Failure To Disagree PDF
Failure To Disagree PDF
Failure To Disagree PDF
A Failure to Disagree
This article reports on an effort to explore the differences still separated in many ways: by divergent attitudes, pref-
between two approaches to intuition and expertise that are erences about facts, and feelings about fighting words such
often viewed as conflicting: heuristics and biases (HB) and as “bias.” If we are to understand the differences between
naturalistic decision making (NDM). Starting from the our respective communities, such emotions must be taken
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
The intellectual traditions that we have traced to deGroot’s There is an obvious difference in the primary form of research
(1946/1978) studies of chess masters (NDM) and to Meehl’s conducted by the respective research communities. The mem-
(1954) research on clinicians (HB) are alive and well today. bers of the HB community are mostly based in academic
They are reflected in the approaches of our respective intel- departments, and they tend to favor well-controlled experi-
lectual communities. In this section we consider three impor- ments in the laboratory. The members of the NDM commu-
tant contrasts between the two approaches: the stance taken by nity are typically practitioners who operate in “real-world”
the NDM and HB researchers toward expert judgment, the use organizations. They have a natural sympathy for the ecolog-
of field versus laboratory settings for decision-making re- ical approach, first popularized in the late 1970s, which ques-
search, and the application of different standards of perfor- tions the relevance of laboratory experiments to real-world
mance, which leads to different conclusions about expertise. situations. NDM researchers use methods such as cognitive
task analysis and field observation to investigate judgments
Stance Regarding Expertise and
and decision making under complex conditions that would be
Decision Algorithms
difficult to recreate in the laboratory.
There is no logical inconsistency between the observations There is no logically necessary connection between these
that inspired the NDM and HB approaches to professional methodological choices and the nature of the hypotheses and
judgment: The intuitive judgments of some professionals models being tested. As the examples of the preceding section
are impressively skilled, while the judgments of other illustrate, the view that heuristics and biases are only studied
professionals are remarkably flawed. Although not contra- and found in the laboratory is a caricature.1 Similarly, the
dictory, these core observations suggest conflicting gener- RPD model could have emerged from the laboratory, and it
alizations about the utility of expert judgment. Members of has been tested there (Johnson & Raab, 2003; Klein, Wolf,
the HB community are of course aware of the existence of Militello, & Zsambok, 1995). In addition, a number of NDM
skill and expertise, but they tend to focus on flaws in researchers have reported studies of the performance of pro-
human cognitive performance. Members of the NDM com- ficient decision makers in realistically simulated environments
munity know that professionals often err, but they tend to (e.g., Smith, Giffin, Rockwell, & Thomas, 1986).
stress the marvels of successful expert performance.
The basic stance of HB researchers, as they consider
experts, is one of skepticism. They are trained to look for 1
Among many other examples, see Slovic (2000) for applications to
opportunities to compare expert performance with perfor- the study of responses to risk; Guthrie, Rachlinski, and Wistrich (2007)
mance by formal models or rules and to expect that experts and Sunstein (2000) for applications in the legal domain; Croskerry and
Norman (2008) for medical judgment; Bazerman (2005) for managerial
will do poorly in such comparisons. They are predisposed judgments and decision making; and Kahneman and Renshon (2007) for
to recommend the replacement of informal judgment by political decision making. The collection assembled by Gilovich, Griffin,
algorithms whenever possible. Researchers in the NDM and Kahneman (2002) includes other examples.
the best practitioners define the standard. As Shanteau The distinction between Systems 1 and 2 plays an
(1992) suggested, “Experts are operationally defined as important role in both the HB and NDM approaches. In the
those who have been recognized within their profession as RPD model, for example, the performance of experts in-
having the necessary skills and abilities to perform at the volves both an automatic process that brings promising
highest level” (p. 255). For example, captains of firefight- solutions to mind and a deliberate activity in which the
ing companies are evaluated not only by their ability to execution of the candidate solution is mentally simulated in
extinguish fires, but also by other criteria, such as the a process of progressive deepening. In the HB approach,
amount of damage created before the fire is controlled. System 2 is involved in the effortful performance of some
When colleagues say, “If Person X had been there instead reasoning and decision-making tasks as well as in the
of Person Y, the fire would not have spread as far,” then continuous monitoring of the quality of reasoning. When
Person X counts as an expert within that organization. The there are cues that an intuitive judgment could be wrong,
use of peer judgments can distinguish highly competent System 2 can impose a different strategy, replacing intu-
decision makers from mediocre ones who may have the ition by careful reasoning.2
same amount of experience and from novices who have The NDM and HB approaches share the assumption
little experience. This level of differentiation is sufficient that intuitive judgments and preferences have the charac-
for most NDM studies. teristics of System 1 activity: They are automatic, arise
In several of the studies that Meehl (1954) reviewed, effortlessly, and often come to mind without immediate
the quality of expert performance was evaluated by com- justification. However, the two approaches focus on differ-
paring the accuracy of decisions made by experts with the ent classes of intuition. Intuitive judgments that arise from
accuracy of optimal linear combinations. If the predictions experience and manifest skill are the province of NDM,
generated by a linear combination of a few variables are which explores the cues that guided such judgments and the
more accurate (in a new sample) than those of a profes- conditions for the acquisition of skill. In contrast, HB
sional who has access to the same information, the perfor- researchers have been mainly concerned with intuitive
mance of the professional is certainly suboptimal. Note that judgments that arise from simplifying heuristics, not from
the optimality criterion is significantly more demanding specific experience. These intuitive judgments are less
than the criteria by which expertise is evaluated in NDM likely to be accurate and are prone to systematic biases.
