Rational choice theorists and political psychologists agree that
psychology explains only deviations from rationality. The primary task
of political scientists, according to Downs, is to study “rational
political behavior, not psychology or the psychology of political
behavior.”1
According to this view, rational
explanations should avoid psychology because psychology explains only
mistakes. As Satz and Ferejohn note, “mental entities need not
figure in the best rational-choice explanations of human
action.”
Harsanyi suggests that when
behavior appears to be irrational, then as a last resort analysts might
turn to “specific assumptions about the psychological mechanisms
underlying human behavior.”
Yet even in this case,
psychology can help only after the establishment of a rational
baseline.
Knowing irrational behavior depends on knowing rational behavior. For
example, psychology can explain wars that result from misperceptions,
but only after the development of a rational model.
Political psychologists
agree that psychology addresses irrational processes that result in
mistaken judgments. For example, Lebow has shown how emotion and
cognition subvert rational deterrence, McDermott has shown how
systematic cognitive biases encourage policymakers to make imprudent
gambles, and Jervis has shown how cognitive biases cause misperceptions
that undermine (or inadvertently support) policy.
In each case, psychology explains deviations from rationality.
The ubiquitous yet inaccurate belief in international relations
scholarship that cognitive biases and emotion cause only mistakes
distorts the field's understanding of the relationship between
rationality and psychology in three ways. First, the belief that
psychology explains mistakes (or deviations from rationality) has led
to the belief that rationality itself must be free of psychology.
Second, the belief that psychology explains mistakes has led
international relations theorists to believe that psychological
explanations need rational baselines. Finally, the belief that
psychology explains mistakes has led theorists to believe that
psychology cannot explain accurate judgments. Each belief depends on
the other. For example, rationality must be free of psychology if
psychology explains only deviations from rationality; analysts can only
know what constitutes a deviation from rationality after establishing a
rational baseline; they can only insist on establishing a rational
baseline to assess the role of psychology if that baseline is itself
free of psychology. This view of the relationship between rationality
and psychology is coherent and logical, but wrong. Although undermining
one of these three beliefs is sufficient to undermine the others, I
address each belief—or myth—in turn.
It
explains how one should reason in order to be rational. Analysts can
only know what is not rational—the domain of
psychology—after establishing what is rational. Demanding the
existence of a rational baseline before considering a psychological
explanation makes sense if that baseline is free of psychology and if
sthat baseline provides the best means to a given end. The second
section of this article critiques this rational baseline argument. I
use rational reputation models to illustrate the difficulty of
excluding psychology from psychological concepts and the danger of
relying on unacknowledged psychological assumptions.
and it should help political psychologists overcome their
disinterest in accurate judgments.
Rationality Depends on Psychology
The belief that psychology explains mistakes implies that rationality
is free of psychology: because rational models cannot be built on an
irrational (or psychological) foundation, each must be distinct from
the other. I argue that rational models are not distinct from, but
rather depend on, psychological assumptions. I use Elster's
definition of rationality as using the best means to achieve a given
end.9
I use accurate outcomes (or judgments) to
mean obtaining a given end and mistaken outcomes to mean not obtaining
a given end. I also distinguish “process” from
“outcome,” which makes it possible for nonrational (or
nonnormative) decision-making processes to yield accurate
judgments.
10 In other words, just as a psychological process
can result in either accurate or inaccurate judgments, a rational
process can do so as well. Not only can actors make mistakes by failing
to adhere to a rational process, but analysts can also make mistakes
(or cause policymakers to make mistakes) by presenting as rational a
process that actually undermines rationality.
11 To avoid the terminological confusion that may result from
the observation that psychological processes can be rational and that
rational (choice) processes can be irrational, I adhere to the
conventional view that models, processes, and baselines are rational to
the extent that they conform to a rational choice (or
“coherence”) standard of rationality. When necessary, I
substitute “rationalist” for “rational choice”
to distinguish it from “rational.”
Rational choice
theorists and political psychologists can agree on Elster's
definition of rationality while disagreeing over how to determine which
“means to a given end” are best. After establishing two
approaches to rationality—a coherence approach advanced by
rationalists and a correspondence approach advanced by
psychologists—I address why both rational choice theorists and
political scientists who use psychology view rationality as free of
psychology.
12 For exceptions, see Sears, Huddy, and Jervis 2003, 10–12; and
Bueno de Mesquita and McDermott 2004. I do
not address the relationship between constructivism, psychology, and
rationality. The constructivist focuses on the emergence and role of
ideational structures and typically relies on the “logic of
appropriateness” instead of the “logic of
consequence” (or instrumental rationality) to explain behavior.
For a discussion of both logics, as well as of strategic social
construction, see Finnemore and Sikkink 1998.
For a review of psychology and constructivism, see Goldgeier and Tetlock 2001.
Competing Approaches to Rationality
Rationalists and empiricists debated epistemology in the seventeenth,
eighteenth, and nineteenth centuries. Rationalists were committed to
deductive rigor, to logic, to mathematical proofs; empiricists were
committed to direct experience. As summarized by Robinson, a
psychologist and historian of psychological ideas: “If the
empiricist traditionally has been content to discover what is, the
rationalist considers what must be. The empiricist's evidence has
always been the data of experience; the rationalist's, the
necessary proofs following from axioms and propositions.”13
The empiricists won the debate in psychology and the
rationalists won the debate in economics. Economists have attempted to
exclude psychology and its empirical tradition from their discipline
and embraced the certainty of logic. This epistemological debate, which
I portray as one between rationalists and psychologists, led to
different approaches to rationality.
