Psychology and International Relations: Early Obstacles to Diffusion

Almost four decades ago, a small group of scholars in international relations, intrigued by the cognitive turn in psychology, explored how these new models could be applied to big questions in international security.Footnote ¹ Cognitive psychologists had identified multiple patterns of thinking that systematically deviated even from relaxed concepts of rational choice. Political scientists drew on these results to examine how loss aversion informed anomalies in the behavior of the parties in a deterrence relationship, how cognitive biases explained counterintuitive outcomes in bargaining, and how poor information processing and probability estimation led to unplanned wars and crisis escalation. They also explored the impact of group dynamics on collective choice that led to crisis escalation and an inability to anticipate unlikely but damaging consequences.Footnote ² Scholars in international relations continued to develop and deepen psychological explanations of international behavior but remained a minority within a corner of the larger field. Why did this early scholarship not diffuse more widely in research and journals? Both analytic and methodological factors seem to have been at play.Footnote ³

Rational choice, the predominant analytical model in the field that enabled formalization and the clarity of a deductive structure, provided a valuable normative baseline for evaluating choice. These are considerable advantages in a discipline that does not have a unified theory. Although cognitive psychology had established empirical support for a large number of systematic deviations from rational inference and choice, it was difficult to establish which bias would matter when and to specify its impact on choice. The failure to identify scope conditions was common to both rational choice and psychological theories, but it did not trouble theorists who developed deductive rather than empirical theories of choice.

A new wave of experimental research is establishing scope conditions for different biases and, in some cases, identifying thresholds when biases can be reversed. Research is also identifying heterogeneity across situations and individuals. The articles in this special issue address these questions to move beyond the sterile debate between rational choice and psychology. Instead they ask: under what conditions are decision makers likely to be rational and when are they likely to behave in ways that behavioral theories expect?

Several of the papers in this issue identify the conditions that enable the predictions of behavioral theories of choice. Renshon, Lee, and Tingley use citizens as a convenience sample for leaders and show that emotional arousal reduces the rate at which individuals select strategically optimal policies. Other contributors analyze the impact of identity on information processing and choice. Hermann finds that strong attachment to national identities leads people to interpret information through an emotional lens and individuals with distinct identities are therefore unlikely to converge as they are exposed to congruent information. Bayram demonstrates that actors with cosmopolitan social identities are more likely to accept the legitimacy of international law and their belief in legal obligation then attenuates their sensitivity to compliance costs. Strategic rationality in compliance, she argues, applies only to a particular set of actors. Closely related, Rathbun, Kertzer, and Paradis find that individuals with a pro-social orientation are less likely to behave in rational self-interested ways than are those with a pro-self orientation.Footnote ⁴

A number of papers explore the conditions that predispose individuals to behave as rationalists expect. Experience is one of these conditions. Saunders shows that inexperience biases information processing and inference.Footnote ⁵ Even when sufficient information exists in their strategic environment, less-experienced decision makers have more difficulty than those with greater experience in acquiring and in updating information. Rathbun, Kertzer, and Paradis also demonstrate in their experiments that experience attenuates the relationship between low epistemic motivation and biased choice. Kertzer shows that variation in individuals’ time and risk preferences affects how individuals update their beliefs about the desirability of conflict when they are confronted with new information about its associated costs.Footnote ⁶

I draw on one of these enabling conditions—the variation in risk preferences—and show how experimental research using a behavioral theory of risk has diffused outside the laboratory and deepened over the last three decades by identifying scope conditions. I use “prospect theory,” a theory developed through experimental research by psychologists and economists, as a case study for several reasons. First, prospect theory was central to the first wave of behavioral research and therefore provides a useful case study of how behavioral theories diffused in the analysis of international politics. Second, it deals explicitly with risk preferences and choice, a central concern in international relations theory, and speaks directly to rational choice. Third, it makes predictions based on situational and individual variables, engaging both structure and agent.