research. NDM researchers compare the performance of We discuss the two classes of judgment in sequence.
professionals with that of the most successful experts in First, we describe the process of skill acquisition that
their field, whereas HB researchers prefer to compare the supports the intuitive judgments and preferences of genuine
judgments of professionals with the outcome of a model experts. In particular, we explore two necessary conditions
that makes the best possible use of available information. It for the development of skill: high-validity environments
is entirely possible for the predictions of experienced cli- and an adequate opportunity to learn them. Next, we dis-
nicians to be superior to those of novices but inferior to a cuss heuristic-based intuitions and some of the biases to
linear model or an intelligent system. which they are prone. Finally, we address the question of
Sources of Intuition the critique of intuition: How can skilled intuitions be
distinguished from heuristic-based intuitions?
The judgments and decisions that we are most likely to call
intuitive come to mind on their own, without explicit 2
awareness of the evoking cues and of course without an The contrast between System 1 and System 2 has given rise to its
own literature. For example, J. St. B. T. Evans (2007) has asserted that
explicit evaluation of the validity of these cues. The fire- System 1 is affected by the tendency to contextualize problems in the light
fighter feels that the house is very dangerous, the nurse of prior knowledge and belief and that System 2 is affected by the
feels that an infant is ill, and the chess master immediately tendency to satisfice without considering alternatives.
small child recognizes that an animal is a dog, not a cat. It useful notion of wicked environments, in which wrong
may be worth noting that this description of pattern recog- intuitions are likely to develop. His most compelling ex-
nition and the skilled pattern recognition described in the ample (borrowed from Lewis Thomas) is the early 20th-
RPD model are different from the recognition heuristic century physician who frequently had intuitions about pa-
discussed by Goldstein and Gigerenzer (1999), which is a tients in the ward who were about to develop typhoid. He
special-purpose rule of thumb. confirmed his intuitions by palpating these patients’
The recognition model implies two conditions that tongues, but because he did not wash his hands the intui-
must be satisfied for an intuitive judgment (recognition) to tions were disastrously self-fulfilling.
be genuinely skilled: First, the environment must provide High validity does not imply the absence of uncer-
adequately valid cues to the nature of the situation. Second, tainty, and the regularities that are to be discovered are
people must have an opportunity to learn the relevant cues. sometimes statistical. Games such as bridge or poker count
For the first condition, valid cues must be specifiable, at as high-validity situations. The mark of these situations is
least in principle— even if the individual does not know that skill, the ability to identify favorable bets, improves
what they are. The child relies on valid cues to identify a without guaranteeing that every attempt will succeed. The
dog, without any ability to state what the cues are. Simi- challenge of learning bridge and poker is not essentially
larly, the nurse and the firefighter are also guided by valid different from the challenge of learning chess, where the
cues they find in the environment. No magic is involved. A uncertainty arises from the enormous number of possible
crucial conclusion emerges: Skilled intuitions will only developments.
