So, for example, one might derive something like the following principle of reasoning from the conjunction rule of probability theory:
Conjunction Principle: One ought not assign a lower degree of probability to the occurrence of event A than one does to the occurrence of A and some (distinct) event B (Stein 1996, 6).
Given principles of this kind, one can evaluate the specific judgments issued by human subjects and the mechanisms that produce them. To the extent that a person's judgments accord with the principles of the Standard Picture, these judgments are rational and to the extent that they violate such principles, the judgments fail to be rational. Similarly, to the extent that a reasoning mechanism produces judgments that accord with the principles of the Standard Picture, the mechanism is rational and to the extent that it fails to do so, it is not rational. As Piattelli-Palmarini puts the point:
The universal principles of logic, arithmetic, and probability calculus ....tell us what we should ... think, not what we in fact think.... If our intuition does in fact lead us to results incompatible with logic, we conclude that our intuition is at fault. (Piattelli-Palmarini, 1994, 158)
2.2.1 The Standard Picture and the Evaluation of Judgments
Proponents of the heuristics and biases program often appear to be in the business of evaluating the intuitive judgments that subjects make against the yardstick of the Standard Picture. As we noted earlier, Kahneman and Tversky say that "although errors of judgment are but a method by which some cognitive processes are studied, the method has become a significant part of the message." (Kahneman and Tversky 1982, 124) And the method-turned-message appears to be that many of our probabilistic judgments systematically deviate from the norms of rationality prescribed by the Standard Picture, specifically from those norms derived from probability theory (Kahneman and Tversky 1972, 431; Piatelli-Pallmarini 1994, 140). A recurrent theme in the heuristics and biases literature is that many of our intuitive judgments about probabilities deviate from the canons of probability theory in such a way that the deviations can be reliably reproduced under a wide range of circumstances that are related in their possession of certain key characteristics -- e.g. the manner in which information is presented to people or the content of the information about which people are asked to reason.
At first sight, this would appear to be a claim that evolutionary psychologists reject. Thus Gigerenzer asserts that "most so-called errors or cognitive illusions are, contrary to the assertions of the literature, in fact not violations of probability theory" (Gigerenzer 1991, 86). But on closer scrutiny, it is hard to see how evolutionary psychologists could reject the claim that many of our intuitive judgments systematically deviate from norms derived from probability theory. This is because some of the central features of their research program commit them to saying that human judgments do systematically deviate from these norms. In order to make this point we will focus on two features of the evolutionary psychological research program: (i) the empirical thesis that formulating probabilistic problems in terms of frequencies improves performance, and (ii) the ameliorative project of improving statistical reasoning by teaching subjects to reformulate probabilistic problems in terms of frequencies.
As we saw in Section 1, evolutionary psychologists maintain that when problems are explicitly formulated in terms of frequencies, performance improves dramatically. Consider, for example, the experiments on base-rate neglect. In 1.1 we discussed Cascells et al.'s "Harvard Medical School problem" and noted that it appears to show that, under certain circumstances, human beings systematically ignore information about base-rates when performing diagnostic tasks (Cascells et al. 1978). For our present purposes, the crucial point to notice about the Cascells et al. experiment is that the problem was formulated in a nonfrequentist format. Subjects were asked about the probability of single events -- the probability that a specific person has a disease -- and they were provided with probabilistic information in percentile and decimal formats. The results were disconcerting: 82% of subjects failed to provide the appropriate bayesian answer to the problem. By contrast, in 1.2 we saw that when presented with variants of the Harvard Medical School problem in which frequencies rather than percentages and single event probabilities were emphasized, subjects performed far better than they did in the original Cascells experiment. Although a number of different factors affect performance, according to Cosmides and Tooby, two predominate: "Asking for the answer as a frequency produces the largest effect, followed closely by presenting the problem information as frequencies." (58)
One central conclusion that evolutionary psychologists have wanted to draw from
these experiments is that human probabilistic judgment improves when problems are
reformulated in terms of frequencies.(5)
So, for example, Cosmides and Tooby claim that
"good statistical reasoning reappears, when problems are posed in frequentist terms."
(Cosmides and Tooby 1996, 62) This, however, poses a serious problem for the view that
evolutionary psychologists reject the heuristics and biases thesis that human beings
perform poorly in many judgmental tasks involving probabilities. After all, it's hard to
make sense of the claim that probabilistic judgment improves or that good statistical
reasoning reappears in frequentist tasks unless performance on non-frequency
problems
was poor, or at any rate less good, in the first place. Moreover, it is clear that the
metric
that evolutionary psychologists are employing in order to evaluate whether or not
probabilistic judgment improves is precisely the same as the one adopted by proponents of
the heuristics and biases program, namely: the standard axioms and theorems of
probability theory. It is precisely because judgments on many frequentist tasks
accord
with Bayes' theorem (and judgments on nonfrequentist tasks do not) that Cosmides and
Tooby claim that good statistical reasoning reappears when problems are posed in terms
of frequencies. The interpretation that evolutionary psychologists impose on their own
experimental data -- viz. that performance improves in frequentist tasks -- commits
them
to accepting the heuristics and biases thesis that many of our probabilistic judgments
deviate from appropriate norms of probabilistic reasoning. A similar point applies to another central feature of evolutionary psychological
research on human reasoning -- the ameliorative project of trying to improve human
probabilistic inference. In addition to providing empirical hypotheses about the cognitive
mechanisms responsible for inductive reasoning, evolutionary psychologists have also been
concerned with trying to improve the quality of probabilistic inference. This practical
project has been vigorously pursued by Gigerenzer and his colleagues. And in a series of
papers with titles like "How to improve bayesian reasoning without instruction: Frequency
formats" and "How to Improve Diagnostic Inferences in Physicians" they have shown how
probabilistic judgment can be improved by teaching subjects to convert problems into a
frequentist format (Gigerenzer & Hoffrage 1995; Hoffrage, Gigerenzer & Ebert in
press).