Finally, a close examination of the applications of prospect theory to puzzles in international security demonstrates the hard challenge of applying theory developed through experimentation outside the laboratory. Much of the current experimental work that is fueling the development of behavioral theories will face this challenge of external validity. Drawing on the example of prospect theory, I argue that moving experimental theories outside the laboratory is slow and demanding work. The story is nevertheless encouraging; prospect theory has diffused over the last three decades and now informs the way many important puzzles in international security are analyzed. Ongoing experimental research is also developing an increasingly robust set of scope conditions of the relevance of prospect theory outside the laboratory, a critical requirement for the more general advance of behavioral theories in the study of international relations.

Prospect Theory and International Security

“Prospect theory” is considered to be the best description of how people evaluate risk in experimental settings.Footnote ⁷ Its four central elements are well known: reference dependence, loss aversion, diminishing sensitivity, and probability weighting. Kahneman and Tversky demonstrated that people's choices among prospects are shaped by “framing effects,” the method for or sequence of how options are presented.Footnote ⁸ People frame their choices around a reference point and consider relative gains and losses from that reference point rather than estimate the net expected value of their assets. They are also much more sensitive to losses than they are to equivalent gains; the value function is concave in the region of gains but convex in the region of losses. The impact of loss aversion is amplified by systematic distortions in the weighting of probabilities. Assessment of risks becomes skewed when outcome probabilities are very high or very low. Depending on their reference points, individuals tend either to overweight or underweight these kinds of probabilities.Footnote ⁹

Moving experimental results to explain puzzles outside the laboratory is difficult. As scholars of international relations drew on psychological concepts that were generated through laboratory experiments, they faced four issues. First, economists and psychologists provide the frame in prospect theory in their experiments and then study the impact; outside the laboratory, the frame is not given and prospect theory provides only a partial account of framing through theories of “mental accounting.”Footnote ¹⁰ The absence of a strong theory of framing has slowed the diffusion of prospect theory in the field.Footnote ¹¹ Second, reference points must be defined ex ante, independently of outcomes, to properly test theories.Footnote ¹² Third, risk has emotional as well as probability dimensions. Emotion influences the choice of reference points and processes of probability estimation.Footnote ¹³ Risks that are particularly vivid or salient are systematically overestimated. Emotion also influences loss aversion directly; people feel the pain of loss more intensely than they feel the pleasure of equivalent gain and are more likely to take risks to recover a loss and less likely to see concessions as compensating for the injury when emotion inflates the loss and feelings of injury.Footnote ¹⁴ Finally, prospect theory, developed in the lab, paid little attention to the scope conditions that would shape its impact outside.

Scholars in international relations first used prospect theory to explain decision-making behavior during crises.Footnote ¹⁵ Using archival material, memoirs, and interviews, early research demonstrated the counterintuitive impact of loss aversion on deterrence strategy and examined its implication for deterrence theory.Footnote ¹⁶ For example, Farnham explained the changes in President Roosevelt's reaction to the Munich crisis and McDermott compared the decision making of Prime Minister Eden in Suez in 1956 and President Carter during the Iran hostage crisis.Footnote ¹⁷

That early research met strong resistance from rational choice theorists who discounted the impact of single case studies. But new work developed more robust tests of loss aversion's impact. In the analysis of deterrence outcomes, scholars in international relations used careful process tracing, first to identify the reference point of decision makers ex ante, then to establish the gains and losses that decision makers perceived themselves facing relative to that reference point, and then to assess whether decision makers’ behavior was as predicted.Footnote ¹⁸ A second promising approach has used the hedonic tone of the problem to suggest an obvious or natural frame.Footnote ¹⁹ Analysts then assess the material consequences of policy choices as gains when challengers are likely to judge that they are doing well and as losses when they are likely to do poorly. Both formulations find support for the proposition that loss aversion promotes challenges to deterrence that are not expected by rational choice theorists.Footnote ²⁰ In the analysis of deterrence, scholars now routinely use concepts of loss aversion to develop theory and loss aversion has now been built into policy.Footnote ²¹