develop in an environment of sufficient regularity, which As the examples of competitive games illustrate, the
provides valid cues to the situation. The ways in which second necessary condition for the development of recog-
skilled judgments take advantage of environmental regu- nition (and of skilled intuition) is an adequate opportunity
larities have been discussed by, among others, Brunswik to learn the relevant cues. It has been estimated that chess
(1957) and Hertwig, Hoffrage, and Martingnon (1999). masters must invest 10,000 hours to acquire their skills
Validity, as we use the term, describes the causal and (Chase & Simon, 1973). Fortunately, most of the skills can
statistical structure of the relevant environment. For exam- be acquired with less practice. A child does not need
ple, it is very likely that there are early indications that a thousands of examples to learn to discriminate dogs from
building is about to collapse in a fire or that an infant will cats. The skilled pediatric nurse has seen a sufficient num-
soon show obvious symptoms of infection. On the other ber of sick infants to recognize subtle signs of disease, and
hand, it is unlikely that there is publicly available informa- the experienced fireground commander has experienced
tion that could be used to predict how well a particular numerous fires and probably imagined many more, during
stock will do—if such valid information existed, the price years of thinking and conversing about firefighting. With-
of the stock would already reflect it. Thus, we have more out these opportunities to learn, a valid intuition can only
reason to trust the intuition of an experienced fireground be due to a lucky accident or to magic—and we do not
commander about the stability of a building, or the intui- believe in magic.
tions of a nurse about an infant, than to trust the intuitions Two conditions must be satisfied for skilled intuition
of a trader about a stock. We can confidently expect that a to develop: an environment of sufficiently high validity and
detailed study of how professionals think is more likely to adequate opportunity to practice the skill. Ericsson, Char-
reveal useful predictive cues in the former cases than in the ness, Hoffman, and Feltovich (2006) have described a
latter. range of factors that influence the rate of skill development.
Determining the validity of an environment is not These include the type of practice people employ, their
always easy. When Tetlock (2005) embarked on his ambi- level of engagement and motivation, and the self-regula-
tious study of long-term forecasts of strategic and eco- tory processes they use. Even when the circumstances are
nomic events by experts, the outcome of his research was favorable, however, some people will develop skilled in-
of three words. The task has a wide range of difficulty: The uate schools (MIT, Princeton, Harvard) to 90% in some-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
item cottage/swiss/cake is easy, but few people can quickly what less selective schools. It can be argued that the setting
find the answer to the item dive/light/rocket—although of this problem is not typical of the challenges that people
everyone recognizes the answer as valid (it is above us and face in the real world, but the phenomenon that Frederick
is blue in good weather; Mednick, 1962). The RAT brings studied is hardly restricted to puzzles. A common genre of
us back to Simon’s observation that the regularities on business literature celebrates successful leaders who made
which intuitions depend are represented in memory. The strategic decisions on the basis of gut feelings and intui-
situation of the RAT has high validity: Widely shared tions that they did not adequately check, but many of these
patterns of associations exist, which everyone can recog- successes owe more to luck than to genius (Rosenzweig,
nize although few can find them without prompting. 2007).
The anchoring phenomenon is another case in which a
Imperfect Intuition bias in the operations of memory causes intuitions to go
astray. Suppose some participants in an experiment are first
We have seen that reliably skilled intuitions are likely to asked “Is the average price of German cars more or less
develop when the individual operates in a high-validity than $100,000?” before they are required to provide a
environment and has an opportunity to learn the rules of numerical estimate of the average cost of German cars.
that environment. These conditions often remain unmet in Other respondents encounter a different anchoring ques-
professional contexts, either because the environment is tion: They are first asked whether the average cost of
insufficiently predictable (as in the long-term forecasting of German cars is more or less than $30,000, and then they are
political events) or because of the absence of opportunities to give an estimate of the average. We can expect the
to learn its rules (as in the case of firefighters exposed to a estimates of the two groups to differ by as much as half the
fire in a skyscraper with unexpected damage to the heat difference between the anchors—in this case the expected
shielding of its structural support). We both agree that most anchoring effect would be $35,000 (Jacowitz & Kahneman,
of the intuitive judgments and decisions that System 1 1995). The mechanism of anchoring is well understood
produces are skilled, appropriate, and eventually success- (Mussweiler & Strack, 2000). The original question with
ful. But we also agree that not all intuitive judgments are the high anchor brings expensive cars to the respondents’
skilled, although our hunches about the frequency of ex- mind: Mercedes, BMWs, Audis. The lower anchor is more
ceptions differ. People, including experienced profession- likely to evoke the image of a beetle and the name Volks-
als, sometimes have subjectively compelling intuitions wagen. The initial question therefore biases the sample of
even when they lack true skill, either because the environ- cars that come to mind when people next attempt to esti-
ment is insufficiently regular or because they have not mate the average price of German cars. The process of
mastered it. Lewis (2003) described the weaknesses in the estimating the average is a deliberate, System 2 operation,
ability of baseball scouts and managers to judge the capa- but the bias occurs in the automatic phase in which in-
bilities, contributions, and potential of players. Despite stances are retrieved from memory. The resulting anchor-
ample opportunities to acquire judgment skill, scouts and ing effect is large and robust. The answers that come to
managers were often insensitive to important variables and mind are typically held with substantial confidence, and the
overly influenced by such factors as the player’s appear- victims of anchoring manipulations confidently deny any
ance—a clear case of prediction by representativeness. effect of the anchor. The common criticism of laboratory
When intuitive judgments do not come from skill, experiments hardly applies here, because large anchoring
where do they come from? This is the question that stu- effects have been demonstrated in the courtroom, in real
dents of heuristics and biases have explored, mostly in estate transactions, and in other real-world contexts.