Rationalists view coherence as necessary to rational judgment. The
coherence view of rationality rests on well-established norms about how
one should make decisions, where a good decision is one that rests on
logic and adheres to the principles of statistical inference. An
incoherent judgment is one that, for example, allows the framing of
information to influence judgment, incorrectly assesses probability, or
violates the axioms of utility theory. Rational judgments must be free
of internal contradictions.14
A coherence model of
rationality implies a relationship between how one makes decisions and
the desirability (or utility) of the outcome. Rational choice theories
explain how one should reason, not how one actually reasons. The closer
one adheres to the model, the more rational one's judgment; the
more one deviates from the model, the more incoherent and thus
irrational one's judgment. A perfectly rational process can yield
an undesired outcome, such as in a prisoners' dilemma (PD), but
this will be the product of the situation, not a property of the
individual's judgmental process. Rational choice theory is
primarily normative, not positive, for it explains how one should make
decisions to reach a given outcome. Violation of these rules and
standards means a process is incoherent, and rationalists assume that
an incoherent process is likely to result in undesired outcomes. By
assuming that people want to be rational, rationalists bridge the
normative to the positive and make the approach testable.
15 A correspondence approach to rationality, which dominates in
psychology, is both positive and nonnormative. It is positive as it
concerns itself above all with actual decisions. It is nonnormative
because it is agnostic on how one should reason: it does not advance
rules or procedures but seeks to understand how people reason. Hammond,
a psychologist, suggests “correspondence theorists are interested
in the way the mind works in relation to the way the world works, while
coherence theorists are interested in the way the mind works in
relation to the way it ought to work.”16
Correspondence
theorists study accurate as well as mistaken judgments. They
investigate how people routinely make good decisions even though their
brains are not loaded with statistical software. The mind can solve
some problems better than others and typically relies on different
heuristics in particular situations to help with decisions.
17 Whereas rationalists emphasize how a rational process, ceteris
paribus, produces accurate judgments, psychologists emphasize how
nonnormative processes can result in accurate judgments. Thus the key
difference between rationalists and psychologists is not over what they
explain, but over how they explain it. Rationalists rely on deduction,
statistics, and probability theory, whereas psychologists rely on
induction and description of how the mind actually works. Neither
focuses primarily on mistakes, though each emphasizes a different way
of obtaining the same desired outcome.
Rationalists and Psychology
Behaviorists and neoclassical economists attempted to eliminate the
“mind” from causal explanations of human behavior. It was
not that mental phenomena caused mistakes, but that it was impossible
to know what role mental phenomena played. Behaviorists and economists
rejected the metaphysical in favor of the observable and material. They
aimed to overthrow “folk psychology” and replace it with
science. Although behaviorists did not address rationality directly,
their attempt to substitute observable stimuli for mental phenomena is
central to contemporary rational choice theory. Methodological reasons
also drove economists to eliminate mental phenomena from their
explanations. Their apparent success, coupled with a coherence
epistemology, led them to believe that psychology is useful only to
explain deviations from rationality. These two traditions explain
rational choice theorists' belief that psychology explains
mistakes.
Eliminating the mind from explanations of behavior required
overthrowing folk psychology. The folk psychology of human action, also
known as intuitive psychology, is the commonsense belief that
one's desires and beliefs explain one's actions.18
For example, why did Bob go
to the opera? Because he desires to hear an opera and believes he will
hear one at the opera house. Using Bob's beliefs and desires
inferred from his trip to the opera to explain that trip is
teleological: the behavior explains the purpose for that behavior. If
Bob desires to hear jazz, but goes to the opera, then analysts must
either view him as irrational or “fix” Bob's desires
and beliefs to make them consistent with his action. Because beliefs
and desires constitute action and cause action, folk psychology is
unfalsifiable. The theory has a few other problems. For example, it is
indeterminate (without numerous ceteris paribus conditions) and it
hinges on access to mental states that are difficult to know. These
(and other) problems make folk psychology problematic and unscientific.
Understandably, behaviorists, economists, and rational choice theorists
sought to replace folk psychology with a causal theory of human
action.
Behaviorism emerged in the early twentieth century as a response to
the problematic and unscientific nature of folk psychology.19
In contrast, behavioralism emerged in the
1950s from the efforts of political scientists to apply scientific
methods to a range of issues, including cognition and affect. See Merkl 1969; and Dahl
1969.
Behaviorists study behavior, not the mind. People
respond to incentives, and to understand incentives is to understand
behavior. In what became the behaviorist manifesto, Watson declared:
“Psychology as the behaviorist views it is a purely objective
experimental branch of natural science. Its theoretical goal is the
prediction and control of behavior. Introspection forms no essential
part of its methods, nor is the scientific value of its data dependent
on the readiness with which they lend themselves to interpretation in
terms of consciousness.”
20 Psychologists should
study behavior, not mental processes. As Skinner observed: “We do
not need to try to discover what personalities, states of mind,
feelings, traits of character, plans, purposes, intentions, or other
perquisites of autonomous man really are in order to get on with a
scientific analysis of behavior.”
21 Analysts do not need to study personality
because, as Skinner discovered, “Thinking is behaving. The
mistake is in allocating the behavior to the mind.”
22 Behaviorists accepted that private mental states exist,
but rejected their causal power.
23 Despite behaviorism's thirty-year dominance in psychology, it
failed to liberate psychology from the mind.24
One can teach a food-deprived pigeon a
lot—behaviorists recently trained pigeons to discriminate between
Monet and Picasso—and pigeon behavior is easy to predict in
controlled settings.
25 However, animals do
more than respond to incentives. Chimpanzees demonstrate what Gestalt
psychologists called “insight” by using sticks to create
tools, and boxes to create ladders, before obtaining a reward.
26 Even a rat's behavior in a maze exhibits
variations that require the introduction of the rat's mind.
27 Edward Tolman discovered what he called
“purposive behaviorism” and demonstrated that rats really
do think. Innis 1999.
Behaviorists
believed they had discovered a universal law of behavior:
“Pigeon, rat, monkey, which is which? It doesn't
matter.”
28 But it matters
enormously whether one is talking about rats or about chimpanzees, let
alone about humans. For example, the human need to justify past
behavior leads to thoughts, feelings, and choices inconsistent with
behaviorist expectations: the more unpleasant an initiation ritual to
join a group, the better liked the group.