A second important development was the effort to specify the scope conditions of the theory. Experimental research has demonstrated that the strength of framing effects varies across conditions.Footnote ²² Domain has an impact on the strength of loss aversion: the tendency toward risk seeking is greater in human life problems than in money problems.Footnote ²³ Prospect theory is more likely to be relevant, therefore, to decisions about war than it is to decisions about finance.Footnote ²⁴ When their survival or the survival of those they represent is not assured, decision makers are especially likely to be risk acceptant; their predisposition to risk is in part a function of the domain.Footnote ²⁵

Missing information, a common attribute of problems in international security, also enhances framing effects.Footnote ²⁶ Outcome probabilities also affect the strength of framing effects: certain/risky bimodal choices produce greater effects than choices between two risky outcomes.Footnote ²⁷ The condition of a certain/risky choice, combined with the endowment effect, helps to explain the conditions that favor deterrence over compellence success. Behavioral economists have identified the endowment effect that leads people to overvalue what they currently possess and therefore exaggerate the cost and losses of concessions.Footnote ²⁸ The endowment effect amplifies loss aversion in the context of compellence. Other things being equal, the endowment effect makes it more difficult for leaders to give up something that they already possess than to prevent them from taking something they do not currently have. The threat of loss of status has been used as well to explain leaders’ tendency to commit additional resources to a conflict once costs are sunk.Footnote ²⁹ Experiments demonstrate that when leaders are in the domain of loss, they tend to make the risk-acceptant choice of committing additional resources.Footnote ³⁰

The use of prospect theory to explain puzzles in international security demonstrates that the diffusion of psychological concepts developed in the laboratory has been slow and bumpy. The challenges in moving experimental results from highly structured problems in the lab to ill-structured problems in international security have been real. There is still no consensus on how to theorize reference points. Complex multidimensional choices, the need to construct options, and imperfect information make the transfer out of the lab difficult. Yet progress in testing alternative theories of framing has been significant, in micro-economics as well as in international security.Footnote ³¹ Scholars have developed alternative formulations of these issues and, working with a range of methodologies from large-N studies to archival case studies to an increasing emphasis on experiments both in the laboratory and the field, are using prospect theory to explain a large number of puzzles. Especially important has been the deepening knowledge about the scope conditions, both situational and individual, for the selection of reference points and the thresholds of framing effects.Footnote ³² Experiments have demonstrated that domain, missing information, outcome descriptors, outcome probabilities, and individual differences in emotion, identities, experience, and risk preferences can shift the selection of reference points and the impact of framing effects in systematic ways.Footnote ³³ Identification of these scope conditions, encompassing both situational and individual differences, and of threshold effects helps to establish when individuals are likely to behave as psychological theories predict and when they are more likely to approximate rational choosers.

Aggregation and Disaggregation

Moving experimental results outside the laboratory is not the only significant challenge behavioral theories face. As Powell's contribution to this issue demonstrates, processes of aggregation and disaggregation are generally considered an obstacle to the diffusion of psychological explanations of international behavior.Footnote ³⁴ In the analysis of international behavior, the units of analysis tend to be the state, international institutions, or the system as a whole, although increasingly individuals are the unit of analysis in the micro-analysis of the behavior of states with strong executive systems. Psychologists—and behavioral economists—are methodological individualists; in cognitive psychology and micro-economics, explanations generally remain at the level of the individual and the problem of inference goes away. But theoretical propositions drawn from individual-level analysis do not move easily to “higher-level” units such as states.