So, for example, Gigerenzer and his colleagues suggest that if physicians convert
diagnostic problems into a frequentist format, then they are more likely to be accurate in
their diagnoses. This sort of ameliorative project, once again, poses a serious problem for the
contention that evolutionary psychologists reject the heuristics and biases thesis that
human beings perform poorly in many judgmental tasks involving probabilities. For it is
extremely hard to see how we can make sense of the idea that performance can be
improved by converting problems into a frequency format unless subjects
were previously
doing something wrong. If there was nothing wrong, for example, with the answers that
physicians provided to diagnostic problems that were formulated in nonfrequentist terms,
then diagnosis couldn't be improved by formulating the problem in a frequentist
format.(6)
This is, we think, an entirely uncontroversial conceptual point. According to conventional
wisdom "If it ain't broken, don't fix it." Our point is rather more basic: If it ain't broken
you can't fix it. It is hard, then, to sustain the view that evolutionary psychologists reject the claim
that many of our probabilistic judgments deviate from the norms of probability theory.
What about a disagreement in the other direction? Do proponents of the heuristics and
biases tradition deny the evolutionary psychologists' claim that many of our intuitive
judgments about probability accord with the principles of probability theory? This is
a
suggestion that is hard to take seriously in the light of overwhelming textual evidence to
the contrary. Kahneman, Tversky and other advocates of the heuristics and biases
program note repeatedly that normatively problematic heuristics like "representativeness"
often get the right answer. Moreover, Kahneman and Tversky
maintain (correctly) that
they were responsible for discovering that formulating many judgmental problems in terms
of frequencies leads to a dramatic improvement in performance (Kahneman and Tversky,
1996). And, as we'll see later on, they have also attempted to explain this phenomenon by
providing an analysis of how the "extensional cues" provided by frequentist formulations
of probabilistic problems, facilitate reasoning. It is, therefore, singularly implausible to
maintain that proponents of heuristics and biases deny that there are many probabilistic
problems in which subjects' judgments accord with the probability calculus. We conclude
that if there is a dispute between evolutionary psychologists and proponents of heuristics
and biases, it is not located in the first box in Figure 1. So it is time to replace Figure 1
with Figure 2. 2.2.2 The Standard Picture and the Evaluation of
Mechanisms If there is no substantive disagreement between evolutionary psychologists and
proponents of the heuristics and biases tradition over whether or not our probabilistic
judgments accord with the principles of the Standard Picture, then perhaps a
disagreement
exists over whether or not the cognitive mechanisms that subserve probabilistic
reasoning
accord with these principles? Certainly, much of what has been said by participants in the
debate suggests such a disagreement. Thus, for example, Cosmides and Tooby explicitly
represent their project as a challenge to what they see as "the conclusion most common in
the literature on judgment under uncertainty -- that our inductive reasoning mechanisms
do not embody a calculus of probability." (Cosmides and Tooby 1996, 1) But when one
considers the issue more carefully it becomes difficult to sustain the view that there is any
genuine disagreement here -- or so we shall argue. In order to defend this claim, we'll start by arguing that the positive accounts of
probabilistic reasoning that evolutionary psychologists and proponents of heuristics and
biases have developed are not incompatible. Indeed, rather than being incompatible, the
views that have emerged from these two research programs about the nature of
probabilistic reasoning mechanisms are to a surprising degree complementary. For while
the heuristics and biases program has been primarily concerned with finding cases where
subjects do a bad job in their probabilistic reasoning and proposing mechanisms to explain
these shortcomings, evolutionary psychologists have been more concerned with positing
mechanisms in order to explain those instances in which our probabilistic reasoning is
normatively unproblematic. In short, the two research programs have simply focused on
different phenomena. Evolutionary psychologists have endorsed a range of claims about the mechanisms
that subserve probabilistic inference in human beings. One often repeated claim is that the
human mind contains a "multitude of elegant machines" for inductive reasoning: "many
different ones, each appropriate to a different kind of decision-making problem."
(Cosmides and Tooby 1996, 63) Moreover, evolutionary psychologists contend that at
least some of these mechanisms -- specifically, frequentist mechanisms -- are normatively
appropriate relative to precisely the same standard that the heuristics and biases program
endorses, viz. their input-output patterns match what would be required by the Bayesian
theory of probability. Thus Cosmides and Tooby suggest that "people do have reliably
developing mechanisms that allow them to apply the calculus of probability." (Cosmides
and Tooby 1996, 18). It is important to stress, however, that these frequentist mechanisms are supposed
to be format restricted; they are only able to process information that is presented
in the
appropriate format. More specifically, frequentist mechanisms "are designed to accept
probabilistic information when it is in the form of a frequency, and to produce a frequency
as their output."(ibid.) When probabilistic problems are presented in a non-frequentist
format, however, evolutionary psychologists contend that our judgments will deviate from
those prescribed by the calculus of probability because the frequentist mechanisms will be
unable to process the information.(7)
In short: according to evolutionary psychology,
whether or not our probabilistic reasoning mechanisms produce judgments that accord
with the probability calculus depends crucially on the format in which the information is
presented. The previous two paragraphs provide a brief description of the main positive
theses that evolutionary psychologists endorse about probabilistic inference in humans.
But it is important to stress that this cannot be the entire story. Nor, for that matter, do
evolutionary psychologists suggest that it is. Indeed they insist that we may well
need to
posit a wide range of other inductive mechanisms -- each of which operates according to
different principles -- in order to explain human reasoning (Cosmides and Tooby 1996,
63). One class of phenomena that is clearly in need of explanation are those instances in
which subjects respond to probabilistic problems in ways that deviate from the norms of
the probability calculus. These responses are not random but systematic in character.
And presumably a complete account of human probabilistic reasoning needs to explain the
inferential patterns that occur when we deviate from the probability calculus as well as
those that occur when we get things right. Though evolutionary psychologists clearly
accept this point and are prepared to posit additional mechanisms in order to explain the
results, they have, as yet, provided no detailed theory which accounts for these results.(8)
Nevertheless, they require an explanation. And presumably the explanation will need to
invoke mechanisms in addition to the frequentist mechanisms discussed above. Moreover,
these additional mechanisms will not map inputs onto the same outputs that the probability
calculus would and, hence, they will be normatively problematic by the lights of the
Standard Picture. Is there any reason to think that proponents of the heuristics and biases program
would or should disagree with any of this? As far as we can see, the answer is no.
First of
all, it is important to see that, according to the above picture of our reasoning architecture,
the total system will yield lots of mistakes, though it will also yield lots of correct answers.
