Which of the following is not one of the conditions that leads to the impulse to use heuristics group of answer choices?

1. Introduction

Pioneering the modern literature on heuristics in cognition,

Amos Tversky and Daniel Kahneman contended that “peo-

ple rely on a limited number of heuristic principles which

reduce the complex tasks of assessing probabilities and pre-

dicting values to simpler judgmental operations” (Tversky

& Kahneman 1974, p. 1124). Intense controversy has de-

veloped over the virtues and vices of such heuristics, most

of them “fast and frugal,” that play a role in many areas (see

Gigerenzer et al. 1999; Gilovich et al. 2002). But the rele-

vant literature has only started to investigate the possibility

that in the moral and political domains, people also rely on

simple rules of thumb that often work well but that some-

times misfire (see Baron 1994a; 1998; Messick 1993). In

fact, the central point seems obvious. Much of everyday

morality consists of simple, highly intuitive rules that gen-

erally make sense, but that fail in certain cases. It is wrong

to lie or steal, but if a lie or a theft would save a human life,

lying or stealing is probably obligatory. Not all promises

should be kept. It is wrong to try to get out of a longstand-

ing professional commitment at the last minute, but if your

child is in the hospital, you may be morally required to do

exactly that.

One of my major goals in this article is to show that

heuristics play a pervasive role in moral, political, and legal

judgments, and that they sometimes produce significant

mistakes. I also attempt to identify a set of heuristics that

now influence both law and policy, and try to make plausi-

ble the claim that some widely held practices and beliefs are

a product of those heuristics. Often moral heuristics repre-

sent generalizations from a range of problems for which

they are indeed well-suited (see Baron 1994a), and hence,

most of the time, such heuristics work well. The problem

comes when the generalizations are wrenched out of con-

text and treated as freestanding or universal principles, ap-

plicable to situations in which their justifications no longer

operate. Because the generalizations are treated as free-

standing or universal, their application seems obvious, and

those who reject them appear morally obtuse, possibly even

monstrous. I contend that the appearance is misleading and

even productive of moral mistakes. There is nothing obtuse,

or monstrous, about refusing to apply a generalization in

contexts in which its rationale is absent.

Because Kahneman and Tversky were dealing with facts

and elementary logic, they could demonstrate that the

heuristics sometimes lead to errors. Unfortunately, that

cannot easily be demonstrated here. In the moral and po-

litical domains, it is hard to come up with unambiguous

cases in which the error is both highly intuitive and on re-

flection uncontroversial – that is, cases in which people can

ultimately be embarrassed about their own intuitions.

Nonetheless, I hope to show that whatever one’s moral

commitments, moral heuristics exist and indeed are om-

nipresent. We should not treat the underlying moral intu-

itions as fixed points for analysis, rather than as unreliable

and potentially erroneous. In the search for reflective equi-

librium, understood to be coherence among our judgments

at all levels of generality (Rawls 1971), it is important to see

BEHAVIORAL AND BRAIN SCIENCES (2005) 28, 531–573

Printed in the United States of America

© 2005 Cambridge University Press 0140-525X/05 $12.50

531

Moral heuristics

Cass R. Sunstein

University of Chicago Law School and Department of Political Science,

University of Chicago, Chicago, IL 60637.

http://www.law.uchicago.edu/faculty/sunstein/

Abstract: With respect to questions of fact, people use heuristics – mental short-cuts, or rules of thumb, that generally work well, but that

also lead to systematic errors. People use moral heuristics too – moral short-cuts, or rules of thumb, that lead to mistaken and even ab-

surd moral judgments. These judgments are highly relevant not only to morality, but to law and politics as well. Examples are given from

a number of domains, including risk regulation, punishment, reproduction and sexuality, and the act/omission distinction. In all of these

contexts, rapid, intuitive judgments make a great deal of sense, but sometimes produce moral mistakes that are replicated in law and pol-

icy. One implication is that moral assessments ought not to be made by appealing to intuitions about exotic cases and problems; those

intuitions are particularly unlikely to be reliable. Another implication is that some deeply held moral judgments are unsound if they are

products of moral heuristics. The idea of error-prone heuristics is especially controversial in the moral domain, where agreement on the

correct answer may be hard to elicit; but in many contexts, heuristics are at work and they do real damage. Moral framing effects, in-

cluding those in the context of obligations to future generations, are also discussed.

Cass R. Sunstein is Karl N. Llewellyn Distinguished

Service Professor, Law School and Department of Po-

litical Science, University of Chicago. His many books

include Laws of Fear (2005), Why Societies Need Dis-

sent (2003), and Risk and Reason (2002). Much of his

research explores the intersection of psychology, eco-

nomics, and law; he has been particularly interested in

the role of cognitive errors and social influences on ju-

ries, legislators, judges, and those who are the objects of

law’s commands. He is now working on a book on the

production of social knowledge, with special reference

to information aggregation across persons.

that some of our deeply held moral beliefs might be prod-

ucts of heuristics that sometimes produce mistakes.

