The core of the text is an evolutionary learning model, intended to “explain how states arrive at foreign policy decisions” (p. 1). In the model, agents (policy elites) are initially randomly assigned policy positions. Policy evolves as the weighted average of the recommendations of policy elites succeeds or fails and the weighting of recommendations is adjusted in response to performance. A later version of the model allows elites to shift their own recommendations in response to success or failure. Propositions derived from the model include the insight that extreme policy positions may improve policy (p. 101) and the finding that multiple advocacy can be counterproductive (p. 149). These propositions are in tension, however, if including extreme positions involves more policy advocates.

Evolutionary metaphors are often used to justify key “as if” assumptions of rational theory. Noting that selection generates “fitness,” the quality of being successful by appropriate evolutionary criteria, Farkas is able to show that even automatons can approximate optimizing behavior. This is an appealing finding, and one that emphasizes complementarity between the rational and the biological approaches. Unfortunately, Farkas presents the evolutionary model as an alternative to rationality.

Farkas appears to criticize a particular kind or rationality, ignoring opportunities to address his concerns, while adopting practices or attributes that parallel rationalist approaches or are equally subject to criticism. For example, Farkas challenges the plausibility of assumptions used in rational choice models but assumes that agents in his game all have heterogeneous preferences over goals. While defensible on tractability grounds, it is clearly not true (Stalin and Trotsky had different goals) and it alludes to the fundamentally problematic nature of arguing over assumptions. More to the point, policy elites in the evolutionary model are egoistic and purposive, seeking to enhance their influence over policy outcomes by adopting recommendations that they believe will be successful. The critical differences, then, between the evolutionary approach and an explicitly rational theory are that elites are initially randomly assigned policies rather than being able to choose policies, that elites update recommendations based solely on past performance, and that selection is exogenous.

There are more minor lapses in the text that perpetuate common errors. For example, in the introduction Farkas asks how people “who individually do not act as if they were rational” (p. 2) can behave as such collectively. Rationality, of course, is not what Farkas is addressing. Since rationality is judged by conformity to a postulated set of objectives (derived from a preference ordering and constraints), we can say that a particular formulation of rationality is wrong, but not typically that people are not rational. Prospect theory, for example, offers a different utility function, not an assertion that individuals do not have preferences or that they fail to seek preferred outcomes.

A more important set of issues has to do with evolution and politics. As the text points out, “The evolutionary model of state learning … depends on an analogy between beliefs and genes and policy and positions and phenotypes” (p. 62). At the core of the evolutionary model is the selection mechanism that adjusts the influence of policy elites. In biological systems, selection may be an exogenous process, but in politics seeking to manipulate the criteria by which one succeeds or fails is an important part of the game. Whether selection occurs in the consciousness of the leader (where contrary to claims in the text it would involve complex calculations), or in the institutions and norms of the state, an agent that can influence the way weights are assigned is evolutionarily advantaged. Since this requires cognitive skills (one must anticipate events and the actions of other agents), it is evolutionarily desirable in politics not to be an automaton. The assertion that the assumption of rationality is overly restrictive then appears somewhat more restrictive than allowing actors to think.

Another concern with, or potential expansion of, the model is that it assumes that objectives do not change. Policy evolves in the learning model to respond better to existing policy concerns. This is problematic if, for example, one is dealing with nuclear weapons. The ability to think and act strategically may be superior if learning from experience is dangerous or costly. Similarly, whether policy elites view the world through the insights of predecessors depends on whether these insights continue to be germane. If relevant processes remain substantially the same over time, then decision making based on inherited heuristics is sound. If, instead, environments change rapidly (as apparently occurs in modernity), then the value of inherited heuristics declines and individuals have an incentive to depend on other sources (peers, their intellect) for insight. This implies that the environments that are most ripe for the evolutionary perspective are those that do not change much or often.

In natural or sexual selection, as the rate or intensity of change increases, individuals or groups with traits that bias toward survival or desirability will increasingly predominate. However, in politics, the rapidity or intensity of change affects the salience of the evolutionary metaphor. As the rate or intensity of change increases, individuals or groups that depend on inherited knowledge will tend to do less well than those who reason. If allowed to exist, rational agents will predominate. After all, human cognitive abilities appear out of natural selection. Thus, the irony is that it would be at odds with the evolutionary metaphor to advocate the current evolutionary model of learning. In systems where knowledge is likely to be time-bound, adopting inherited heuristics biases against selection. The approach with greater fitness is to avoid heuristics and adopt rational choice.

The text emphasizes that the validity of a model depends on how much it explains. “Careful readers who are concerned with the misapplication of biological models would do well to ask how accurately a model captures the behavior it purports to represent and whether the predictions made by the model are supported by the data available to test it” (p. 67). The text offers many anecdotes, but there is no systematic assessment in the context of international relations. Further, there is substantial overlap between the evolutionary approach and existing rationalist explanations for international behavior. “The evolutionary model of learning produces outcomes that are identical—or nearly so—with those predicted by rational-choice models” (p. 76). I am willing to believe that the model has empirical leverage, but let me suggest at least one framework for a more general test.

Why not assess implications of the evolutionary approach for government institutions, rules, SOPs, etc.? There is little discussion of institutional and informal constraints on decision making in the text. This is striking because rules and other constraints are well suited to the propagation of knowledge. It is clear why incumbents seek to impose their will on future decision makers through institutional constraints. What is less immediately obvious is why subsequent decision makers would defer to such constraints. Rules carry knowledge of the relationship between policies and outcomes and so impact decision making. The evolutionary perspective suggests why and how rules develop and when they are maintained. Further, these are easily observable properties of states, so that tests of implications of the evolutionary approach should be manageable. Finally, this is an interesting aspect of politics that rationalist theories seek to explain, but not without some difficulty.

One learns from State Learning and International Change that there is much still to learn about learning in international politics. Farkas forces readers to examine conventional assumptions about foreign policy decision making. Rationality is not necessary to generate behavior that is consistent with the claims of rational models, nor need one adopt rational theory to engage in interesting, informative research. However, while contrasting rational and evolutionary approaches is of considerable value, of even greater value would be an exploration of the boundaries between, and synthesis of, rational and evolutionary models. Farkas offers a promising glimpse of the value of an exploration of learning in international politics and a hint at what the synthesis between these two perspectives can achieve. This alone should encourage readers to have a look.