And this is entirely consistent with the heuristics and biases account. Moreover,
proponents of the heuristics and biases program will clearly not want to reject the claim
that we possess cognitive mechanisms that fail to produce the input-output
mappings that
are sanctioned by the probability calculus. That there are such mechanisms is a central
claim of the heuristics and biases approach to human probabilistic reasoning. Indeed, it
would appear that the positive views that evolutionary psychologists endorse about the
nature of our reasoning architecture is consistent with the claim that the systems
responsible for producing nonbayesian judgments employ the sorts of heuristics that
Kahneman, Tversky and their followers have invoked in order to explain deviations from
the probability calculus. So, for example, it may be the case that some of the normatively
problematic mechanisms that evolutionary psychologists must posit to explain normatively
problematic judgments implement the representativeness and availability heuristics. At this point it might be suggested that proponents of the heuristics and biases
program reject the existence of mechanisms that operate according to principles of the
probability calculus. This could be either because (a) they reject the existence of more
than one reasoning mechanism or (b) while they accept the existence of more than one
reasoning mechanism, they deny that any of them operate according to the principles of
probability. Let's consider these options in turn. Evolutionary psychologists sometimes appear to suggest that proponents of the
heuristics and biases program are wedded to the assumptions that there are no domain
specific or modular mechanisms for reasoning and that all reasoning is subserved by
general-purpose processes and mechanisms. So, for example, Cosmides and Tooby
appear to attribute to the heuristics and biases program "a certain old-fashioned image of
the mind: that it has the architecture of an early model, limited-resource general-purpose
computer." (Cosmides and Tooby 1996, 13) There is plenty of textual evidence, however,
that proponents of the heuristics and biases program do not endorse such a picture of the
mind. So, for example, in a passage that anticipates a central theme in the work of
evolutionary psychologists, Kahneman and Tversky compare the processes involved in the
solving of probabilistic problems "with the operation of a flexible computer program that
incorporates a variety of potentially useful subroutines."(9)
(Kahneman and Tversky 1983,
88) Elsewhere, they are even more explicit on the matter and claim that "the actual
reasoning process is schema-bound or content-bound so that different operations or
inferential rules are available in different contexts" and that "consequently, human
reasoning cannot be adequately described in terms of content-independent formal rules."
(Kahneman and Tversky 1983, 499). Piattelli-Palmarini is still more explicit in his
endorsement of a domain-specific conception of human reasoning and goes so far as to
suggest (rightly or wrongly) that judgmental errors are "a demonstration of what modern
cognitive science calls the 'modularity' of the mind." (Piattelli-Palmarini 1994, 32) In
other words, Piattelli-Palmarini appears to be endorsing the claim that we possess modules
for reasoning. So proponents of the heuristics and biases program do not appear to be adverse to
the idea that human reasoning is subserved by a variety of domain specific cognitive
mechanisms. Do they, perhaps, deny that any of these mechanisms operate according to
the principles of the probability calculus? If they did maintain this position, then there
would be a genuine disagreement between evolutionary psychologists and proponents of
the heuristics and biases program. But there is, in fact, no reason to suppose that they do
hold such a view. First of all, nowhere in the heuristics and biases literature have we been
able to find a single passage in which it is explicitly denied that we possess some
cognitive
mechanisms that operate according to the principles of the probability calculus. What we
do find, however, are passages that may be interpreted as suggesting that there are
no
such mechanisms. So, for example, as we noted earlier, Kahneman and Tversky (1973)
claim that This and other similar passages in the heuristics and biases literature might be thought to
have the conversational implicature that we only use normatively problematic
heuristics in
our probabilistic reasoning and hence possess no reasoning mechanisms that operate
according to the principles of the probability calculus. We maintain, however, that there are extremely good reasons to treat such claims
as instances of rhetorical excess. First, as we pointed out in 1.1, the claim that we possess
no normatively unproblematic mechanisms for probabilistic reasoning is clearly not
supported by the available empirical evidence. Such a claim is vastly stronger than
anything the available evidence will support. And this provides us with some reason to
treat it as a rhetorical flourish as opposed to a core claim of the heuristics and biases
research program. Second, all the quotations from the heuristics and biases literature that we have
found which suggest that humans possess no normatively appropriate reasoning
mechanisms, manifest a tendency that Kahneman and Tversky have, themselves, lamented
-- the tendency to overstate one's position by "omitting relevant quantifiers." (Kahneman
and Tversky 1996, 589) Kahneman and Tversky raise this point in response to
Gigerenzer's claim that cognitive illusions disappear when problems are formulated in
terms of frequencies. They suggest that "because Gigerenzer must be aware of the
evidence that judgments of frequency ... are subject to systematic error, a charitable
interpretation of his position is that he has overstated his case by omitting relevant
quantifiers." (Kahneman and Tversky 1996) We maintain that much the same may be
said of the position that Kahneman, Tversky and their followers sometimes appear to
endorse regarding the normative status of our reasoning mechanisms. Consider, for
example, the above quotation from Kahneman and Tversky (1973). The natural reading
of this passage is that Kahneman and Tversky are claiming that humans always
"rely on a
limited number of heuristics" (Kahneman and Tversky 1973, 48). But notice that the
relevant quantifier is omitted. It is left unspecified whether they are claiming that we
always use normatively problematic heuristics as opposed to (for example) claiming
that
we typically or often use such heuristics. And because they must
know that the truth of
the natural reading is vastly underdetermined by the data, it is surely charitable to interpret
this as an instance of rhetorical excess -- an overstatement of their position that results
from omitting relevant quantifiers. Moreover, this point generalizes: In all the
passages
from the heuristics and biases literature that we have found which suggest that humans
possess no normatively appropriate reasoning mechanisms, relevant quantifiers are
systematically omitted. We suggest, therefore, that because proponents of the heuristics
and biases program are presumably aware that the available evidence fails to support the
claim that humans possess no normatively unproblematic reasoning mechanisms, the
charitable interpretation of these quotations is that they overstate the position by omitting
relevant quantifiers. A final point that further supports the conclusion of the previous paragraph is that,
in their more reflective moments --when quantifiers are not omitted -- advocates of the
heuristics and biases tradition make it clear that they are not maintaining that we
always
use normatively problematic heuristics and mechanisms in our intuitive reasoning. Instead
they explicitly claim only that we sometimes or often use such
heuristics and mechanisms.