If moral heuristics are in fact pervasive, then people with

diverse foundational commitments should be able to agree,

not that their own preferred theories are wrong, but that

they are often applied in a way that reflects the use of

heuristics. Utilitarians ought to be able to identify heuris-

tics for the maximization of utility; deontologists should be

able to point to heuristics for the proper discharge of moral

responsibilities; and those uncommitted to any large-scale

theory should be able to specify heuristics for their own

more modest normative commitments. And if moral

heuristics exist, blunders are highly likely not only in moral

thinking, but in legal and political practice as well. Con-

ventional legal and political arguments are often a product

of heuristics masquerading as universal truths. Hence, I will

identify a set of political and legal judgments that are best

understood as a product of heuristics, and that are often

taken, wrongly and damagingly, as a guide to political and

legal practice, even when their rationale does not apply.

2. Ordinary heuristics and an insistent

homunculus

2.1. Heuristics and facts

The classic work on heuristics and biases deals not with

moral questions but with issues of fact. In answering hard

factual questions, those who lack accurate information use

simple rules of thumb. How many words, in four pages of a

novel, will have “ing” as the last three letters? How many

words, in the same four pages, will have “n” as the second-

to-last letter? Most people will give a higher number in re-

sponse to the first question than in response to the second

(Tversky & Kahneman 1984) – even though a moment’s re-

flection shows that this is a mistake. People err because they

use an identifiable heuristic – the availability heuristic – to

answer difficult questions about probability. When people

use this heuristic, they answer a question of probability by

asking whether examples come readily to mind. How likely

is a flood, an airplane crash, a traffic jam, a terrorist attack,

or a disaster at a nuclear power plant? Lacking statistical

knowledge, people try to think of illustrations. For those

without statistical knowledge, it is far from irrational to use

the availability heuristic; the problem is that this heuristic

can lead to serious errors of fact, in the form of excessive

fear of small risks and neglect of large ones.

Or consider the representativeness heuristic, in accor-

dance with which judgments of probability are influenced

by assessments of resemblance (the extent to which A

“looks like” B). The representativeness heuristic is famously

exemplified by people’s answers to questions about the

likely career of a hypothetical woman named Linda, de-

scribed as follows: “Linda is 31 years old, single, outspoken,

and very bright. She majored in philosophy. As a student,

she was deeply concerned with issues of discrimination and

social justice and also participated in antinuclear demon-

strations” (see Kahneman & Frederick 2002; Mellers et al.

2001). People were asked to rank, in order of probability,

eight possible futures for Linda. Six of these were fillers

(such as psychiatric social worker, elementary school

teacher); the two crucial ones were “bank teller” and “bank

teller and active in the feminist movement.”

More people said that Linda was less likely to be a bank

teller than to be a bank teller and active in the feminist

movement. This is an obvious mistake, a conjunction error,

in which characteristics A and B are thought to be more

likely than characteristic A alone. The error stems from the

representativeness heuristic: Linda’s description seems to

match “bank teller and active in the feminist movement” far

better than “bank teller.” In an illuminating reflection on

the example, Stephen Jay Gould observed that “I know [the

right answer], yet a little homunculus in my head continues

to jump up and down, shouting at me – ‘but she can’t just

be a bank teller; read the description’” (Gould 1991,

p. 469). Because Gould’s homunculus is especially inclined

to squawk in the moral domain, I shall return to him on sev-

eral occasions.

Of course, the early work on heuristics has been subject

to intense criticism, sometimes with the claim that in the

real world, most heuristics work quite well (Gigerenzer

2000; Gigerenzer et al. 1999). In this view, many findings of

cognitive errors are an artifact of the laboratory setting and

of clever experimental designs involving unfamiliar prob-

lems; in the real world, people may be much less likely to

err, perhaps because the heuristics are sensible, perhaps

because they are not applied indiscriminately (Krueger &

Funder 2004; cf. Kahneman & Tversky 1996). In addition,

some of the key ideas, including availability and represen-

tativeness, have been challenged as inadequately specified

and as subject to ad hoc applications (Gigerenzer 1996; cf.

Kahneman & Tversky 1996, p. 591). For present purposes,

it is unnecessary to resolve these debates here. No one de-

nies that with respect to facts, human beings use simple

rules of thumb that can produce serious mistakes; even the

“recognition heuristic,” said to enable people to make re-

markably accurate judgments about the size of cities or

prospects in sports events (Goldstein & Gigerenzer 2002),

will produce severe and predictable errors. If this is true for

facts, it is highly likely to be true for political and moral

judgments as well.

With respect to moral heuristics, existing work is sugges-

tive rather than definitive; a great deal of progress remains

to be made, above all through additional experimental work

on moral judgments. Some of the moral heuristics that I

shall identify might reasonably be challenged as subject to

ad hoc rather than predictable application. One of my pri-

mary hopes is to help stimulate further research, testing

when and whether people use moral heuristics that pro-

duce sense or nonsense in particular cases.

2.2. Attribute substitution and prototypical cases

What is a heuristic? Kahneman and Frederick have recently

suggested that heuristics are mental shortcuts used when

people are interested in assessing a “target attribute” and

when they substitute a “heuristic attribute” of the object,

which is easier to handle (Kahneman & Frederick 2002).