29 Behaviorists thought they eliminated the mind from their
explanations, for they focused on what they imagined to be law-like
relationships between stimulus and effect. Animals respond to
incentives, such as corn pellets or cash rewards, and this allows
analysts to explain and predict behavior without reference to mental
processes. However, as Chomsky observed, analysts cannot identify a
stimulus without first identifying a response, in which case a stimulus
is not a property of the environment but of the individual's
beliefs and desires. This observation makes clear, as Chomsky notes,
“that the talk of ‘stimulus control’ simply disguises
a complete retreat to mentalistic psychology.”30
Rather than help analysts escape from psychology, stimulus-and-response
approaches depend on understanding an actor's mental state.
Even if researchers control the environment and provide only one
stimulus, how subjects respond to that stimulus depends not on its
physical attributes but on the subjective understanding (or construal)
of the stimulus.31
Researchers can reliably predict that a
food-deprived chicken will respond to a lever that gives it food, or
that a person will respond to a $10 bill on the ground by picking it
up, but prediction becomes unreliable in slightly more complex
settings. How students respond to very low (but passing) grades differs
dramatically: some students work harder, some blame the exam, and some
pump their fist and say “Yes!” Despite behaviorists'
attempts to rely only on behavior, they nonetheless relied on folk
psychology. Behaviorists eliminated from their explanations neither
desires (I want that corn pellet) nor beliefs (at the end of the third
maze on the left is food). Without knowing desires and beliefs, one
cannot know what “works” as an incentive.
As with their colleagues in psychology, economists coveted scientific
progress, which meant developing a causal theory of human behavior.
Although marginal utility theory began by expressing in mathematical
terms Bentham's view that the source of all human behavior is
avoiding pain and seeking pleasure, economists believed that measuring
hedonic states—my pleasure in consumption is two and a half times
your pleasure—was an illusion.32
Progress required abandoning the metaphysical and
focusing on the material and observable. Switching to a simple ranking
of preferences (ordinal utility) did not address the problem;
economists continued to assume individual cardinal utility, which
requires introspection to assign precise numbers to experienced
utility. Samuelson's introduction of revealed preference theory
appeared to solve the psychological problem.
33 Instead
of ranking preferences to explain behavior, Samuelson used behavior to
rank preferences. Like behaviorists, economists denied the causal power
of the mind without denying the existence of the mind.
Inferring preferences from behavior resulted in empirical and
theoretical problems. Empirically, economists have found so little
support for the axiom of revealed preference that it is
“empirically useless.”34
Theoretically, reliance on
revealed preferences meant that economists abandoned causal theory.
Revealed preferences explain action by inferring preferences from that
action, just as folk psychology explains action by inferring desires
and beliefs from that action.
35 Both are teleological
arguments. The problems of revealed preferences encouraged economists
to junk both behaviorism and any claim about the psychological
plausibility of their assumptions.
36
Whereas behaviorists believed that relying on mental phenomena was a
mistake, economists believed that mental phenomena caused mistakes.
Because psychology explains deviations from rationality, and given the
assumption that psychological biases are idiosyncratic rather than
systematic, psychology becomes unimportant with a large enough sample.
Markets are free of psychology even if individuals are not.
Friedman's famous “as if” assumption gave economists
license to stop worrying about psychology and learn to love aggregate
data.
37 When explaining market behavior, psychology
happily, perfectly, and conveniently washes out.
38 The
emergence of behavioral (or psychological) economics reinforces the
view that psychology explains mistakes. Behavioral economists attack
neoclassical approaches for ignoring psychology, and typically
reintroduce psychology to explain irrational behavior.
39 The market's
imperfections reflect individuals' psychological limitations.
The rational actor approach of microeconomics grounds rational choice
theory.40
Like behaviorists and neoclassical
economists, Satz and Ferejohn accept the existence of psychological
states but believe these states are unnecessary to explain human
action; rational choice theory has a “remote connection to
psychology.”
41 If Skinner is
correct that “thinking is behaving,” then as Riker argues,
analysts can infer preferences from behavior: “Choice reveals
preference because one can infer backwards from an outcome and a
consistent process to what the goals must have been to get to that
outcome.”
42 Intransitive preferences indicate either
irrationality or a change in taste. In other words, analysts infer
desires and beliefs from action. When those desires and beliefs are
inconsistent with action, then the actor is either irrational or the
actor has changed her desires. As Sen points out, the claim that
revealed preferences explain behavior without reference to
psychological assumptions is “pure rhetoric.”
43 Revealed preferences rely on the folk psychology
of human action, as does expected utility theory. Given an actor's
choices, analysts must either know beliefs about probability to
determine utilities, or know utilities to determine beliefs about
probabilities.
44 Just as
desires and beliefs explain and constitute action, analysts can measure
probability beliefs or utilities only by assuming the validity of
expected utility theory.
45
Expected utility theory depends on the assumption that desires and
beliefs explain actions. It depends on folk psychology.
Because rationalists reject the mind as causing behavior, the
actor's environment carries the explanatory burden. Hardin advises
that analysts should study the incentives that cause ethnic conflict,
not the psychology of the participants.46
Achen and Snidal note that rational deterrence theory never refers to
mental calculations because decision makers need not calculate:
“If they simply respond to incentives in certain natural ways,
their behavior will be describable by utility functions.”
47 Incentives explain the behavior of actors
engaged in ethnic cleansing, the behavior of decision makers in a
nuclear crisis, or even, as Satz and Ferejohn note, the behavior of
pigeons in a Skinner box: “Studies have shown that pigeons do
reasonably well in conforming to the axioms of rational-choice
theory.”
48 Whether rational choice explains
pigeon behavior, or vice versa, is beside the point.
49 Operant conditioning provides many of the same basic
predictions as microeconomics; see Alhadeff
1982.
Even if people are pigeons and incentives explain
behavior, analysts cannot know what counts as an incentive without
reference to the beliefs and desires of the animal. Adults, children,
and pigeons respond differently to the opportunity to drink scotch,
watch cartoons, or eat a corn pellet. Even if it were true that all
humans are pleasure seekers and pain avoiders, this simple hedonic
calculus restates the problem but does not solve it, unless one
believes that humans have identical views of pain and pleasure.