So, for example, when they are being careful, Kahneman and Tversky claim only that
"intuitive predictions and judgments are often mediated by a small number of
distinct
mental operations ... [or] ... judgmental heuristics." (Kahneman and Tversky 1996) But
this position is entirely compatible with the evolutionary psychological view that we also
possess some normatively unproblematic reasoning mechanisms. In short: when
proponents of the heuristics and biases tradition express their views carefully and fill in the
appropriate quantifiers, they end up maintaining a position about the normative status of
our reasoning mechanisms that does not conflict with the claims of evolutionary
psychologists. It is time, then, to replace Figure 2 by Figure 3. 2.3 Accuracy-Based Assessments Though the Standard Picture is the normative yardstick most commonly invoked in
the dispute between evolutionary psychology and the heuristics and biases program, it is
not the only one. Another kind of normative standard is suggested by Gigerenzer's
discussion of "Take the Best" and other members of a class of satisfying algorithms that he
calls "fast and frugal" procedures (Gigerenzer, Hoffrage and Kleinbölting, (1991);
Gigerenzer and Goldstein 1996). According to Gigerenzer, a central consideration when
evaluating reasoning is its accuracy (Gigerenzer and Goldstein 1996, 665). And
because
fast and frugal algorithms get the correct answer at least as often as other computationally
more expensive, "rational"(10)
methods (such as standard statistical linear models)
Gigerenzer clearly thinks that they are normatively unproblematic. Indeed he thinks that
the fact that these simple algorithms are accurate constitutes a refutation of the claim that
only "rational" algorithms can be accurate and goes some way towards overcoming the
"opposition between the rational and the psychological and to reunite the two."
(Gigerenzer and Goldstein 1996, 666) Although the notion of accuracy applies to both judgments and cognitive
mechanisms, Gigerenzer and other evolutionary psychologists are concerned primarily
with the accuracy of mechanisms (Gigerenzer and Goldstein 1996; Cosmides and Tooby
1996). Moreover, it is also clear that once we address the issue of whether or not
evolutionary psychologists and proponents of the heuristics and biases tradition disagree
about the accuracy of our cognitive mechanisms, the same considerations apply mutatis
mutandis to the putative disagreement over judgments. For this reason we will focus
primarily on whether or not there is any genuine disagreement between evolutionary
psychology and the heuristics and biases program over the accuracy of our cognitive
mechanisms. When applied to cognitive mechanisms, Gigerenzer's accuracy-based
criterion for epistemic evaluation bears an intimate relationship to the reliabilist
tradition in epistemology according to which (very roughly) a cognitive mechanism
is rational just in case it tends to produce true beliefs and avoid producing false
ones (Goldman 1986; Nozick 1993).(11)
One frequently observed consequence of
reliabilist and accuracy-based approaches to the evaluation of cognitive
mechanisms is that assessments must be relativized to some environment or
domain of information (Goldman 1986; Nozick 1993; Stich 1990). A visual
system, for example, is not reliable or unreliable simpliciter, but only reliable or
accurate relative to a (set of) environment(s) or a domain of information.(12)
Moreover, there is an indefinitely wide range of environments or domains to which
evaluations might be relativized. For present purposes, however, let's focus on
two that have been suggested by Dan Sperber to be particularly relevant to
understanding the evolutionary psychological approach to reasoning -- what he
calls the actual domain and the proper domain for a cognitive
mechanism (Sperber
1994). The actual domain for a given reasoning module is "all the information in
the organism's environment that may (once processed by perceptual modules, and
possibly by other conceptual modules) satisfy the module's input conditions (51-2). By "input
conditions" Sperber means those conditions that must be satisfied in
order for the module to be able to process a given item of information. So, for
example, if a module requires that a problem be stated in a particular format, then
any information not stated in that format fails to satisfy the module's input
conditions. By contrast, the proper domain for a cognitive mechanism is all the
information that it is the mechanism's "biological function to process." ( 52) The
proper domain is the information that the mechanism was designed to process by
natural selection. In recent years, many philosophers of biology have come to
regard the notion of a biological function as a particularly slippery one.(13)
For
current purposes we can rely on the following very rough characterization: The
biological functions of a system are the activities or effects of the system in virtue
of which the system has remained a stable feature of an enduring species. Do evolutionary psychologists and proponents of the heuristics and biases
tradition disagree about the accuracy of reasoning mechanisms in their proper
domains? Clearly not. For while evolutionary psychologists have maintained that
cognitive mechanisms will tend to perform accurately in their proper domains -- on
the kinds of information that they are designed to process -- the heuristics and
biases tradition has been entirely silent on the issue. Determining the accuracy of
cognitive mechanisms in the proper domain is simply not the line of work that
proponents of heuristics and biases are engaged in. So, there could be no
disagreement here. It is similarly implausible to maintain that evolutionary psychologists and
advocates of the heuristics and biases tradition disagree over the accuracy of our
reasoning mechanisms in the actual domain. Clearly, evolutionary psychologists
think that some of our reasoning mechanisms are accurate in the actual domain.
But it is equally clear that they do not claim that all of them are. They certainly
cite no evidence that could support the claim that all of our reasoning mechanisms
are accurate in the actual domain. And, what is more important, such a claim
would be patently incompatible with their ameliorative project. If all our reasoning
mechanisms are accurate in the actual domain, then there is little room for
systematically improving human reasoning. So it must be the case that what
evolutionary psychologists want to claim is that some but not all of our reasoning
mechanisms are accurate in the actual domain. Do proponents of the heuristics and biases tradition reject this claim? As
far as we can see, the answer is no. They clearly think that some of
our cognitive
mechanisms are inaccurate in the actual domain. This, after all, is a central
message of their research program. But they have been largely silent on the issue
of whether or not we possess other reasoning mechanisms that are accurate in the
actual domain. And this is simply because, as we mentioned earlier on, proponents
of the heuristics and biases tradition have primarily focused on explaining instances
of incorrect judgment as opposed to explaining instances of successful inference.