Psychological theories are not alone, however, in facing the challenge of aggregation. Rational choice theorists continue to assume, despite the now overwhelmingly contradictory empirical evidence, that the state is a unitary rational actor. Or, as the introduction suggests, they use a “boxes-in-boxes” approach where preferences—and politics—at one level connect to and constrain preferences at the next level.Footnote ³⁵ The advantages of this approach are clear: rationalists have to deal with only one source of bias—politics—while psychological approaches have to deal with two—the impact of politics at one level on another and the impact of psychological processes that may vary across individuals and situations. This adds orders of magnitude of complexity to the specification of psychological models of international behavior.Footnote ³⁶

However, the difficulty remains that rational choice models are also inherently individual when they specify how actors estimate and choose. States cannot think, process information, estimate probabilities, or calculate; only their leaders can. Yet rational choice theorists routinely assume a unitary decision maker and go up the ladder of analysis to attribute these properties to the state. In so doing, rational choice theorists assume that the impact of psychological processes is small and can be easily managed through standard “as if” assumptions. Yet the evidence increasingly suggests their impact is large, that it varies across individuals and conditions, and that under certain conditions, it is strong enough not only to affect but even to reverse preferences. There is then a considerable cost to assuming away the impact of psychological processes in theorizing choice.Footnote ³⁷

Second, scholars are increasingly paying explicit attention to when and how psychological processes are amplified or diminished in collective settings.Footnote ³⁸ It is important to ask how individual heterogeneity is aggregated into collective choice. In a group context, choice shifts are specific cases of the more general phenomenon of group polarization where group discussion leads to adoption of a more extreme position than the group average in the initially preferred direction.Footnote ³⁹ Drawing on social identity theory, theories of social contagion, strong pro-social orientations, social comparison, and persuasive arguments, researchers have specified how individual biases are modified or exaggerated by participation in different kinds of collective settings.Footnote ⁴⁰ Social comparison explanations of aggregation assume that individuals are motivated to present themselves as socially desirable and these processes of amplification become mutually reinforcing within the group; the average at the end of the group process has shifted toward a more extreme collective choice.Footnote ⁴¹ Theories of persuasive arguments emphasize the validity and novelty of the content of argument as two factors that shift individual choices in a group; novel arguments that are valid are especially persuasive.Footnote ⁴² Theories of social contagion expect that information, evaluation, and affect circulate among members of a group, building upon the fertile ground of the “group within the individual” to forge loyalty and consensus.Footnote ⁴³ Saunders explicitly links individual-level variation in bias with principal-agent models to explain why some biases are expressed or mitigated in certain cases rather than others.Footnote ⁴⁴ Rather than ignore these processes in the name of parsimony and assign individual-level attributes to states, as Powell suggests, social and political psychologists are modeling alternative formulations of the impact of aggregation processes on shifts in both the direction and magnitude of choice.

Conclusion: How Different Is this Time?

The diffusion and deepening of knowledge that I have shown in the use of prospect theory to explain important puzzles in international security suggests that, at the intersection of psychology, behavioral economics, and international politics, empirical theories can continue to develop a fruitful research agenda of puzzles using new methods and designs. Scholars of international relations now have the opportunity to design and execute experimental studies, both in the laboratory and in the field. Social psychologists have long worked with colleagues in international politics on experimental studies, but they were restricted to the lab, relied almost exclusively on students, and were very few in number. As a next best approximation, psychologists, political scientists, and historians worked together to analyze counterfactuals, but satisfactory controls were difficult to achieve.Footnote ⁴⁵ It is far easier today to do web-based experiments with a broad range of groups: in this issue, Herrmann embeds a series of experiments in a national survey in the United States to demonstrate that beliefs are not independent of preferences; Rho and Tomz use web-based experiments to analyze the impact of information on voter preferences about trade policy; and Renshon, Lee, and Tingley use experimental vignettes to examine voters’ responses to shifts in bargaining power globally.Footnote ⁴⁶ These experimental studies allow direct assessments of “interventions” as well as tighter control of confounding variables than other kinds of methodologies that scholars of international relations use.