Behaviorists are still responding to Chomsky's observation that
one's beliefs and desires determine what one views as an
incentive.
50 Rational deterrence rests on the manipulation of incentives, which
are properties of individuals, not environments. Thus what is rational
for me to do depends on what inferences I believe you will make from my
actions, and this cannot be specified in the abstract. How you respond
to my threats or promises depends on your image of me and of
yourself.51
Treating beliefs as exogenous can be sensible but
does not solve the problem of inference: What will you infer from my
behavior? Relying on Bayesian updating works well when the new
information is objective and subject to one interpretation: I'll
recognize a poker chip as red even if I was expecting to see a green
chip and will update my belief about the probable distribution of red
and green chips in a bag. However, in most cases that interest students
of international politics, new information is ambiguous, subject to
multiple interpretations, and interpreted through one's beliefs,
expectations, and desires. As Jervis has emphasized, people do not
simply revise their preexisting beliefs based on new evidence, but they
use their preexisting beliefs to interpret that evidence.
52 Unless one is in the lab predicting the color of poker
chips in a bag, to make a rational choice one needs to understand
beliefs, desires, and their influence on the interpretation of
evidence.
The problems with folk psychology led behaviorists, neoclassical
economists, and rational choice theorists to rely on logic, deduction,
and observable behavior. Without denying the existence of the mind,
scholars in these traditions view the mind as exogenous to the study of
rationality. Yet incentives depend on beliefs and desires, revealed
preferences cannot be causal, and making preferences (or desires)
exogenous still requires knowledge of beliefs and action to assess
rationality. Analysts cannot hope to exclude the mind from rationality
if their explanations of rational behavior depend on the mind.
Eliminating the mind from one's explanations would be an enormous
accomplishment; inserting the mind by assumption is not. Of course, not
all approaches to rationality depend on folk psychology. Often
statistics or probability theory provides correct (or rational) answers
to problems. As I discuss in the next section, this approach to
rationality dominates much of the social psychology literature on
judgment and helps to explain why political psychologists, like
rationalists, believe that psychology explains mistakes.
Political Psychology and Mistakes
Whereas neoclassical economics still dominates economics departments,
the 1960s' cognitive revolution overthrew behaviorism in
psychology and reintroduced the mind as key to explaining not behavior,
but mistakes. Bruner introduced cognition to psychology with his
research on how motivation, such as expectations or desires, influence
perceptions.53
Ten
years later, Heider developed attribution theory by observing that
people predictably overattribute to disposition and underweigh
situation.
54 Heider's early observations led to what is
known as the “fundamental attribution error.”
55 After losing its mind to behaviorism, psychology
welcomed the mind back into its explanations, but these explanations
focused on deviations from rationality. Tversky and Kahneman follow in
this research tradition and have influenced how political psychologists
approach rationality.
Tversky and Kahneman's “biases and heuristics”
approach to judgment demonstrates that people systematically deviate
from rational standards.56
They
premised their research on reasoning backwards from mistakes (in judgment)
to errors (in process). The biases and heuristics approach compares
subjects' answers to reasoning problems to the normatively correct
answer that statistics or probability theory provides. People routinely
answer reasoning questions incorrectly. Researchers then identify the
difference between the normative answer and the actual answer as a reasoning
bias needing explanation. Shortcuts in the reasoning process—or
heuristics—explain the bias. Tversky and Kahneman found that the
statistically savvy avoid elementary errors, but “their intuitive
judgments are liable to similar fallacies in more intricate and less
transparent problems.”
57 Tversky
and Kahneman created well-defined experiments that allowed them to use
statistics and probability theory, not psychology, to create a rational
baseline; they then used psychology to explain why most people, most of
the time, do not provide the normatively correct answers. The biases
and heuristics program has cataloged such an impressive array of
pervasive judgmental errors and mistakes that Funder, a critic of the
approach, entitled his review of a social psychology text in this
tradition, “Everything You Know Is Wrong.”
58 The biases and heuristics research program encouraged political
psychologists to focus on procedural errors that led to mistaken
judgments. However, extending this approach to study decision making in
international politics results in two problems. The first problem is
identifying correctly the procedural error that led to the mistaken
judgment. Analysts can attribute bad decisions to the need for
cognitive consistency, improper assimilation of new data to old
beliefs, the desire to avoid value trade-offs, groupthink,
idiosyncratic schemas, motivated or emotional bias, reliance on
heuristics because of cognitive limitations, incorrect use of
analogies, the framing of information, feelings of shame and
humiliation, or a miserable childhood.59
For cognitive consistency and assimilation, see Jervis 1976; tradeoffs, see Lebow 1981; groupthink, see Janis 1982; idiosyncratic schemas, see Goldgeier 1994, and Larson
1985; motivated biases, see Lebow and Stein
1987; heuristics, see Mintz et al.
1997, and Simon 1982; analogies, see
Houghton 2001, and Khong
1992; frames, see McDermott 1998;
humiliation, see Steinberg 1996; and
childhood, see George and George 1964, and
Post 2003.
Discovering the correct
error is difficult in an uncontrolled field setting when so many errors
are possible.
The second problem is identifying the mistaken judgment. Was
President George W. Bush mistaken to believe that Iraqi President
Saddam Hussein posed an imminent threat to the United States in spring
2003? If Bush was correct, then analysts do not need to bother with
heuristics (unless he was right for the wrong reasons).60
Rosen argues that Kennedy adhered to a
nonnormative decision-making process during the Cuban Missile Crisis.
Rosen 2004.
If he was wrong, then
heuristics might be helpful (unless he knew he was wrong and simply
lied). As Jervis observed, “It is hard to know what a
person's perceptions were and even harder to know whether they
were correct.”
61 Experts can even disagree
over the correct answer to statistical problems. For example,
Gigerenzer contends that Bayesian approaches provide one answer to
Tversky and Kahneman's logic problems, but a statistician using a
“frequentist” approach would arrive at a different
answer.