Heuristics therefore operate through a process of attribute

substitution. The use of heuristics gives rise to intuitions

about what is true, and these intuitions sometimes are bi-

ased, in the sense that they produce errors in a predictable

direction. Consider the question of whether more people

die from suicides or homicides. Lacking statistical informa-

tion, people might respond by asking whether it is easier to

recall cases in either class (the availability heuristic). The

approach is hardly senseless, but it might also lead to errors,

a result of “availability bias” in the domain of risk percep-

Sunstein: Moral heuristics

532 BEHAVIORAL AND BRAIN SCIENCES (2005) 28:4

tion (see Kuran & Sunstein 1999). For the size of cities, the

recognition heuristic, which is a close cousin of the avail-

ability heuristic, has the same problem, leading to what

might be called “recognition bias.” Sometimes heuristics

are linked to affect, and indeed affect has even been seen

as a heuristic (Slovic et al. 2002); but attribute substitution

is often used for factual questions that lack an affective

component.

Similar mechanisms are at work in the moral, political,

and legal domains. Unsure of what to think or do about a tar-

get attribute (what morality requires, what the law is), peo-

ple might substitute a heuristic attribute instead – asking,

for example, about the view of trusted authorities (a leader

of the preferred political party, an especially wise judge, or

a religious figure). Every law professor in the United States

knows that in approaching difficult constitutional questions,

many law students are drawn to a kind of “Justice Antonin

Scalia heuristic.” If students are unsure how to analyze a

constitutional problem, they might ask instead what Justice

Scalia (an influential conservative on the United States

Supreme Court) thinks – and either follow him or do the op-

posite. In the areas of morality, politics, and law, attribute

substitution is pervasively involved. Often the process works

by appeal to prototypical cases. Confronted by a novel and

difficult problem, observers often ask whether it shares fea-

tures with a familiar problem. If it seems to do so, then the

solution to the familiar problem is applied to the novel and

difficult one. Of course it is possible that in the domain of

values as well as facts, real-world heuristics generally per-

form well in the real world – so that moral errors are re-

duced, not increased, by their use, at least compared to the

most likely alternatives (see my remarks on rule-utilitarian-

ism later). The only claim here is that some of the time, our

moral judgments can be shown to misfire.

The principal heuristics should be seen in light of dual-

process theories of cognition (Kahneman & Frederick

2002). Those theories distinguish between two families of

cognitive operations, sometimes labeled System I and Sys-

tem II. System I is intuitive; it is rapid, automatic, and ef-

fortless (and it features Gould’s homunculus). System II, by

contrast, is reflective; it is slower, self-aware, calculative,

and deductive. System I proposes quick answers to prob-

lems of judgment, and System II operates as a monitor, con-

firming or overriding those judgments. Consider, for exam-

ple, someone who is flying from New York to London in the

month after an airplane crash. This person might make a

rapid, barely conscious judgment, rooted in System I, that

the flight is quite risky; but there might well be a System II

override, bringing a more realistic assessment to bear. Sys-

tem I often has an affective component, but it need not; for

example, a probability judgment might be made quite

rapidly and without much affect at all.

There is growing evidence that people often make auto-

matic, largely unreflective moral judgments, for which they

are sometimes unable to give good reasons (see Greene &

Haidt 2002; Haidt 2001; cf. Pizarro & Bloom 2003). Moral,

political, or legal judgments often substitute a heuristic at-

tribute for a target attribute; System I is operative here as

well, and it may or may not be subject to System II over-

ride. Consider the incest taboo. People have moral revul-

sion against incest even in circumstances in which the

grounds for that taboo seem to be absent; they are subject

to “moral dumbfounding” (Haidt et al. 2004), that is, an in-

ability to give an account for a firmly held intuition. It is

plausible, at least, to think that System I is driving their

judgments, without System II correction. The same is true

in legal and political contexts as well.

3. Heuristics and morality

To show that heuristics operate in the moral domain, we

have to specify some benchmark by which we can measure

moral truth. On these questions I want to avoid any espe-

cially controversial claims. Whatever’s one view of the foun-

dations of moral and political judgments, I suggest, moral

heuristics are likely to be at work in practice. In this section

I begin with a brief account of the possible relationship be-

tween ambitious theories (understood as large-scale ac-

counts of the right or the good) and moral heuristics. I sug-

gest that for those who accept ambitious theories about

morality or politics, it is tempting to argue that alternative

positions are mere heuristics; but this approach is un-

promising, simply because any ambitious theory is likely to

be too contentious to serve as the benchmark for measur-

ing moral truth. Progress is best made, not by opposing

(supposedly correct) ambitious theories to (supposedly

blundering) common sense morality, but in two more mod-

est ways: first, by showing that moral heuristics are at work

in any view about what morality requires; and second, by

showing that such heuristics are at work in a minimally con-

tentious view about what morality requires. I will identify a

number of such heuristics in Section 4.

3.1. Theories, heuristics, and ambitious starts

Many utilitarians, including John Stuart Mill and Henry

Sidgwick, argue that ordinary morality is based on simple

rules of thumb that generally promote utility but that some-

times misfire (see Mill 1861/1971, pp. 28–29; Sidgwick

1907/1981, pp. 199–216; cf. Hare 1981; Smart 1973). For

example, Mill emphasizes that human beings “have been

learning by experience the tendencies of experience,” so

that the “corollaries from the principle of utility” are being

progressively captured by ordinary morality (Mill 1861/

1971, p. 29).