Nonetheless, as we saw in 2.2, theorists working within the heuristics and biases
tradition are not adverse to the idea that we have reasoning mechanisms other than
the ones which employ normatively problematic heuristics and, to the extent that
they say anything about these other mechanisms, they seem amenable to the idea
that they may be accurate. So, for example, Kahneman and Tversky seem entirely
comfortable with the idea that mechanisms that employ correct rules of
probabilistic inference can produce highly accurate judgments in contexts where
the problem is transparent and "extensional" cues are effective (Kahneman and
Tversky 1983). The situation is similar when we turn to the issue of whether or not
evolutionary psychologists and advocates of the heuristics and biases approach
disagree over the accuracy of our judgments. Evolutionary psychologists think
that we tend to be accurate in the proper domain whereas proponents of the
heuristics and biases program are simply silent on the issue. And both parties
appear to think that many but not all of our judgments are accurate in the
actual
domain. There are, of course, lots of issues of detail where the two research
programs disagree. So, for example, Gigerenzer has challenged some of the
interpretations that advocates of the heuristics and biases program have imposed
on specific experiments. We will consider some of these cases in Section 3. But
we maintain that these disagreements are merely matters of detail and ought not to
distract from the genuine consensus between evolutionary psychology and the
heuristics and biases program. Both programs clearly accept that many of our
judgments in the actual domain are inaccurate; that we are subject to systematic
errors. This is a central claim of the heuristics and biases program and
evolutionary psychology is similarly committed to this view by virtue of endorsing
the ameliorative project. Moreover, neither program insists that all of our
judgments are inaccurate. Both, for example, think that our judgments about
frequency can be highly accurate. Again, there is no disagreement. So we can
now replace Figure 3 with Figure 4. 2.4 Constrained-Optimality Assessments A final normative standard that has been invoked by participants in the debate
between evolutionary psychology and the heuristics and biases tradition is one that applies
only to the evaluation of cognitive mechanisms and not to the judgments that these
mechanisms produce. The standard in question maintains that a reasoning mechanism is
normatively unproblematic to the extent that it is optimal given the constraints to which it
is subject. This proposal is alluded to by Gigerenzer when he suggests that some
reasoning mechanisms may be optimal in the way that Herman von Helmholtz and Richard
Gregory propose that visual processing mechanisms are optimal: they are the best systems
available for acquiring an accurate picture of the world given the constraints under
which they must operate. One crucial point to stress is that the best system (given the
constraints under which it operates) need not be a system that never make mistakes. As
Gigerenzer points out, such "systems can be fooled and may break down when stable,
long-term properties of the environment to which they are adapted change." (Gigerenzer
1998b, 10) So, for example, Gregory maintains that visual "illusions will be a necessary
part of all efficiently designed visual machines" -- even the best designed visual
systems
(Gregory, quoted in Gigerenzer 1991, 228). Similarly, Gigerenzer suggests that, given
the constraints under which real cognitive systems must operate, "cognitive illusions" or
"biases" will be a necessary part of an efficiently designed reasoning mechanism. Thus the
Helmholtzian view "allows both for optimal cognitive functioning and for systematic
illusions" (Gigerenzer 1991, 240). Is there any disagreement between evolutionary psychologists and proponents of
the heuristics and biases program on the issue of whether or not we possess mechanisms
that are optimally well-designed (given the appropriate constraints) for probabilistic
reasoning? Once again, we maintain, the answer is no. While evolutionary
psychologists
have suggested that we possess mechanisms that are optimal in the relevant sense,
proponents of the heuristics and biases program need not and do not deny this claim. To
see why, it is important to note that when evolutionary psychologists suggest that we
possess reasoning mechanisms that are optimal given the constraints, they typically appear
to have in mind the claim that we possess cognitive mechanisms that are optimally well-designed
for processing information in their proper domains (and under conditions similar
to those our evolutionary ancestors would have encountered) and not the claim that we
possess mechanisms that are optimally well-designed for processing information in their
actual domains. Thus, for example, Cosmides and Tooby suggest that "our minds
come
equipped with very sophisticated intuitive competences that are well-engineered solutions
to the problems humans normally encountered in natural environments ... and that
ecologically valid input (e.g. frequency formats) may be necessary to activate these
competences" (Cosmides and Tooby 1996, 9). But if the notion of optimality invoked by
evolutionary psychologists is indexed to the proper domain, then, as we have already seen
in 2.2, proponents of the heuristics and biases program do not disagree. The heuristics
and biases program simply is not concerned with the performance of cognitive mechanisms
in their proper domains. Suppose however that, contrary to appearances, evolutionary psychologists do
wish to maintain that we possess reasoning mechanisms that are optimal relative to the
actual domain. Even so, they clearly could not maintain that all of our reasoning
mechanisms are optimal since, once again, such a view would render their ameliorative
project impossible. If all our reasoning mechanisms were the best that they could be, then
we couldn't make them better. Here the (dis)analogy between visual systems and
reasoning systems is illuminating. It is plausible to claim that when functioning normally
our visual systems are optimal in the sense that they simply cannot be improved. By
contrast, we can improve our reasoning -- hence the ameliorative project. So, the
most
that evolutionary psychologists could be claiming is that some or, perhaps many, of our
cognitive mechanisms are optimal relative to the actual domain. But this is not a claim
that proponents of the heuristics and biases tradition either do or should reject. To the
best of our knowledge, proponents of the heuristics and biases program have never denied
that we possess some reasoning mechanisms that are optimal in this sense. What they do
deny is that all of the cognitive mechanisms subserving reasoning are optimal in the
sense
that they always produce judgments that are correct and/or accord with the principles of
the probability calculus. This, however, is a very different notion of optimality -- a notion
of optimality that does not take into consideration the constraints under which our
reasoning systems must operate. There is no reason to suppose that the heuristics and
biases program is committed to denying that we possess cognitive mechanisms that are
optimal in the actual domain given the constraints under which they operate. For,
as we
have already seen, the claim that a reasoning system is optimal (given the appropriate
constraints) is perfectly consistent with the view that it is subject to lots of biases and
cognitive illusions. Thus proponents of the heuristics and biases program need not and do
not deny that some or even many of our cognitive mechanisms may be optimal in the
Helmholtzian sense that Gigerenzer and other evolutionary psychologists have in mind.
And if this correct, then we can replace Figure 4 with Figure 5. The main burden of this paper has been to dispel the illusion that there is any
substantive disagreement between evolutionary psychologists and advocates of the
heuristics and biases tradition concerning the extent of human rationality. We do not
intend to suggest, however, that there is nothing left for evolutionary psychologists and
proponents of the heuristics and biases program to disagree about. Clearly there is.