62 A “frequentist”
rejects the Bayesian view of probability as a subjective measure of
belief, and instead uses the relative frequency of repeatable events to
explain probability. See Gigerenzer 1991; and
Hammond 1996.
Because the answers
are equally valid, what a Bayesian views as a judgmental mistake
needing explanation, a frequentist might view as an accurate judgment.
One cannot use statistics or probability theory to provide a rational
baseline for problems that lack unique logical or statistical
solutions.
When logic yields multiple answers to the same problem, analysts must
look elsewhere, such as to folk psychology, to assess rationality.
Assessing the rationality of a decision-making process demands that
three questions be addressed: What did the actor want, believe, and do?
After establishing an actor's desires, beliefs, and actions,
analysts can address how well the actor adhered to the rules of
inference.63
Did the actor consider alternative
explanations or hunt for disconfirming evidence? Analysts might also
ask how well grounded were the actor's beliefs, or how sensible
were the actor's desires. Whereas social psychologists can
establish rationality without reference to the mind, in most cases
political psychologists establish rationality only after determining an
actor's beliefs and desires. In other words, political
psychologists typically rely on folk psychology to understand
rationality.
Cognitive biases and emotion can be sources of mistakes, and
understanding how cognitive biases and emotion subvert rationality is
important. Yet the biases and heuristics research design presupposes
unique solutions to statistical problems. In the absence of unique
solutions, both rationalists and political psychologists must first
establish an actor's desires and beliefs before judging behavior
as rational. Only with the help of cognitive dissonance theory can one
acknowledge that rationality depends on mental phenomena while
simultaneously disparaging mental phenomena as a source of
irrationality. Because rationality depends on psychology, psychology
must do more than explain mistakes.
Psychology and Rational Baselines
The belief that psychology explains mistakes implies that
psychological explanations require rational baselines. Analysts must
know what is rational before they can know what is not rational.
Rational models provide standards that, ceteris paribus, explain how
one should reason and what constitutes a rational judgment. For
example, because of misconceptions about chance, people believe that a
coin toss is more likely to result in the pattern H-T-H-T-H-T than in
H-H-H-T-T-T.64
This belief is a mistake. A coin
does not have a memory and probability theory tells one what to expect.
If it turns out that a particular coin is more likely to land
H-T-H-T-H-T, then the problem is with the coin or the toss, but not
with the laws of probability. Similarly, rationalists argue that
analysts can understand the influence of psychology as a cause of war
only after establishing an “ideal” explanation of how
rational actors should behave.
65 When a rational process fails to yield an
accurate judgment, the problem is with the actor or the environment,
but not with the baseline itself. I argue, to the contrary, that
rational models of problems more complex than coin tosses will
sometimes be a source of mistakes. In some cases, analysts can turn to
psychology to prevent mistakes that rational baselines cause.
For example, rational reputation models advise decision makers that
reputations form predictably and systematically, and are worth
defending.66
Common sense, formal models, and logic tell
rationalists how reputations form and thus how rational actors should
behave. What is striking about the rational reputation literature is
that it should even exist. Dictionaries define a reputation as a
judgment about someone's character, yet character or disposition
is psychology's home turf. Riker understood that it makes
“no sense in terms of rational choice theory to attempt to
attribute consistent character traits to actors.”
67 Driving the cause of war and peace back into the
personality of decision makers smacks of “mentalism,” which
rationalists and behaviorists gave up long ago in the name of science.
Behaviorists would say that food, not an internal state of hunger,
causes a pigeon to peck a bar. Or, brittle glass breaks because it is
hit by a stone, not because of an inherent quality of brittleness.
68 Moore 1999,
64–65. This behaviorist explanation works only if one takes the
characteristics of glass for granted; a rock dropped on a sheet of
steel has a different effect. Jervis, personal communication.
Rationalists praised behaviorists for liberating psychology from the
mind and creating a true science.
69 In the
spirit of the behaviorists, rationalists improved the concept of
reputation by making it rational, which meant eliminating the
psychology. Once purged of all things mental, rationalists can use
reputation to address everything from market failure to collective
action and commitment problems.
70 Insisting that a rational baseline precede a
psychological explanation makes sense if that baseline first, is free
of psychology, and second, provides the best means to a desired end.
Rational reputation arguments fall short on both counts.
First, eliminating psychology from a psychological concept also
eliminates its explanatory power. For example, some rationalists define
reputation as a judgment about someone's situation.71
In this case, a reputation is a function not of mental properties but
of the costs and benefits of a situation. If a “mindless”
cost-benefit calculus explains behavior, then the situation, not a
reputation, explains behavior. “Situational reputation” is
an oxymoron. Reputation's explanatory power depends on explaining
either why an actor behaves similarly in different situations or why
different actors respond differently to the same situation. When
situational incentives govern—such as when people flee a burning
building—reputation is irrelevant. When situational incentives do
not govern, or when an actor behaves contrary to compelling situational
incentives—such as when someone stays inside a burning
building—then an actor's beliefs, desires, subjective
calculations of costs and benefits, or reputation may be relevant.
72 Draining psychology from a psychological concept
guts its explanatory power.
Second, an analyst's reliance on unacknowledged psychological
assumptions can cause mistakes. Just as behaviorists used terms such as
“drive” to replace ordinary terms such as
“desire,” rationalists use elaborate definitions of
reputation to evade psychology. For example, a reputation becomes a
belief about the probability that another is a particular
“type.” A “type” is a player's private
information about its expected utility for particular outcomes. Its
expected utility is its payoff for a particular outcome, and observers
infer probable payoffs from behavior.73
In order for a revealed reputation argument to explain or
predict behavior, payoffs must be a function, at least in part, of an
actor's disposition: beliefs, desires, character, and all things
internal, mental, and unobservable. If an actor's payoffs are a
function of the environment, then one is stuck with the oxymoron of
“situational reputation.” Substituting “type”
for disposition allows rationalists to rely on psychological
assumptions (such as the existence of a type that always keeps or
always breaks promises) while avoiding a psychological term.