1

Is ordinary morality a series of heuristics for

what really matters, which is utility?

With the aid of modern psychological findings, utilitari-

ans might be tempted to make exactly this argument (see

Baron 1998). They might contend that ordinary moral com-

mitments are a set of mental shortcuts that generally work

well, but that also produce severe and systematic errors

from the utilitarian point of view. Suppose that most peo-

ple reject utilitarian approaches to punishment and are in-

stead committed to retributivism; this is their preferred

theory. Are they responding to System I? Might they be

making a cognitive error? (Is Kantianism a series of cogni-

tive errors?) Note that with respect to what morality re-

quires, utilitarians frequently agree with their deontologi-

cal adversaries about concrete cases; they can join in

accepting the basic rules of criminal and civil law. When de-

ontologists and others depart from utilitarian principles,

perhaps they are operating on the basis of heuristics that

usually work well but sometimes misfire.

But it is exceedingly difficult to settle large-scale ethical

debates in this way. In the case of many ordinary heuristics,

based on availability and representativeness, a check of the

facts, or of the elementary rules of logic, will show that peo-

Sunstein: Moral heuristics

BEHAVIORAL AND BRAIN SCIENCES (2005) 28:4 533

ple err. In the moral domain, this is much harder to demon-

strate. To say the least, those who reject utilitarianism are

not easily embarrassed by a demonstration that their moral

judgments can lead to reductions in utility. For example,

utilitarianism is widely challenged by those who insist on the

importance of distributional considerations. It is far from

clear that a moderate utility loss to those at the bottom can

be justified by a larger utility gain for many at the top (Rawls

1971; see also Nussbaum 1984; Sen 1980–1981).

Emphasizing the existence of moral heuristics, those who

reject utilitarianism might well turn the tables on their utili-

tarian opponents. They might contend that the rules recom-

mended by utilitarians are consistent, much of the time, with

what morality requires – but also that utilitarianism, taken

seriously, produces serious mistakes in some cases. In this

view, utilitarianism is itself a heuristic, one that usually works

well but leads to systematic errors. And indeed, many de-

bates between utilitarians and their critics involve claims, by

one or another side, that the opposing view usually produces

good results, but also leads to severe mistakes and should be

rejected for that reason (see Smart & Williams 1973).

These large debates are not easy to resolve, simply be-

cause utilitarians and deontologists are most unlikely to be

convinced by the suggestion that their defining commit-

ments are mere heuristics. Here, there is a large difference

between moral heuristics and the heuristics uncovered in

the relevant psychological work, in which the facts or sim-

ple logic provide a good test of whether people have erred.

If people tend to think that more words, in a given space,

end with the letters “ing” than have “n” in the next-to-last

position, something has clearly gone wrong. If people think

that some person Linda is more likely to be “a bank teller

who is active in the feminist movement” than a “bank

teller,” there is an evident problem. If citizens of France or

Germany think that New York University is more likely to

have a good basketball team than St. Joseph’s University,

because they have not heard of the latter, then a simple ex-

amination of the record will show that they are wrong. In

the moral domain, factual blunders and simple logic do not

provide such a simple test.

3.2. Neutral benchmarks and weak consequentialism

My goal here is, therefore, not to show that common sense

morality is a series of heuristics for the correct general the-

ory (as suggested by Sidgwick and Mill), but more cau-

tiously, that in many particular cases, moral heuristics are at

work – and that this point can be accepted by people with

diverse general theories, or with grave uncertainty about

which general theory is correct. In the cases catalogued in

Section 5, I contend that it is possible to conclude that a

moral heuristic is at work without accepting any especially

controversial normative claims. In several of the examples,

that claim can be accepted without accepting any con-

testable normative theory at all. Other examples will re-

quire acceptance of what I shall call “weak consequential-

ism,” in accordance with which the social consequences of

the legal system are relevant, other things being equal, to

what law ought to be doing.

Weak consequentialists need not be utilitarians; they do

not have to believe that law and policy should attempt to

maximize utility. They might agree that violations of rights

count among the consequences that ought to matter, so that

deontological considerations play a role in the overall as-

sessment of what should be done. Consider Amartya Sen’s

frequent insistence that consequentialists can insist that

consequences count without accepting utilitarianism and

without denying that violations of rights are part of the set

of relevant consequences (see Sen 1982; 1985). Thus Sen

urges an approach that “shares with utilitarianism a conse-

quentialist approach (but differs from it in not confining at-

tention to utility consequences only)” while also attaching

“intrinsic importance to rights (but...not giving them

complete priority irrespective of other consequences)”

(Sen 1996, p. 1038). Weak consequentialism is in line with

this approach. In evaluating decisions and social states,

weak consequentialists might well be willing to give a great

deal of weight to nonconsequentialist considerations.

Of course, some deontologists will reject any form of con-

sequentialism altogether. They might believe, for example,

that retribution is the proper theory of punishment, and

that the consequences of punishment are never relevant to

the proper level of punishment. Some of my examples will

be unpersuasive to deontologists who believe that conse-

quences do not matter at all. But weak consequentialism

seems to me sufficiently nonsectarian, and attractive to suf-

ficiently diverse people, to make plausible the idea that in

the cases at hand, moral heuristics are playing a significant

role. And for those who reject weak consequentialism, it

might nonetheless be productive to ask whether, from their

own point of view, certain rules of morality and law are re-

flective of heuristics that sometimes produce serious errors.