Indeed there are a number of different disputes that remain. One of these disputes focuses
on the issue of how we ought to apply probability theory to specific problems in the
heuristics and biases literature -- e.g. the Lawyer/Engineer problem and the Harvard
Medical School problem -- and whether or not probability theory provides a uniquely
correct answer to these problems. Though authors in the heuristics and biases tradition
often appear to assume that there is only one normatively correct answer to these
problems, Gigerenzer has argued that there are typically a number of equally reasonable
ways of applying probability theory to the problems and that these different analyses result
in distinct but equally correct answers (Gigerenzer 1991 & 1994). Another very real dispute concerns the adequacy of the explanations proposed by
proponents of the heuristics and biases tradition -- explanations that invoke heuristics,
such as availability and representativeness, in order to explain cognitive phenomena.
Evolutionary psychologists have maintained that these "heuristics are too vague to count
as explanations" and that psychologists working in the heuristics and biases tradition have
failed to "specify precise and falsifiable process models, to clarify the antecedent
conditions that elicit various heuristics, and to work out the relationship between
heuristics." (Gigerenzer 1996, 593) Proponents of the heuristics and biases tradition have
responded by arguing that evolutionary psychologists have "missed the point." (Kahneman
and Tversky 1996) They maintain that representativeness and other heuristics "can be
assessed experimentally" and that testing the hypothesis that probability judgments are
mediated by these heuristics "does not require a theoretical model." (ibid.) On our view, both of these disputes raise deep and interesting questions which we
plan to address elsewhere. In the present section we propose to focus on a third very real
dispute between evolutionary psychologists and proponents of the heuristics and biases
tradition, one which has often been center stage in the literature. This is the disagreement
over what interpretation of probability theory to adopt. There has been a long-standing disagreement between proponents of the heuristics
and biases program and evolutionary psychologists over what we should recognize as the
correct interpretation of probability theory. In contrast with psychologists in the heuristics
and biases tradition, Gigerenzer has urged that probability theory ought to be given a
frequentist interpretation according to which probabilities are construed as relative
frequencies of events in one class to events in another. As Gigerenzer points out,
according to "this frequentist view, one cannot speak of a probability unless a reference
class is defined." (Gigerenzer 1993, 292-293) So, for example, "the relative frequency of
an event such as death is only defined with respect to a reference class such as 'all male
pub-owners fifty-years old living in Bavaria'." (ibid.) One consequence of this that
Gigerenzer is particularly keen to stress is that, according to frequentism, it makes no
sense to assign probabilities to single events. Claims about the probability of a single
event are literally meaningless: For a frequentist ... the term "probability", when it refers to a single event,
has no meaning at all for us (Gigerenzer 1991, 88). Moreover, Gigerenzer maintains that because of this "a strict frequentist" would argue
that "the laws of probability are about frequencies and not about single events" and,
hence, that "no judgment about single events can violate probability theory." (Gigerenzer
1993, 292-293) In stark contrast with Gigerenzer's frequentism, Kahneman, Tversky and their
followers insist that probability theory can be meaningfully applied to single events
and
hence that judgments about single events (e.g. Jack being a engineer or, in another well
known problem, Linda being a bank teller(14)) can violate probability
theory. This
disagreement emerges very clearly in Kahneman and Tversky (1996) where they argue
that Gigerenzer's treatment of judgment under uncertainty "appears far too restrictive"
because it "does not apply to events that are unique for the individual and, therefore,
excludes some of the most important evidential and decision problems in people's lives."
(Kahneman and Tversky 1996, 589) Instead of adopting frequentism, Kahneman and
Tversky suggest that some "subjectivist" or "Bayesian" account of probability may be
preferable. This disagreement over the interpretation of probability raises complex and
important questions in the foundations of statistics and decision theory about the scope
and limits of our formal treatment of probability. Moreover, the dispute between
frequentists and subjectivists has been a central debate in the foundations of probability for
much of the Twentieth century (von Mises 1957; Savage 1972). Needless to say, a
satisfactory treatment of these issues is beyond the scope of the present paper. But we
would like to comment briefly on what we take to be the central role that issues about the
interpretation of probability theory play in the dispute between evolutionary psychologists
and proponents of the heuristics and biases program. In particular, we will argue that
Gigerenzer's use of frequentist considerations in this debate is deeply problematic. Questions about the interpretation of probability entered the debate between
evolutionary psychology and the heuristics and biases tradition primarily because it was
realized by some theorists -- most notably Gigerenzer -- that these questions bear on the
issue of whether or not human reasoning violates appropriate norms of rationality. As we
have already seen, Gigerenzer argues that if frequentism is true, then statements about the
probability of single events are meaningless and, hence, that judgments about single events
cannot violate probability theory (Gigerenzer 1993, 292-293). Gigerenzer clearly
thinks
that this conclusion can be put to work in order to dismantle part of the evidential base for
the claim that human judgments and reasoning mechanisms violate appropriate norms.