laboratory experiments. The answer, of course, is that For a final example, consider this question: “Julie is a
incorrect intuitions, like valid ones, also arise from the graduating senior. She read fluently at age 4. What is your
operations of memory. Three phenomena that have been best guess of her GPA [grade point average]?” Most people
Frederick, 2002). livestock judges, astronomers, test pilots, soil judges, chess
This document is copyrighted by the American Psychological Association or one of its allied publishers.
Of course, the mechanisms that produce incorrect masters, physicists, mathematicians, accountants, grain in-
intuitions will only operate in the absence of skill. If people spectors, photo interpreters, and insurance analysts. In con-
have a skilled response to the task with which they are trast, Shanteau noted poor performance by experienced
charged, they will apply their skill. But even in the absence professionals in another large set of occupations: stockbro-
of skill an intuitive response may come to their minds. The kers, clinical psychologists, psychiatrists, college admis-
difficulty is that people have no way to know where their sions officers, court judges, personnel selectors, and intel-
intuitions came from. There is no subjective marker that ligence analysts. Shanteau searched for task characteristics
distinguishes correct intuitions from intuitions that are pro- that distinguished the domains in which experts did well
duced by highly imperfect heuristics. An important char- from those in which experts did poorly. The factors that we
acteristic of intuitive judgments, which they share with identified—the predictability of outcomes, the amount of
experience, and the availability of good feedback—were
perceptual impressions, is that a single response initially
included in his list. In addition, Shanteau pointed to static
comes to mind. Most of the time we have to trust this first
(as opposed to dynamic) stimuli as favorable to good
impulse, and most of the time we are right or are able to
performance.
make the necessary corrections if we turn out to be wrong,
Three professions—nurses, physicians, and audi-
but high subjective confidence is not a good indication of
tors—appeared on both of Shanteau’s (1992) lists. These
validity (Einhorn & Hogarth, 1978). Checking one’s intu-
professionals exhibited genuine expertise in some of their
ition is an effortful operation of System 2, which people do activities but not in others. We refer to such mixed grades
not always perform—sometimes because it is difficult to do for professionals as “fractionated expertise,” and we be-
so and sometimes because they do not bother. lieve that the fractionation of expertise is the rule, not an
Intuitions that originate in heuristics are not necessar- exception. For example, auditors who have expertise in
ily wrong. Indeed, the original statement of the HB ap- “hard” data such as accounts receivable may do much less
proach asserted, “In general these heuristics are quite use- well with “soft” data such as indications of fraud (J. Shan-
ful, but sometimes they lead to severe and systematic teau, personal communication, February 12, 2009).
errors” (Tversky & Kahneman, 1974, p. 1124). The HB There are a few activities, such as chess, in which a
claim is not that intuitions that arise in heuristics are always master will not encounter challenges that are genuinely
incorrect, only that they are less trustworthy than intuitions new. In most domains, however, professionals will occa-
that are rooted in specific experiences. Unfortunately, peo- sionally have to deal with situations and tasks that they
ple are not normally aware of the origins of the thoughts have not had an opportunity to master. Physicians, as is
that come to their minds, and the correlation between the well known, encounter from time to time diagnostic prob-
accuracy of their judgments and the confidence they expe- lems that are entirely new to them—they have expertise in
rience is not consistently high (Arkes, 2001; Griffin & some diagnoses but not in others. Similarly, weather fore-
Tversky, 1992). Subjective confidence is often determined casters are more successful in the routine prediction of
by the internal consistency of the information on which a temperature and precipitation than in forecasting hail
judgment is based, rather than by the quality of that infor- (Stewart, Roebber, & Bosart, 1997).