The internal validity of the results from well-designed experiments tends to be high, but their external validity, as I have argued, is much more challenging. There is good reason to be careful about the inferences that are drawn from tightly structured experimental studies and applied in very different contexts, across situations and individuals. When scholars are assessing the impact of beliefs and information processes on voter preferences toward international issues, there is every reason to be confident in the external validity of the results of carefully designed experiments where participants represent the broader population. Better understanding of how publics think matters to the analysis of the constraints that decision makers face and to their assessment of audience costs. It is also central to the specification of a process of aggregation in a “box-to-box” approach.Footnote ⁴⁷

The challenge comes in drawing inferences from these kinds of experiments to the behavior of leaders. Just as psychologists have had to be careful not to claim too much from experiments with undergraduate students, scholars of international politics must be careful about the claims they make about leaders’ behavior based on experiments with students or undifferentiated publics. Saunders and Tetlock and Gardner have shown that experience can profoundly influence judgment.Footnote ⁴⁸ Field studies and interviews with experienced decision makers, as well as archival work, will continue to be important to validate generalizations from laboratory experiments. Scholars are now using large-N studies to test the impact of psychological concepts on leaders’ behavior. Multiple streams of evidence will help to refine theories about the impact of psychological concepts on international behavior. Finally, as I have argued, it is critically important to specify the scope conditions of theories that are designed and tested in the laboratory.

The opportunities for future research at the intersection of psychology and rational choice are considerable. In the last wave of research, rational choice was the default option against which deviance was measured. There was value to that kind of research to establish empirically the robustness of different patterns of thinking and choosing. That work is now largely accomplished and the next generation of research can examine how and when rational and psychological processes explain choice and behavior.Footnote ⁴⁹ Behavioral game theory is now drawing on psychology and behavioral economics to model the processes that deepen or dampen escalatory behavior. The analysis of a broad range of collective action problems can benefit from this kind of integrated approach.

Psychological approaches will also be useful in uncovering the micro-foundations of instrumentally rational strategic behavior. In an example of exactly this kind of research, Rathbun, Kertzer, and Paradis bridge the rational and psychological approaches to explore individual differences.Footnote ⁵⁰ Using both an experimental bargaining game and archival research, they find that the psychological micro-foundations of strategic rationality are a pro-self, social-value orientation and a high level of epistemic motivation. In a related study, Kertzer and Rathbun find that social preferences regarding reciprocity and fairness explain why some actors are better able than others to avoid bargaining failure.Footnote ⁵¹ Bayram finds that cosmopolitans comply with international law because of their sense of legal obligation, unlike those who are instrumentally rational and sensitive to the payoff structure of compliance.Footnote ⁵² All these studies, drawing on cognitive and social psychology, explore individual differences and engage the puzzle of who is instrumentally rational and when.

A third useful direction for research across the two approaches will be to establish thresholds for when psychological processes change the instrumentally rational outcome. Psychologists and behavioral economists have found, through experimental methods, the threshold when people shift from systematically overweighting to systematically underweighting the probability of outcomes. Using puzzles in international politics, experimental work can be done on other important cognitive biases to establish the thresholds when decisions shift.Footnote ⁵³ These kinds of experiments can begin to answer the question of how sensitive rational choice models are to empirically established processes of judgment and choice.

The research agenda for those using psychological approaches to explain anomalies in international behavior extends beyond the fruitful dialogue with rational choice. By moving beyond the assumption of universal self-interested maximizing behavior, behavioral theories suggest sources of cooperation other than those in the standard bargaining literature, including the impact of a broader set of social preferences, the emotional attachments to these preferences, and collective identities and moods on behavior. Scholars can examine when and how these preferences and identities mitigate the tragedy of the commons, a ubiquitous problem in a loosely governed international system where governance varies across sectors. Psychological approaches can also be helpful in explaining high levels of compliance even when reporting mechanisms are weak and enforcement mechanisms are negligible.

Generally, building more complex models of judgment and choice that are tested through combinations of experiments, large-N studies, and archival research can only improve the accuracy of the baselines we use to identify the important anomalies and puzzles in international politics.Footnote ⁵⁴ What is anomalous or puzzling can only be determined against a baseline, and the more empirically grounded the theories that specify the baselines, the better the choice of puzzles and the more productive the research agenda.