3.3. Evolution and rule-utilitarianism: Simple heuristics

that make us good?

Two clarifications before we proceed. First, some moral

heuristics might well have an evolutionary foundation (de

Waal 1996; Katz 2000; Sober & Wilson 1999). Perhaps

natural selection accounts for automatic moral revulsion

against incest or cannibalism, even if clever experiments,

or life, can produce situations in which the revulsion is

groundless. In the case of incest, the point is straightfor-

ward: The automatic revulsion might be far more useful,

from the evolutionary perspective, than a more fine-

grained evaluation of contexts (Stein 2001). In fact, an evo-

lutionary account might be provided for most of the heuris-

tics that I explore here. When someone has committed a

harmful act, evolutionary pressures might well have incul-

cated a sharp sense of outrage and a propensity to react in

proportion to it. As a response to wrongdoing, use of an out-

rage heuristic might well be much better than an attempt

at any kind of consequentialist calculus, weak or strong. Of

course, many moral commitments are a product not of evo-

lution but of social learning and even cascade effects (see

Sunstein 2003); individuals in a relevant society will in-

evitably be affected by a widespread belief that it is wrong

to tamper with nature (discussed later), and evolutionary

pressures need not have any role at all.

Second, and related, some or even most moral heuristics

might have a rule-utilitarian or rule-consequentialist de-

fense (see Hooker 2000). The reason is that in most cases

they work well despite their simplicity, and if people at-

tempted a more fine-grained assessment of the moral issues

involved, they might make more moral mistakes rather than

fewer (especially because their self-interest is frequently at

stake). Simple but somewhat crude moral principles might

lead to less frequent and less severe moral errors than com-

Sunstein: Moral heuristics

534 BEHAVIORAL AND BRAIN SCIENCES (2005) 28:4

plex and fine-grained moral principles. Compare the avail-

ability heuristic. Much of the time, use of that heuristic pro-

duces speedy judgments that are fairly accurate, and those

who attempt a statistical analysis might make more errors

(and waste a lot of time in the process). If human beings use

“simple heuristics that make us smart” (Gigerenzer et al.

1999), then they might also use “simple heuristics that make

us good.” I will offer some examples in which moral heuris-

tics seem to me to produce significant errors for law and pol-

icy, but I do not contend that we would be better off with-

out them. On the contrary, such heuristics might well

produce better results, from the moral point of view, than

the feasible alternatives – a possibility to which I will return.

4. The Asian disease problem and moral framing

In a finding closely related to their work on heuristics, Kah-

neman and Tversky find “moral framing” in the context of

what has become known as “the Asian disease problem”

(Kahneman & Tversky 1984). Framing effects do not in-

volve heuristics, but because they raise obvious questions

about the rationality of moral intuitions, they provide a valu-

able backdrop. Here is the first component of the problem:

Imagine that the U.S. is preparing for the outbreak of an

unusual Asian disease, which is expected to kill 600 people.

Two alternative programs to combat the disease have been

proposed. Assume that the exact scientific estimates of the

consequences are as follows:

If Program A is adopted, 200 people will be saved.

If Program B is adopted, there is a one-third probability

that 600 people will be saved and a two-thirds probability

that no people will be saved.

Which of the two programs would you favor?

Most people choose Program A.

But now consider the second component of the problem,

in which the same situation is given, but followed by this de-

scription of the alternative programs:

If Program C is adopted, 400 people will die.

If Program D is adopted, there is a one-third probability

that nobody will die and a two-thirds probability that 600

people will die.

Most people choose Problem D. But a moment’s re-

flection should be sufficient to show that Program A and

Program C are identical, and so too for Program B and Pro-

gram D. These are merely different descriptions of the

same programs. The purely semantic shift in framing is suf-

ficient to produce different outcomes. Apparently, people’s

moral judgments about appropriate programs depend on

whether the results are described in terms of “lives saved”

or in terms of “lives lost.” What accounts for the difference?

The most sensible answer begins with the fact that human

beings are pervasively averse to losses (hence the robust

cognitive finding of loss aversion, Tversky & Kahneman

1991). With respect to either self-interested gambles or

fundamental moral judgments, loss aversion plays a large

role in people’s decisions. But what counts as a gain or a loss

depends on the baseline from which measurements are

made. Purely semantic reframing can alter the baseline and

hence alter moral intuitions (for many examples involving

fairness, see Kahneman et al. 1986).

This finding is usually taken to show a problem for stan-

dard accounts of rationality. Recently, however, it has been

argued that subjects are rationally responding to the infor-

mation provided, or “leaked,” by the speaker’s choice of

frame (McKenzie 2004). Certainly, the speaker’s choice

might offer a clue about the desired response; some sub-

jects in the Asian disease problem might be responding to

that clue. But even if people are generally taking account of

the speaker’s clues,

2

that claim is consistent with the propo-

sition that frames matter a great deal to moral intuitions,

which is all I am stressing here.