mation (Einhorn & Hogarth, 1978; Kahneman & Tversky, Characteristically, we came to the topic of fraction-
1973). As a result, evidence that is both redundant and ated expertise with different examples in mind. GK focuses
flimsy tends to produce judgments that are held with too on the experts who perform a constant task (e.g., putting
much confidence. These judgments will be presented too out fires; establishing a diagnosis) but encounter unfamiliar
assertively to others and are likely to be believed more than situations. The ability to recognize that a situation is anom-
they deserve to be. The safe way to evaluate the probable alous and poses a novel challenge is one of the manifesta-
accuracy of a judgment (our own or someone else’s) is by tions of authentic expertise. Descriptions of diagnostic
considering the validity of the environment in which the thinking in medicine emphasize the intuitive ability of
performance of some tasks—typically those that deal with judgments included the following: college academic per-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
the short term— but the feedback they receive from their formance, presence of throat infections, diagnosis of gas-
failures in long-term judgments is delayed, sparse, and trointestinal disorders, length of psychiatric hospitalization,
ambiguous. The experience of the professionals that DK job turnover, suicide attempts, juvenile delinquency, ma-
has thought about is therefore conducive to overconfidence. lingering, and occupational choice.
These professionals may have strong subjective con- Findings in which the performance of human judges is
fidence in their judgments, but we do not believe that inferior to that of simple algorithms are often cited as
subjective confidence reliably indicates whether intuitive evidence of cognitive ineptitude, but this conclusion is
judgments or decisions are valid. When experts recognize unwarranted. The correct conclusion is that people perform
anomalies, using judgments of typicality and familiarity, significantly more poorly than algorithms in low-validity
they are detecting violations of patterns in the external environments. The tasks reviewed by Grove et al. (2000)
situation. In contrast, people do not have a strong ability to
generally involved noisy and/or highly complex situations.
distinguish correct intuitions from faulty ones. People, even
The forecasts made by the algorithms were often wrong,
experts, do not appear to be skilled in detecting patterns in
albeit less often than the clinical predictions. The studies in
the internal situation in order to identify the basis for their
the Meehl paradigm have not produced “smoking gun”
judgments. Therefore, reliance on subjective confidence
demonstrations in which clinicians miss highly valid cues
may contribute to overconfidence.
that the algorithm detects and uses. Indeed, such an out-
The experts that GK has studied seem less susceptible
to overconfidence, perhaps in part because of the direct come would be unlikely, because human learning is nor-
personal risks it poses. Weather forecasters, engineers, and mally quite efficient. Where simple and valid cues exist,
logistics specialists typically resist requests to make judg- humans will find them if they are given sufficient experi-
ments about matters that fall outside their area of compe- ence and enough rapid feedback to do so— except in the
tence. People in professions marked by standard methods, environments that Hogarth (2001) labeled “wicked,” in
clear feedback, and direct consequences for error appear to which the feedback is misleading. A statistical approach
appreciate the boundaries of their expertise. These experts has two crucial advantages over human judgment when
know more knowledgeable experts exist. Weather forecast- available cues are weak and uncertain: Statistical analysis
ers know there are people in another location who better is more likely to identify weakly valid cues, and a predic-
understand the local dynamics. Structural engineers know tion algorithm will maintain above-chance accuracy by
that chemical engineers, or even structural engineers work- using such cues consistently. The meta-analysis performed
ing with different types of models or materials, are the true by Karelaia and Hogarth (2008) showed that consistency
experts who should be consulted. accounted for much of the advantage of algorithms over
As in the other topics that we have considered, we find humans.
no reason to disagree about either fractionation of expertise The evaluation and approval of personal loans by loan
or overconfidence. As usual, different rules apply to dif- officers is an example of a situation in which algorithms
ferent tasks. should be used to replace human judgment. Identifying the
relatively small number of defaulting loans is a low-valid-
Augmenting Professional Judgment: ity task because of the low base rate of the critical outcome.
The Use of Algorithms Algorithms have largely replaced human judges in this
The attitude toward the Meehl paradigm, in which intui- task, using as inputs objective demographic and personal
tions and professional judgments are set in competition, is data rather than subjective impression of reliability. The
a sore point in conversations between adherents of NDM result is an unequivocal improvement: We have fairer loan
and HB. The idea of algorithms that outdo human judges is judgments (i.e., judgments that are not improperly influ-
a source of pride and joy for members of the HB tribe, but enced by gender or race), faster decisions, and reduced
algorithms are usually distrusted by the NDM community. expenses.
in the algorithmic approach, and (e) a low likelihood that arise from genuine skill—the focus of the NDM
This document is copyrighted by the American Psychological Association or one of its allied publishers.
changing conditions will render the algorithm obsolete. We approach— but that they can also arise from inap-
also agree that algorithms that substitute for human judg- propriate application of the heuristic processes on
ment must remain under human supervision, to provide which students of the HB tradition have focused.