74 This substitution
matters if the unacknowledged psychological assumption causes empirical
inaccuracies.
Psychologists once thought they could type personality the way one
types blood, but this view gave way to mounting evidence that a
person's behavior correlates modestly from situation to situation.
In the 1960s, psychologists suggested that subjective understandings of
situations explained responses to those situations.75
Actors behave characteristically in particular
situations as long as the actor perceives the situation as similar. It
is the actor, not the analyst, who decides which situations, or which
features of a situation, are similar to earlier situations.
76 How one
responds to different stimuli (or to different situations) depends on
how one construes those stimuli. Because rationalists deny their
reliance on psychology, they do not address their crucial assumption
that disposition is diagnostic of future behavior. Although assumptions
can be wrong but useful, rational reputation models fail repeated
empirical tests.
77 Inaccurate predictions should
encourage rationalists to revisit their psychological assumptions.
The two observations above—that psychological concepts drained
of psychology lose explanatory punch and that neglect of psychological
assumptions can result in mistakes—reinforce a final point: no
reason exists to privilege rationalist over psychological explanations.
A correspondence approach and a coherence approach to rationality can
yield competing means to the same desired end, and no principled reason
exists to prefer one over the other. For example, Mercer argues that
because reputations for resolve form rarely, rational actors should not
fight wars over reputation.78
Mercer uses attribution theory to
understand when people will use dispositional explanations to account for
someone's behavior. The only way to know whether reputations
should form, or to understand when reputations do form, is by
understanding how people think. In the case of reputation formation, or
more generally in problems of strategy (such as finance or nuclear
deterrence), knowing how people tend to reason and tend to behave is
crucial to making accurate judgments.
79 Predicting reputations is different than
predicting the distribution of heads or tails in a coin flip. Coins are
mindless; people (generally) are not. The point is not that rational
reputation arguments must be wrong, but that psychological explanations
might be right. Observing that Mercer's reputation argument might
be correct means recognizing that, in principle, psychological
explanations can correct mistakes that rational models cause.
Even on grounds of parsimony and generalizability, no reason exists
to privilege rationalist explanations over psychological ones.80
Rabin points out
that psychological approaches can be “simpler, more tractable,
and more useful (less post hoc) explanations than existing
models…. [T]here is no inherent negative correlation
between psychological realism on the one hand, and taste for
tractability, formalism, parsimony, and simplification on the other
hand.”
81 For example, loss aversion and
mental accounting may serve as a simpler and superior model to expected
utility theory, and a simple attribution theory may beat complex
rational reputation models.
82 Parsimony
and generalizability are attributes of theories, not of disciplines.
The belief that rationalist models necessarily provide the best means
to a given end, that they are free of psychology, and so should precede
psychological explanations, is wrong. Coherence models can be powerful
and revealing and helpful, but so too can correspondence models, and
the use of one does not depend on the existence of the other.
Emotion and Accurate Judgments
The belief that psychology explains mistakes implies that psychology
cannot explain accurate judgments. Scholars typically view cognitive
biases and emotion as undermining rationality, which is why rationality
is thought to be free of psychology and why psychology is thought to
require a rational baseline. By eliminating the mind, analysts can
determine what constitutes rational behavior. This view is mistaken:
rational models typically rely on psychological assumptions; rational
baselines need not precede psychological explanations; and, as I argue
in this section, nonnormative (or nonrationalist) decision-making
processes can provide the best means to a given end. I use emotion to
demonstrate the value of a correspondence approach to rationality.
Political scientists generally treat emotion as distinct from, and
ideally subordinate to, rationality.83
As a character in
Skinner's utopia put it, “We all know that emotions are
useless and bad for our peace of mind and our blood pressure.”
84 Political scientists typically emphasize the primacy of
cognition and allow emotion to play a role only when needed to explain
irrational choice. Scholars view emotion as a cost or benefit to be modeled,
a tool to be manipulated to send signals or impart credibility, or as an
inevitable but unfortunate aspect of human decision making.
85 Frank, an economist, argues that
people have emotional predispositions that commit them to behavior
that, while not in their immediate self-interest, may be in their
long-term interest: emotions are “merely incentives to behave in
a particular way.”
86 Frank's observation
that emotion can be adaptive by providing people with cues to help
navigate their world is insightful. However, recent work in
neuroscience suggests that emotion is not merely a tool of rationality
but instead is necessary to rationality. This perspective encourages a
new look at international relations concepts. I discuss how emotion
makes “trust,” “identity,” and solutions to
collective action problems possible.
Developments in neuroscience—such as brain imaging techniques
that allow scientists to examine emotion in the brain—have
contributed to the return of emotion to psychology and the introduction
of neuroscience to economics.87
See Glimcher 2003; and Wall Street Journal, 15
November 2002, B1.
After decades of research and exposure to
patients such as “Elliot,” the neuroscientist Damasio
rejected the traditional view that separate neural mechanisms operate
for emotion and for reason.
88 After Elliot had a brain
tumor removed, he underwent a radical change in personality and was no
longer able to hold a job, but could not collect disability benefits
because health professionals declared him to be of sound mind and body.
Damasio was asked to determine what, if anything, was wrong with
Elliot. Damasio found that Elliot was a smart and coherent young man
who retained memory, language, and other cognitive skills, but lost his
ability to “feel” or to have emotion. Elliot was “the
coolest, least emotional, intelligent human being one might imagine,
and yet his practical reason was so impaired that it produced …
a succession of mistakes, a perpetual violation of what would be
considered socially appropriate and personally advantageous.”
89 Damasio had other
emotionally flat patients with similar problems. For example, one
patient, when asked to decide between two dates for his next
appointment, was stuck in a “tiresome cost-benefit analysis, an
endless outlining and fruitless comparison of options and possible
consequences.”
90 The patient spun out a mental decision tree for nearly
thirty minutes until he was told that the second of two dates was
better.