For, as we have seen, participants in the debate between evolutionary psychology and the
heuristics and biases tradition typically view probability theory as the source of
appropriate normative constraints on probabilistic reasoning. And if frequentism is true,
then no probabilistic judgments about single events will be normatively problematic (by
this standard) since they will not violate probability theory. In which case Gigerenzer gets
to exclude all experimental results involving judgments about single events as evidence for
the existence of normatively problematic, probabilistic judgments and reasoning
mechanisms. On the face of it, Gigerenzer's strategy is quite persuasive. Nevertheless we think
that it is subject to serious objections. Frequentism itself is a hotly contested view, but
even if we grant, for argument's sake, that frequentism is correct, there are still serious
grounds for concern. First, as we observed in footnote 6, there is a serious tension
between the claim that subjects don't make errors in reasoning about single events and the
ameliorative project that evolutionary psychologists are engaged in. The present point is
not that frequentism is false but merely that evolutionary psychologists cannot comfortably
maintain both (a) that we don't violate appropriate norms of rationality when reasoning
about the probabilities of single events and (b) that reasoning improves when single event
problems are converted into a frequentist format. A second and perhaps more serious problem with Gigerenzer's use of frequentist
considerations is that it is very plausible to maintain that even if statements about
the
probabilities of single events really are meaningless and hence do not violate the
probability calculus, subjects are still guilty of making some sort of
error when they deal
with problems about single events. For if, as Gigerenzer would have us believe,
judgments about the probabilities of single events are meaningless, then surely the correct
answer to a (putative) problem about the probability of a single event is not some
numerical value or rank ordering, but rather: "Huh?" or "That's utter nonsense!" or
"What on earth are you talking about?" Consider an analogous case in which you are
asked to answer a question like: "Is Linda taller than?" or "How much taller than is
Linda?" Obviously these questions are nonsense because they are incomplete. In order to
answer them we must be told what the other relatum of the "taller than" relation is
supposed to be. Unless this is done, answering "yes" or "no" or providing a numerical
value would surely be normatively inappropriate. Now according to the frequentist, the
question "What is the probability that Linda is a bank teller?" is nonsense for much the
same reason that "Is Linda taller than?" is. So when subjects answer the single event
probability question by providing a number they are doing something that is clearly
normatively inappropriate. The normatively appropriate answer is "Huh?", not "Less than
10 percent". It might be suggested that the answers that subjects provide in experiments
involving single event probabilities are an artifact of the demand characteristics of the
experimental context. Subjects (one might claim) know, if only implicitly, that single
event probabilities are meaningless. But because they are presented with forced choice
problems that require a probabilistic judgment, they end up giving silly answers. So, one
might think that the take-home message is "Don't blame the subject for giving a silly
answer. Blame the experimenter for putting the subject in a silly situation in the first
place!" But this proposal is implausible for two reasons. First, as a matter of fact,
ordinary people use judgments about single event probabilities in all sorts of circumstances
outside of the psychologist's laboratory. So it is implausible to think that they view
single
event probabilities as meaningless. But, second, even if subjects really did think that single
event probabilities were meaningless, presumably we should expect them to provide more
or less random answers and not the sorts of systematic responses that are observed in the
psychological literature. Again, consider the comparison with the question "Is Linda taller
than?" It would be a truly stunning result if everyone who was pressured to respond said
"Yes." The main aim of this paper has been to dispel an illusion: the illusion that
evolutionary psychology and the heuristics and biases tradition are deeply divided in their
assessments of human reasoning. We started by outlining the two research programs and
disentangling their core claims from the rhetorical flourishes that have obscured an
emerging consensus between the two programs about the scope and limits of human
rationality and about the cognitive architecture that supports it. We then showed that,
contrary to appearances, there is no substantial disagreement between evolutionary
psychologists and advocates of the heuristics and biases program over the extent of human
rationality. On a number of different readings of what the dispute is supposed to be,
neither research program denies the core claims of the other and, in many cases, it is
clear that they should and do endorse each other's core claims. Finally, we briefly
focused on some of the points of disagreement that remain once the illusory dispute has
disappeared. Though there are some important issues dividing evolutionary psychologists
and advocates of the heuristics and biases program, there is also a surprising degree of
consensus. Moreover, and this has been our central theme, they do not really have any
deep disagreement over the extent of human rationality. Acknowledgments An earlier version of this paper was discussed at a Workshop on the Evolution of Mind at
the Hang Seng Centre for Cognitive Studies at the University of Sheffield. We are
grateful for the many helpful comments and criticisms that were offered on that occasion.
Special thanks are due to George Botterill, Richard Byrne, Peter Carruthers, Gerd
Gigerenzer, Brian Loar, Adam Morton and Michael Segal. References Barkow, J., Cosmides, L., and Tooby, J. (eds.), (1992). The Adapted Mind:
Evolutionary Psychology and the Generation of Culture. Oxford: Oxford University
Press. Baron-Cohen, S. (1995). Mindblindness: An Essay on Autism and Theory of
Mind.
Cambridge, MA: MIT Press. Bazerman, M. and Neale, M. (1986). Heuristics in negotiation. In H. Arkes & K
Hammond (eds.), Judgment and Decision Making: An Interdisciplinary Reader.
Cambridge: Cambridge University Press. Bower, B. (1996). Rational mind design: research into the ecology of thought treads on
contested terrain. Science News, 150, 24-25. Casscells, W., Schoenberger, A. and Grayboys, T. (1978). Interpretation by physicians of
clinical laboratory results. New England Journal of Medicine, 299, 999-1000. Cosmides, L. (1989). The logic of social exchange: Has natural selection shaped how
humans reason? Studies with Wason Selection Task. Cognition, 31, 187-276. Cosmides, L. and Tooby, J. (1992). Cognitive adaptations for social exchange. In
Barkow, Cosmides and Tooby (1992), 163-228. Cosmides, L. and Tooby, J. (1994). Origins of domain specificity: The evolution of
functional organization. In Hirschfeld and Gelman (1994), 85-116. Cosmides, L. and Tooby, J. (1996). Are humans good intuitive statisticians after all?
Rethinking some conclusions from the literature on judgment under uncertainty."
Cognition, 58, 1, 1-73. Fodor, J. (1983). The Modularity of Mind. Cambridge, MA: MIT Press. Gigerenzer, G. (1991a). How to make cognitive illusions disappear: Beyond 'heuristics
and biases.' European Review of Social Psychology, 2, 83-115. Gigerenzer, G. (1991b). On Cognitive Illusions and Rationality. Poznan Studies in
the
Philosophy of the Sciences and the Humanities, Vol. 21, 225-249. Gigerenzer, G. (1993). The bounded rationality of probabilistic mental models. In K.
Manktelow and D. Over, eds., Rationality: Psychological and Philosophical
Perspectives. London: Routledge. Gigerenzer, G. (1994). Why the distinction between single-event probabilities and
frequencies is important for psychology (and vice versa). In G. Wright and P. Ayton, eds.,
Subjective Probability. New York: John Wiley. Gigerenzer, G. (1996). On narrow norms and vague heuristics: A reply to Kahneman
and Tversky (1996). Psychological Review, 103, 592-596. Gigerenzer, G. (1998a). The modularity of social intelligence. In A. Whiten and R. Byrne
(eds.), Machiavellian Intelligence II. Cambridge: Cambridge University Press. Gigerenzer, G. (1998b). Ecological intelligence: An adaptation for frequencies. In D.