continuous monitoring of their performance and of relevant ● Skilled judges are often unaware of the cues that
changes in the environment. Maintaining adequate super- guide them, and individuals whose intuitions are not
vision of algorithms can be difficult, because there is evi- skilled are even less likely to know where their
dence that human operators become more passive and less judgments come from.
vigilant when algorithms are in charge—a phenomenon ● True experts, it is said, know when they don’t know.
that has been labeled “automation bias” (Skitka, Mosier & However, nonexperts (whether or not they think
Burdick, 1999, 2000). they are) certainly do not know when they don’t
We agree that the introduction of algorithms and other know. Subjective confidence is therefore an unreli-
formal decision aids in organizations will often encounter able indication of the validity of intuitive judgments
opposition and unexpected problems of implementation. and decisions.
Few people enjoy being replaced by mechanical devices or ● The determination of whether intuitive judgments
by mathematical algorithms, and many devices and algo- can be trusted requires an examination of the envi-
rithms function less well in the real world than on the ronment in which the judgment is made and of the
planning board (Yates, Veinott, & Patalano, 2003). Even opportunity that the judge has had to learn the
decision aids and procedures that leave the authority of the regularities of that environment.
decision maker intact— decision analysis is a salient exam- ● We describe task environments as “high-validity” if
ple—are often resisted, for both good and bad reasons. there are stable relationships between objectively iden-
Naturally, we have somewhat different attitudes toward tifiable cues and subsequent events or between cues
these problems of implementation, with DK usually view- and the outcomes of possible actions. Medicine and
ing them as obstacles to be overcome and GK seeing them firefighting are practiced in environments of fairly
as reasons to be skeptical about the value of formal methods. high validity. In contrast, outcomes are effectively
Despite our different attitudes toward formal methods, unpredictable in zero-validity environments. To a
we agree on the potential of semi-formal strategies. An good approximation, predictions of the future value of
example is the premortem method (Klein, 2007) for reduc- individual stocks and long-term forecasts of political
ing overconfidence and improving decisions. Project teams events are made in a zero-validity environment.
using this method start by describing their plan. Next they ● Validity and uncertainty are not incompatible. Some
imagine that their plan has failed and the project has been environments are both highly valid and substan-
a disaster. Their task is to write down, in two minutes, all tially uncertain. Poker and warfare are examples.
the reasons why the project failed. The facilitator goes The best moves in such situations reliably increase
around the table, getting reasons from each of the team the potential for success.
members, starting with the leader. The rationale for the ● An environment of high validity is a necessary
method is the concept of prospective hindsight (Mitchell, condition for the development of skilled intuitions.
Russo, & Pennington, 1989)—that people can generate Other necessary conditions include adequate oppor-
more criticisms when they are told that an outcome is tunities for learning the environment (prolonged
certain. It also offers a solution to one of the major prob- practice and feedback that is both rapid and un-
lems of decision making within organizations: the gradual equivocal). If an environment provides valid cues
suppression of dissenting opinions, doubts, and objections, and good feedback, skill and expert intuition will
which is typically observed as an organization commits eventually develop in individuals of sufficient talent.
itself to a major plan. The premortem method is consistent ● Although true skill cannot develop in irregular or
with the HB concern for overconfidence while drawing on unpredictable environments, individuals will some-
determine the boundaries of their true expertise. deGroot, A. D. (1978). Thought and choice in chess. The Hague: Mouton.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
● We agree that the weak regularities available in (Original work published 1946)
low-validity situations can sometimes support the Einhorn, H. J., & Hogarth, R. M. (1978). Confidence in judgment:
development of algorithms that do better than Persistence of the illusion of validity. Psychological Review, 85, 395–
chance. These algorithms only achieve limited ac- 416.
Ericsson, K. A. (2006). The influence of experience and deliberate prac-
curacy, but they outperform humans because of tice on the development of superior expert performance. In K. A.
their advantage of consistency. However, the intro- Ericsson, N. Charness, R. R. Hoffman, & P. J. Feltovich (Eds.), The
duction of algorithms to replace human judgment is Cambridge handbook of expertise and expert performance (pp. 39 – 68).
likely to evoke substantial resistance and sometimes New York: Cambridge University Press.
Ericsson, K. A., Charness, N., Hoffman, R. R., & Feltovich, P. J. (Eds.).
has undesirable side effects. (2006). The Cambridge handbook of expertise and expert performance.
New York: Cambridge University Press.
