In these instances, the proposed course of action will be “reduced” or more limited than proposed (examples are the decision of the U.K. government to join the United States in the Second Gulf War in 2003 ).
When the dynamics in both groups will be polythink, the leader will make every effort to “force” his or her preferred method and the decision will reflect his or her choice (e.g., the debate in Israel to attack the missile launchers in west Iraq during the First Gulf War in 1991 ).
When the dynamics in the “small group” (DDG) is of polythink and the dynamics in the “large group” (DAG) is of groupthink, the leader will make every effort to gain support within the “large group” and that course of action will be adopted (e.g., the debate in U.K. government on the preferred way to solve the dispute with Argentina in the Falkland Islands).
Leaders make foreign policy decisions in a variety of contexts and environments. Some decisions are individualistic, whereas others are made in a group. This article surveyed models of foreign policy decision-making (both individual and group): the rational actor model, the cybernetic theory, prospect theory, poliheuristic theory, bureaucratic politics, and organizational politics, as well as groupthink, polythink, and con-div. It discussed the decision rules that underline these models. It also proposed a new, two-group model, which consists of a decision design group and a decision approval group.
The authors thank Eldad Tal-Shir for research assistance.