Moral framing has been demonstrated in the important

context of obligations to future generations (see Frederick

2003), a much-disputed question of morality, politics, and

law (Morrison 1998; Revesz 1999). To say the least, the ap-

propriate discount rate for those yet to be born is not a ques-

tion that most people have pondered, and hence their judg-

ments are highly susceptible to different frames. From a

series of surveys, Maureen Cropper and her coauthors sug-

gest that people are indifferent between saving one life to-

day and saving 45 lives in 100 years (Cropper et al. 1994).

They make this suggestion on the basis of questions asking

people whether they would choose a program that saves

“100 lives now” or a program that saves a substantially larger

number “100 years from now.” It is possible, however, that

people’s responses depend on uncertainty about whether

people in the future will otherwise die (perhaps technolog-

ical improvements will save them?); and other ways of fram-

ing the same problem yield radically different results (Fred-

erick 2003). For example, most people consider “equally

bad” a single death from pollution next year and a single

death from pollution in 100 years. This finding implies no

preference for members of the current generation. The sim-

plest conclusion is that people’s moral judgments about ob-

ligations to future generations are very much a product of

framing effects (for a similar result, see Baron 2000a).

3

The same point holds for the question of whether gov-

ernment should consider not only the number of “lives” but

also the number of “life-years” saved by regulatory inter-

ventions. If the government focuses on life-years, a pro-

gram that saves children will be worth far more attention

that a similar program that saves senior citizens. Is this im-

moral? People’s intuitions depend on how the question is

framed (see Sunstein 2004). People will predictably reject

an approach that would count every old person as worth

“significantly less” than what every young person is worth.

But if people are asked whether they would favor a policy

that saves 105 old people or 100 young people, many will

favor the latter, in a way that suggests a willingness to pay

considerable attention to the number of life-years at stake.

At least for unfamiliar questions of morality, politics, and

law, people’s intuitions are very much affected by framing.

Above all, it is effective to frame certain consequences as

“losses” from a status quo; when so framed, moral concern

becomes significantly elevated. It is for this reason that po-

litical actors, and those involved in law, often phrase one or

another proposal as “turning back the clock” on some social

advance. The problem is that for many social changes, the

framing does not reflect social reality, but is simply a verbal

manipulation.

Let us now turn to examples that are more controversial.

5. Moral heuristics: A catalogue

My principal interest here is the relationship between

moral heuristics and questions of law and policy. I separate

Sunstein: Moral heuristics

BEHAVIORAL AND BRAIN SCIENCES (2005) 28:4 535

the relevant heuristics into four categories: (1) those that

involve morality and risk regulation; (2) those that involve

punishment; (3) those that involve “playing God,” particu-

larly in the domains of reproduction and sex; and (4) those

that involve the act-omission distinction. The catalogue is

meant to be illustrative rather than exhaustive.

5.1. Morality and risk regulation

5.1.1. Cost–benefit analysis.

An automobile company is

deciding whether to take certain safety precautions for its

cars. In deciding whether or not to do so, it conducts a cost–

benefit analysis, in which it concludes that certain precau-

tions are not justified – because, say, they would cost $100

million and save only four lives, and because the company

has a “ceiling” of $10 million per life saved (a ceiling that is,

by the way, significantly higher than the amount the United

States Environmental Protection Agency uses for a statisti-

cal life). How will ordinary people react to this decision?

The answer is that they will not react favorably (see Viscusi

2000, pp. 547, 558). In fact, they tend to punish companies

that base their decisions on cost–benefit analysis, even if a

high valuation is placed on human life. By contrast, they im-

pose less severe punishment on companies that are willing

to impose a “risk” on people, but that do not produce a for-

mal risk analysis that measures lives lost and dollars, and

trades one against another (see Tetlock 2000; Viscusi 2000).

The oddity here is that under tort law, it is unclear that a

company should be liable at all if it has acted on the basis

of a competent cost–benefit analysis; such an analysis might

even insulate a company from a claim of negligence. What

underlies people’s moral judgments, which are replicated in

actual jury decisions (Viscusi 2000)?

It is possible that when people disapprove of trading

money for lives, they are generalizing from a set of moral

principles that are generally sound, and even quite useful,

but that work poorly in some cases. Consider the following

moral principle: Do not knowingly cause a human death. In

ordinary life, you should not engage in conduct with the

knowledge that several people will die as a result. If you are

playing a sport or working on your yard, you ought not to

continue if you believe that your actions will kill others. In-

voking that idea, people disapprove of companies that fail

to improve safety when they are fully aware that deaths will

result. By contrast, people do not disapprove of those who

fail to improve safety while believing that there is a “risk”

but appearing not to know, for certain, that deaths will en-

sue. When people object to risky action taken after cost–

benefit analysis, it seems to be partly because that very

analysis puts the number of expected deaths squarely “on

screen” (see Tetlock 2000).

Companies that fail to do such analysis, but that are

aware that a “risk” exists, do not make clear, to themselves

or to anyone else, that they caused deaths with full knowl-

edge that this was what they were going to do. People dis-

approve, above all, of companies that cause death know-

ingly. There may be a kind of “cold-heart heuristic” here:

Those who know they will cause a death, and do so anyway,

are regarded as cold-hearted monsters.