Another conclusion that we both accept is that the Evans, J. St. B. T. (2007). Hypothetical thinking: Dual processes in
approaches of our respective communities have built-in reasoning and judgment. Hove, East Sussex, England: Psychology
limitations. For historical and methodological reasons, HB Press.
researchers generally find errors more interesting and in- Evans, J. St. B. T., & Frankish, K. (Eds.). (2009). In two minds: Dual
structive than correct performance; but a psychology of processes and beyond. New York: Oxford University Press.
Fogarty, W. M. (1988). Formal investigation into the circumstances
judgment and decision making that ignores intuitive skill is surrounding the downing of a commercial airliner by the U. S. S.
seriously blinkered. Because their intellectual attitudes de- Vincennes (CG 49) on 3 July 1988 [Unclassified Letter Ser. 1320 of 28
veloped in reaction to the HB tradition, members of the July 1988, to Commander in Chief, U.S. Central Command]. Washing-
NDM community have an aversion to the word bias and to ton, DC: U.S. Department of the Navy.
Frederick, S. (2005). Cognitive reflection and decision making. Journal of
the corresponding concept; but a psychology of profes- Economic Perspectives, 19(4), 25– 42.
sional judgment that neglects predictable errors cannot be Gawande, A. (2002). Complications: A surgeon’s notes on an imperfect
adequate. Although we agree with both of these conclu- science. London: Profile Books.
sions, we have yet to move much beyond recognition of the Gilovich, T., Griffin, D., & Kahneman, D. (Eds.). (2002). Heuristics and
problem. DK is still fascinated by persistent errors, and GK biases: The psychology of intuitive judgment. New York: Cambridge
University Press.
still recoils when biases are mentioned. We hope, however, Goldberg, L. R. (1970). Man versus model of man: A rationale, plus some
that our effort may help others do more than we have been evidence, for a method of improving on clinical inferences. Psycholog-
able to do in bringing the insights of both communities to ical Bulletin, 73, 422– 432.
bear on their common subject. Goldstein, D. G., & Gigerenzer, G. (1999). The recognition heuristic:
How ignorance makes us smart. In G. Gigerenzer, P. M. Todd, &
A. B. C. Research Group (Eds.), Simple heuristics that make us smart
REFERENCES (pp. 37–58). New York: Oxford University Press.
Griffin, D. W., & Tversky, A. (1992). The weighing of evidence and the
Arkes, H. R. (2001). Overconfidence in judgmental forecasting. In J. S. determinants of confidence. Cognitive Psychology, 24, 411– 435.
Armstrong (Ed.), Principles of forecasting: A handbook for researchers Groopman, J. (2007). How doctors think. New York: Houghton Mifflin.
and practitioners (pp. 495–516). Boston: Kluwer Academic. Grove, W. M., Zald, D. H., Lebow, B. S., Snitz, B. E., & Nelson, C.
Bazerman, M. H. (2005). Judgment in managerial decision making (6th (2000). Clinical versus mechanical prediction: A meta-analysis. Psy-
ed.). Hoboken, NJ: Wiley. chological Assessment, 12, 19 –30.
Beach, L. R. (1990). Image theory: Decision making in personal and Guthrie, C., Rachlinski, J. J., & Wistrich, A. J. (2007). Blinking on the
organizational contexts. Chichester, England: Wiley. bench: How judges decide cases. Cornell Law Review, 93, 1– 43.
Brunswik, E. (1957). Scope and aspects of the cognitive problem. In H. Hammond, K. R., Hamm, R. M., Grassia, J., & Pearson, T. (1987). Direct
Gruber, K. R. Hammond, & R. Jessor (Eds.), Contemporary ap- comparison of the efficacy of intuitive and analytical cognition in
proaches to cognition (pp. 5–31). Cambridge, MA: Harvard University expert judgment. IEEE Transactions on Systems, Man, and Cybernet-
Press. ics, SMC-17(5), 753–770.
Cannon-Bowers, J. A., & Salas, E. (Eds.). (1998). Making decisions under Hertwig, R., Hoffrage, U., & Martingnon, L. (1999). Quick estimation:
stress: Implications for individual and team training. Washington, DC: Letting the environment do the work. In G. Gigerenzer, P. M. Todd, &
American Psychological Association. A. B. C. Research Group (Eds.), Simple heuristics that make us smart
Chase, W. G., & Simon, H. A. (1973). The mind’s eye in chess. In W. G. (pp. 37–58). New York: Oxford University Press.
Psychological Review, 80, 237–251. Rosenzweig, P. (2007). The halo effect . . . and the eight other business
This document is copyrighted by the American Psychological Association or one of its allied publishers.