Damasio's research on the brain is not a parlor trick. People
without emotion may know they should be ethical, and may know they
should be influenced by norms, and may know that they should not make
disastrous financial decisions, but this knowledge is abstract and
inert and does not weigh on their decisions. They do not care about
themselves or about others, and they neither try to avoid making
mistakes nor are they capable of learning from their mistakes. They
“know” but cannot “feel” and, consequently,
they make mistakes.91
Emotion is necessary to rationality and intrinsic to choice. Emotion
precedes choice (by ranking one's preferences), emotion influences
choice (because it directs one's attention and is the source of
action), and emotion follows choice (which determines how one feels
about one's choice and influences one's preferences). Yet
theories of rational decision making imagine a process free of emotion,
reintroducing it only to explain mistakes. Emotion can, of course, lead
to mistakes, but so can cognition, and no one suggests that cognition
is only a source of mistakes.
If rationality depends on emotion, then correspondence models of
rationality may have an advantage over coherence models. In this case,
idealized (and emotionless) models of decision making may be a source
of mistakes. For example, the rational choice of defection in a PD
game, or in a collective action problem, is normatively self-defeating
(because it reliably fails to maximize expected utility) and positively
wrong (because cooperation is common). Since the mid-1970s,
experimentation on PD and strategically equivalent games (such as
commons dilemmas) has invariably found what one researcher calls
“rampant cooperation.”92
People who irrationally cooperate in PD commonly do better than those
who rationally defect. In a variety of games of strategic interaction,
the rational choice is typically the worst choice.
93 See Colman's discussion of PD, and of Matching,
Centipede, and Ultimatum games. Colman
2003.
Although neurologists and neuroeconomists suggest
that humans may be hardwired to cooperate, I focus on how emotion
solves problems of strategic interaction.
94 Collective action becomes a problem when the benefits of
noncooperation are greater than the benefits of cooperation with the
group, unless no one else cooperates with the group, in which case the
costs of noncooperation exceed the costs of cooperation.95
If incentives encourage people to be selfish, but
collective selfishness undermines desired outcomes or important
societal values, how do people solve these problems? Changing the
payoffs is one solution, but this abolishes rather than solves the
problem.
96 Taylor
notes “external” and “internal” solutions to
collective action problems.
97 External solutions include
reliance on political entrepreneurs or on property rights, whereas
internal solutions rely on, for example, a sense of community. Like
Taylor, I emphasize community or identity as an internal solution to
collective action problems.
Trust is important to solving collective action problems, and I
suggest that emotion is the basis of trust. As noted by philosophers,
trust is a feeling of optimism in another's goodwill and
competence.98
Political scientists suggest that trust means
giving the other the benefit of the doubt, or is a feeling that the
other has benevolent intentions.
99 A social psychologist
finds that trust requires feelings of “warmth and
affection.”
100 I suggest that trust is
an emotional belief. Emotional beliefs are generalizations about
internal, enduring properties of an object that involve certainty
beyond evidence.
101 Trust
requires certainty beyond observable evidence and reliance instead on
how one feels about someone. For example, how people feel influences
their interpretations of another's behavior. People give the
benefit of the doubt to those they trust, and doubt anything beneficial
done by those they distrust. While using internally generated feelings
as evidence for trust departs from standard models of rationality,
people who are incapable of using their feelings are irrational, and
those who ignore their feelings and base cooperation on observable
facts cannot trust.
102 Rationalists do to trust what they do to reputation.103
The
belief that psychology explains mistakes, and thus the fear of
“going mental,” leads rationalists to drain psychology from
psychological concepts. As the philosopher Becker noted, rational
accounts of trust appear to eliminate that which they claim to
describe.
104 Rationalists drain the psychology from trust
by turning it into a consequence of incentives. Emphasizing incentives
as the basis for trust eliminates both the need for trust and the
opportunity to trust. If trust depends on external evidence,
transparency, iteration, or incentives, then trust adds nothing to the
explanation.
105 Equally importantly, emphasizing incentives as
the basis for trust eliminates the opportunity for trust to form. Trust
is a feeling toward someone, not toward an inanimate object or toward a
situation. I trust my friends but rely on my car.
106 If
observers attribute cooperation to the environment rather than the
person, then trust cannot—and need not—develop. My argument
is not that incentives are unimportant—though Chomsky's
observation that incentives depend on beliefs and desires merits
repeating—but that incentive-based behavior is not a substitute
for trust-based behavior.
I argue that identity produces emotion that creates trust, which then
solves collective action problems. Social identity theory (SIT)
dominates psychological approaches to group—not
individual—behavior, and explains emotion's role in
identity.107
Tajfel, a psychologist,
sought to understand the minimal conditions necessary to trigger
intergroup competition.
108 Surprisingly, merely placing
someone in an imaginary group is sufficient to trigger out-group
discrimination. The experimental results led Tajfel and Turner to
create SIT: when I identify with a group, the group becomes a part of
me and, as a result, I seek to view my group (and myself) as different
and better than other groups.
109 Tajfel and Turner 1986. Putting the group in the
individual avoids the problem of aggregating preferences (which in most
cases cannot be done rationally). See Arrow
1963.
Although most recent portrayals of SIT
marginalize or drop emotion, Tajfel recognized the importance of
emotion to identity.
110
Tajfel defined social identity as “that
part of an
individual's self-concept which derives from his knowledge of his
membership of a social group (or groups) together with the value and
emotional significance attached to that membership.”
111 A strong feeling of group identity
leads to sharing, cooperation, perceived mutuality of interests, and
willingness to sacrifice personal interests for group interests. It is
only because people invest groups with emotional significance that they
care about them; if people do not care, they neither cooperate nor
compete.
The emotion in identity provides the basis for trust. The
psychologist Brewer observed that group members “uniformly
evaluate members of their own group as high relative to those from
other groups on characteristics such as trustworthiness, honesty, and
loyalty. This consistency in the content of intergroup
perceptions has led to the speculation that, in addition to serving
functions related to positive personal identity, in-group bias also
serves to solve a pervasive social dilemma—that of the problem of
interpersonal trust.”112
Solving the problem of interpersonal trust solves
collective action problems. Because identification with a group
produces “positive emotions such as admiration, sympathy and
trust,” individuals will forgo their short-term interests and
cooperate to solve a common problem.