4

In this view, critics

of cost–benefit analysis should be seen as appealing to Sys-

tem I, and as speaking directly to the homunculus: “is a cor-

poration or public agency that endangers us to be pardoned

for its sins once it has spent $6.1 million per statistical life

on risk reduction?” (Ackerman & Heinzerling 2004).

Note that it is easy to reframe a probability as a certainty

and vice versa; if I am correct, the reframing is likely to have

large effects. Consider two cases:

(a) Company A knows that its product will kill ten peo-

ple. It markets the product to its ten million customers with

that knowledge. The cost of eliminating the risk would have

been $100 million.

(b) Company B knows that its product creates a one in

one million risk of death. Its product is used by ten million

people. The cost of eliminating the risk would have been

$100 million.

I have not collected data, but I am willing to predict that

Company A would be punished more severely than Com-

pany B, even though there is no difference between them.

I suggest, then, that a moral heuristic is at work, one that

imposes moral condemnation on those who knowingly en-

gage in acts that will result in human deaths. And of course

this heuristic does a great deal of good. The problem is that

it is not always unacceptable to cause death knowingly, at

least if the deaths are relatively few and an unintended

byproduct of generally desirable activity. When govern-

ment allows new highways to be built, it knows that people

will die on those highways; when government allows new

coal-fired power plants to be built, it knows that some peo-

ple will die from the resulting pollution; when companies

produce tobacco products, and when government does not

ban those products, hundreds of thousands of people will

die; the same is true for alcohol. Of course it would make

sense, in all of these domains, to take extra steps to reduce

risks. But that proposition does not support the implausible

claim that we should disapprove, from the moral point of

view, of any action taken when deaths are foreseeable.

There is a complementary possibility, involving the con-

fusion between the ex ante and ex post perspectives. If a life

might have been saved by a $50 expenditure on a car, peo-

ple are going to be outraged, and they will impose punish-

ment. What they will not see or incorporate is the fact, eas-

ily perceived ex ante, that the $50-per-car expenditure

would have been wasted on millions of other people. It is

hardly clear that the ex ante perspective is always prefer-

able. But something has gone badly wrong if the ex post

perspective leads people to neglect the tradeoffs that are

actually involved.

I believe that it is impossible to vindicate, in principle,

the widespread social antipathy to cost–benefit balancing.

5

But here too, “a little homunculus in my head continues to

jump up and down, shouting at me” that corporate cost–

benefit analysis, trading dollars for a known number of

deaths, is morally unacceptable. The voice of the ho-

munculus, I am suggesting, is not reflective, but is instead

a product of System I, and a crude but quite tenacious

moral heuristic.

5.1.2. Emissions trading. In the last decades, those in-

volved in enacting and implementing environmental law

have experimented with systems of “emissions trading”

(Sunstein 2002). In those systems, polluters are typically

given a license to pollute a certain amount, and the licenses

can be traded on the market. The advantage of emissions

trading systems is that if they work well, they will ensure

emissions reductions at the lowest possible cost.

Is emissions trading immoral? Many people believe so.

Political theorist Michael Sandel, for example, urges that

trading systems “undermine the ethic we should be trying

Sunstein: Moral heuristics

536 BEHAVIORAL AND BRAIN SCIENCES (2005) 28:4

to foster on the environment” (Sandel 1997; see also Kel-

man 1981). Sandel contends:

[T]urning pollution into a commodity to be bought and sold

removes the moral stigma that is properly associated with it. If

a company or a country is fined for spewing excessive pollutants

into the air, the community conveys its judgment that the pol-

luter has done something wrong. A fee, on the other hand,

makes pollution just another cost of doing business, like wages,

benefits and rent.

In the same vein, Sandel objects to proposals to open car-

pool lanes to drivers without passengers who are willing to

pay a fee. Here, as in the environmental context, it seems

unacceptable to permit people to do something that is

morally wrong as long as they are willing to pay for the priv-

ilege.

I suggest that, like other critics of emissions trading pro-

grams, Sandel is using a moral heuristic; in fact, he has been

fooled by his homunculus. The heuristic is this: People

should not be permitted to engage in moral wrongdoing for

a fee. You are not allowed to assault someone as long as you

are willing to pay for the right to do so; there are no trad-

able licenses for rape, theft, or battery. The reason is that

the appropriate level of these forms of wrongdoing is zero

(putting to one side the fact that enforcement resources are

limited; if they were unlimited, we would want to eliminate,

not merely to reduce, these forms of illegality). But pollu-

tion is an altogether different matter. At least some level of

pollution is a byproduct of desirable social activities and

products, including automobiles and power plants. Of

course, certain acts of pollution, including those that

amount to the intentional or reckless infliction of harm, are

morally wrong; but the same cannot be said of pollution as

such. When Sandel objects to emissions trading, he is treat-

ing pollution as equivalent to a crime in a way that over-

generalizes a moral intuition that makes sense in other con-

texts. There is no moral problem with emissions trading as

such. The insistent objection to emissions trading systems

stems from a moral heuristic.

Unfortunately, that objection has appeared compelling

to many people, so much as to delay and to reduce the use

of a pollution reduction tool that is, in many contexts, the

best available (Sunstein 2002). Here, then, is a case in

which a moral heuristic has led to political blunders, in the

form of policies that impose high costs for no real gain.