Karelaia, N., & Hogarth, R. M. (2008). Determinants of linear judgment: delusions that deceive managers. New York: Free Press.
A meta-analysis of lens model studies. Psychological Bulletin, 134, Schraagen, J. M. C., Chipman, S. F., & Shalin, V. J. (Eds.). (2000).
404 – 426. Cognitive task analysis. Mahwah, NJ: Erlbaum.
Klein, G. A. (1993). A recognition-primed decision (RPD) model of rapid Shanteau, J. (1992). Competence in experts: The role of task character-
decision making. In G. A. Klein, J. Orasanu, R. Calderwood, & C. E. istics. Organizational Behavior and Human Decision Processes, 53,
Zsambok (Eds.), Decision making in action: Models and methods (pp. 252–262.
138 –147). Norwood, NJ: Ablex. Simon, H. A. (1992). What is an explanation of behavior? Psychological
Klein, G. (1998). Sources of power: How people make decisions. Cam- Science, 3, 150 –161.
bridge, MA: MIT Press. Skitka, L. J., Mosier, K., & Burdick, M. D. (1999). Does automation bias
Klein, G. (2007, September). Performing a project premortem. Harvard decision-making? International Journal of Human–Computer Studies,
Business Review, pp. 18 –19. 51, 991–1006.
Klein, G. A., Calderwood, R., & Clinton-Cirocco, A. (1986). Rapid Skitka, L. J., Mosier, K., & Burdick, M. D. (2000). Accountability and
decision making on the fireground. In Proceedings of the Human automation bias. International Journal of Human–Computer Studies,
Factors and Ergonomics Society 30th Annual Meeting (Vol. 1, pp. 52, 701–717.
576 –580). Norwood, NJ: Ablex. Slovic, P. (Ed.). (2000). The perception of risk. London: Earthscan.
Klein, G. A., Orasanu, J., Calderwood, R., & Zsambok, C. E. (1993). Smith, P. J., Giffin, W. C., Rockwell, T. H., & Thomas, M. (1986).
Decision making in action: Models and methods. Norwood, NJ: Ablex. Modeling fault diagnosis as the activation and use of a frame system.
Klein, G., Wolf, S., Militello, L., & Zsambok, C. (1995). Characteristics Human Factors, 28, 703–716.
of skilled option generation in chess. Organizational Behavior and Stewart, T. R., Roebber, P. J., & Bosart, L. F. (1997). The importance of
Human Decision Processes, 62, 63– 69. the task in analyzing expert judgment. Organizational Behavior and
Lewis, M. (2003). Moneyball: The art of winning an unfair game. New Human Decision Processes, 69, 205–219.
York: Norton. Sunstein, C. R. (Ed.). (2000). Behavioral law and economics. New York:
Lipshitz, R. (1993). Converging themes in the study of decision making in Cambridge University Press.
realistic settings. In G. A. Klein, J. Orasanu, R. Calderwood, & C. E. Tetlock, P. E. (2005). Expert political judgment: How good is it? How can
Zsambok (Eds.), Decision making in action: Models and methods (pp. we know? Princeton, NJ: Princeton University Press.
103–137). Norwood, NJ: Ablex. Tversky, A., & Kahneman, D. (1971). Belief in the law of small numbers.
Mednick, S. A. (1962). The associative basis of the creative process. Psychological Bulletin, 76, 105–110.
Psychological Review, 69, 220 –232. Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heu-
Meehl, P. E. (1954). Clinical vs. statistical prediction: A theoretical ristics and biases. Science, 185, 1124 –1131.
analysis and a review of the evidence. Minneapolis: University of Woods, D. D., O’Brien, J., & Hanes, L. F. (1987). Human factors
Minnesota Press. challenges in process control: The case of nuclear power plants. In G.
Meehl, P. E. (1973). Why I do not attend case conferences. In P. E. Meehl Salvendy (Ed.), Handbook of human factors/ergonomics (pp. 1724 –
(Ed.), Psychodiagnosis: Selected papers (pp. 225–302). Minneapolis: 1770). New York: Wiley.
University of Minnesota Press. Yates, J. F., Veinott, E. S., & Patalano, A. L. (2003). Hard decisions, bad
Mitchell, D., Russo, J., & Pennington, N. (1989). Back to the future: decisions: On decision quality and decision aiding. In S. L. Schneider
Temporal perspective in the explanation of events. Journal of Behav- & J. C. Shanteau (Eds.), Emerging perspectives on judgment and
ioral Decision Making, 2, 25–38. decision research (pp. 13– 63). New York: Cambridge University Press.