Cummins and C. Allen, eds., The Evolution of Mind. New York: Oxford University
Press. Gigerenzer, G. and Goldstein, D. G. (1996). Reasoning the fast and frugal way: Models of
bounded rationality. Psychological Review, 103, 650-669. Gigerenzer, G. and Hug, K. (1992). Domain-specific reasoning: Social contracts,
cheating and perspective change. Cognition, 43, 127-171. Gigerenzer, G., and Hoffrage, U. (1995). How to improve Bayesian reasoning without
instruction: Frequency formats. Psychological Review, 102, 684-704. Gigerenzer, G., Hoffrage, U., and Kleinbölting, H. (1991). Probabilistic mental
models:
A Brunswikean theory of confidence. Psychological Review, 98, 506-528. Gigerenzer, G. and Murray, D. (1987). Cognition as Intuitive Statistics,
Hillsdale, NJ:
Erlbaum. Godfrey-Smith, P. (1994). A modern history theory of functions. Nous, 28,
344-362. Goldman, A. (1986). Epistemology and Cognition, Cambridge, MA: Harvard
University
Press. Gould, S. (1992). Bully for Brontosaurus. Further Reflections in Natural
History.
London: Penguin Books. Harman, G. (1986). Change in View. Cambridge, MA: MIT Press. Hertwig, R. and Gigerenzer, G. (1994). The chain of reasoning in the conjunction task.
Unpublished manuscript. Hirschfeld, L. and Gelman, S. (eds.), (1994). Mapping the Mind. Cambridge:
Cambridge University Press. Hoffrage, U., Gigerenzer, G. and Ebert, . (In press) How to improve diagnostic
inferences in physicians. Kahneman, D. and Tversky, A. (1972). Subjective probability: A judgment of
representativeness. Cognitive Psychology, 3, 340-354. Kahneman, D. and Tversky, A. (1973). On the psychology of prediction.
Psychological
Review, 80, 237-251. Reprinted in Kahneman, Slovic and Tversky (1982). Kahneman, D. and Tversky, A. (1996). On the reality of cognitive illusions: A reply to
Gigerenzer's critique. Psychological Review, 103, 582-591. Kahneman, D., Slovic, P. and Tversky, A. (1982). Judgment Under Uncertainty:
Heuristics and Biases. Cambridge: Cambridge University Press. Lichtenstein, S., Fischoff, B. and Phillips, L. (1982). "Calibration of probabilities: The
state of the art to 1980. In Kahneman, Slovic and Tversky (1982), 306-334. Mises, R. v. (1957) Probability, Statistics and Truth. London: Allen and
Unwin. Neander, K. (1991). The teleological notion of 'function'. Australasian Journal of
Philosophy, 59, 454-468. Nisbett, R. and Borgida, E. (1975). Attribution and the social psychology of prediction.
Journal of Personality and Social Psychology, 32, 932-943. Nisbett, R. and Ross, L. (1980). Human Inference: Strategies and Shortcomings of
Social Judgment. Englewood Cliffs, NJ: Prentice-Hall. Nisbett, R., Fong, G, Lehman, D. and Cheng, P. (1987). Teaching reasoning.
Science,
238, 625-631. Nozick, R. (1993). The Nature of Rationality. Princeton, NJ: Princeton
University
Press. Piattelli-Palmarini, M. (1994). Inevitable Illusions: How Mistakes of Reason Rule Our
Minds. New York: John Wiley & Sons. Pinker, S. (1994). The Language Instinct. New York: William Morrow and
Co. Pinker, S. (1997). How the Mind Works. New York: W. W. Norton. Plantinga, A. (1993). Warrant and Proper Function. Oxford: Oxford
University Press. Samuels, R. (forthcoming). Evolutionary psychology and the massive modularity
hypothesis. To appear in The British Journal for the Philosophy of Science. Samuels, R., Stich, S. and Tremoulet, P. (forthcoming). Rethinking rationality: From
bleak implications to Darwinian modules. To appear in E. LePore & Z. Pylyshyn (eds.)
Rutgers University Invitation to Cognitive Science. Oxford: Basil Blackwell. Savage, L. J. (1972). The Foundations of Statistics. London: J. Wiley. Slovic, P., Fischhoff, B., and Lichtenstein, S. (1976). Cognitive processes and societal
risk taking. In J. S. Carol and J. W. Payne, eds., Cognition and Social Behavior.
Hillsdale, NJ: Erlbaum. Sperber, D. (1994). The modularity of thought and the epidemiology of representations.
In Hirschfeld and Gelman (1994), 39-67. Stein, E. (1996). Without Good Reason. Oxford: Clarendon Press. Stich, S. (1990). The Fragmentation of Reason. Cambridge, MA: MIT
Press. Tooby, J. and Cosmides, L. (1992). The psychological foundations of culture. In
Barkow, Cosmides and Tooby (1992), 19-136. Tooby, J. and Cosmides, L. (1995). Foreword. In Baron-Cohen (1995), xi-xviii. Tversky, A. and Kahneman, D. (1983). Extensional versus intuitive reasoning: The
conjunction fallacy in probability judgment. Psychological Review, 90,
293-315. Tversky, A. and Kahneman, D. (1982). Judgments of and by representativeness. In
Kahneman, Slovic and Tversky (1982), 84-98. We are … blind not only to the extremes of probability but also to
intermediate probabilities -- from which one might well adduce that we are
blind about probabilities. I would like to suggest a simple, general, probabilistic law: Any
probabilistic intuition by anyone not specifically tutored in
probability calculus has a greater than 50 percent chance of being
wrong. (Piattelli-Palmarini 131 & 132) This is not, however, a claim that any other proponents of heuristics and biases have been
prepared to endorse even in their least careful statements. Nor is there any reason to think
that they should since it is utterly implausible to maintain that this thesis is supported by
the available data. We will, therefore, treat Piattelli-Palmarini's "probabilistic law" as a
particularly extreme instance of rhetorical excess and ignore it in the remainder of the
paper.
In making predictions and judgments under uncertainty, people do not
appear to follow the calculus of chance or the statistical theory of prediction.
Instead, they rely on a limited number of heuristics which sometimes yield
reasonable judgments and sometimes lead to severe and systematic errors.
(Kahneman and Tversky 1973, 48)
3. Some Real Disagreements
Conclusion