113 As
Tyler and his colleagues have discussed, the evidence linking identity
and trust to solving collective action problems is extensive.
114 Trusting
individuals cooperate (by restricting their consumption of a shared
resource) even when they have information that others are not
restricting their own consumption.
115 Others have
shown that identity is crucial to reducing competition within the
group, and that identification with a group is linked to a willingness
to conserve communal resources.
116 Even when it is in the interest of cooperators to
leave their group, they do not.
117 The emotion in
identity makes trust available to solve collective action problems.
Because identity demands discrimination, relying on identity and
trust to solve collective action problems is a double-edged sword.
Tajfel singled out the esteem emotions as the driving force behind
competition. It is the feeling of self-esteem, or pride, that makes
people in groups view their group as different and better than other
groups. In general, the more positively one feels about one's
group, the more negatively one feels about rivalrous groups.118
In-group trust does not require out-group
distrust—which is a feeling of pessimism about another's
goodwill and competence—but it does require one to distinguish
between trusting one's group and not trusting an out-group.
119 Whether discrimination is a
consequence of animus or an unintended consequence of preference for
one's group, the result for the out-group is similar. After
slaughtering 12,000 heathen in a day, Joshua gave thanks to God by
carving the commandment, “Thou shall not kill.” Joshua was
not being hypocritical, for the prohibition applied only to the
in-group, just as Moses preached to “love thy neighbor,” as
long as the neighbor was a member of the tribe.
120 Preferential treatment of one's own group is
a source of peace and cooperation within the group, and often of
distrust and hostility between groups.
121 Emotion drives in-group cooperation and out-group discrimination. If
this perspective is correct, then an emphasis on material incentives,
abstract cognitive beliefs, or incomplete information is mistaken.122
Stuart Kaufman reaches similar
conclusions in his study of ethnic conflict. Kaufman
2001.
The emotion in identity explains why group
members may trust each other and why they may distrust out-group
members. For example, should analysts view alliances between states as
instrumental, or should they sometimes view them as reflecting a
security community of shared values, beliefs, and trust? Both realists
and constructivists eliminate emotion from their assessments.
Constructivists, for example, view Deutsch's “we
feeling” (which provides the basis for a security community) as
indicating purely cognitive beliefs free of affect.
123 Because political
scientists typically view emotion as irrational, it is not surprising
that constructivists, no less than rationalists, are eager to purge it
from their explanations.
124 Instead of running from emotion, recognizing
emotion's role in trust and in identity may help analysts better
understand how alliances might work and how security communities might
form. Emotion in identity—or “we feeling”—may
be key to explaining when security communities form and, at least
according to realists, why they form so rarely.
A correspondence approach to rationality ought to be better than a
coherence approach when the explanation for a phenomenon rests on
psychological concepts such as trust, identity, or reputation. For
example, “justice” depends on emotion. As Solomon argued,
“The idea that justice requires emotional detachment, a kind of
purity suited ultimately only to angels, ideal observers, and the
original founders of society has blinded us to the fact that justice
arises from and requires such feelings as resentment, jealousy, and
envy as well as empathy and compassion.”125
If
emotion subverts rationality, then any conception of justice that
depends on emotion must capture irrational behavior. Rationalists drain
emotion from justice and assume self-interested actors that have
preferences over outcomes rather than preferences over strategies.
Rational people care only about the realization of their preferences
and are indifferent to how an outcome is obtained; they care about
their final payoffs but not the payoffs of others.
126 Yet preferences over process
appear to be just as important. How a decision is made, what the
other's motivation appears to be, and whether one is treated with
respect, consideration, and fairness often matter more than the
outcome.
127 Although rationalists
can use nonstandard hedonic values in their utility functions, doing so
risks creating unfalsifiable arguments and, even worse, acknowledging a
reliance on psychology. While political psychologists recognize that
justice depends on emotion, this dependence leads them to view a
concern for justice as a deviation from rationality.
128 It is as if a rational person must be greedy but
indifferent to justice.
Recognizing emotion's central role in justice has policy
implications. U.S. forces in Iraq have accidentally killed thousands of
Iraqi innocents. Their family members demand justice. A rationalist
solution might emphasize monetary compensation, because rational people
are self-interested and have preferences over outcomes. A psychological
solution would attend to process: expressions of regret, recognition
that a grave injustice occurred, and acknowledging changes in procedure
to make accidental deaths less likely. Addressing injustice in Iraq
depends on whether Iraqis are driven more by greed—better known
as self-interest—or, for example, by anger.129
Philosophers in the early seventeenth century began to
view greed as so powerful that, when harnessed with reason, it might
tame the other passions. Thus began the semantic shift from the emotion
“greed” into emotionless (and virtuous)
“interest.” Hirschman
1977.
The mistake is viewing greed as rational and
anger as irrational. They are both emotions and one is no less rational
than the other. Draining emotion from justice leads analysts to
misunderstand the concept and, perhaps, to pursue self-defeating
policies in Iraq.
Elster suggests that because emotion undermines rationality, the only
problems emotion solves are those it causes in the first place.130
Emotion can cause problems (such as out-group
discrimination), but emotion can also solve problems that rational
models cause (such as the rational strategy of defection in PD). The
concepts “identity” and “trust” have causal
power only when analysts recognize their dependence on emotion.
Identification without feeling implies a cold, neutral, bloodless
observation; it inspires no action. Trust without emotion implies an
expectation of trustworthiness based on incentives; trust adds nothing
if incentives explain cooperation. Emotion makes identity, trust, and
solutions to collective action problems possible. Political scientists
should view emotion not only as a problem to be overcome, but also as a
solution that can help individuals and groups overcome problems.
Damasio has shown that rationality depends on emotion, and I have
discussed how emotion may solve collective action problems and may
contribute to understanding alliance formation and justice. Psychology
does more than explain mistakes. It also helps prevent them.