5.1.3. Betrayals. To say the least, people do not like to be

betrayed. A betrayal of trust is likely to produce a great deal

of outrage. If a babysitter neglects a child or if a security

guard steals from his employer, people will be angrier than

if the identical acts are performed by someone in whom

trust has not been reposed. So far, perhaps, so good: When

trust is betrayed, the damage is worse than when an other-

wise identical act has been committed by someone who was

not a beneficiary of trust. And it should not be surprising

that people will favor greater punishment for betrayals than

for otherwise identical crimes (see Koehler & Gershoff

2003). Perhaps the disparity can be justified on the ground

that the betrayal of trust is an independent harm, one that

warrants greater deterrence and retribution – a point that

draws strength from the fact that trust, once lost, is not eas-

ily regained. A family robbed by its babysitter might well

be more seriously injured than a family robbed by a thief.

The loss of money is compounded and possibly dwarfed by

the violation of a trusting relationship. The consequence of

the violation might also be more serious. Will the family

ever feel entirely comfortable with babysitters? It is bad to

have an unfaithful spouse, but it is even worse if the infi-

delity occurred with your best friend, because that kind of

infidelity makes it harder to have trusting relationships with

friends in the future.

In this light, it is possible to understand why betrayals

produce special moral opprobrium and (where the law has

been violated) increased punishment. But consider a find-

ing that is much harder to explain: People are especially

averse to risks of death that come from products (like

airbags) designed to promote safety (Koehler & Gershoff

2003). The aversion is so great that people have been found

to prefer a higher chance of dying, as a result of accidents

from a crash, to a significantly lower chance of dying in a

crash as a result of a malfunctioning airbag. The relevant

study involved two principal conditions. In the first, people

were asked to choose between two equally priced cars, Car

A and Car B. According to crash tests, there was a 2%

chance that drivers of Car A, with Air Bag A, will die in se-

rious accidents as a result of the impact of the crash. With

Car B, and Air Bag B, there was a 1% chance of death, but

also an additional chance of one in 10,000 (0.01%) of death

as a result of deployment of the air bag. Similar studies in-

volved vaccines and smoke alarms.

The result was that most participants (over two-thirds)

chose the higher risk safety option when the less risky one

carried a “betrayal risk.” A control condition demonstrated

that people were not confused about the numbers: when

asked to choose between a 2% risk and a 1.01% risk, peo-

ple selected the 1.01% risk so long as betrayal was not in-

volved. In other words, people’s aversion to betrayals is so

great that they will increase their own risks rather than sub-

ject themselves to a (small) hazard that comes from a de-

vice that is supposed to increase safety. “Apparently, people

are willing to incur greater risks of the very harm they seek

protection from to avoid the mere possibility of betrayal”

(Koehler & Gershoff 2003, p. 244). Remarkably, “betrayal

risks appear to be so psychologically intolerable that people

are willing to double their risk of death from automobile

crashes, fires, and diseases to avoid incurring a small possi-

bility of death by safety device betrayal.”

What explains this seemingly bizarre and self-destructive

preference? I suggest that a heuristic is at work: Punish, and

do not reward, betrayals of trust. The heuristic generally

works well. But it misfires in some cases, as when those who

deploy it end up increasing the risks they themselves face.

An airbag is not a security guard or a babysitter, endanger-

ing those whom they have been hired to protect. It is a

product, to be chosen if and only if it decreases aggregate

risks. If an airbag makes people safer on balance, it should

be used, even if in a tiny percentage of cases it will create a

risk that would not otherwise exist. People’s unwillingness

to subject themselves to betrayal risks, in circumstances in

which products are involved and they are increasing their

likelihood of death, is the moral cousin to the use of the rep-

resentativeness heuristic in the Linda case. Both stem from

a generally sound rule of thumb that leads to systematic er-

rors.

In a sense, the special antipathy to betrayal risks might

be seen to involve not a moral heuristic but a taste. In

choosing products, people are not making purely moral

judgments; they are choosing what they like best, and it just

turns out that a moral judgment, involving antipathy to be-

Sunstein: Moral heuristics

BEHAVIORAL AND BRAIN SCIENCES (2005) 28:4 537

trayals, is part of what they like best. It would be useful to

design a purer test of moral judgments, one that would ask

people not about their own safety but about that of others

– for example, whether people are averse to betrayal risks

when they are purchasing safety devices for their friends or

family members. There is every reason to expect that it

would produce substantially identical results to those in the

experiments just described. Closely related experiments

support that expectation (see Ritov & Baron 2002, p. 168).

In deciding whether to vaccinate their children from risks

for serious diseases, people show a form of “omission bias.”

Many people are more sensitive to the risk of the vaccina-

tion than to the risk from diseases – so much so that they

will expose their children to a greater risk from “nature”

than from the vaccine. (There is a clear connection between

omission bias, trust in nature, and antipathy to “playing

God,” as discussed later.) The omission bias, I suggest, is

closely related to people’s special antipathy to betrayals. It

leads to moral errors, in the form of vaccination judgments,

and undoubtedly others, by which some parents increase

the fatality risks faced by their own children.

5.2. Morality and punishment

5.2.1. Pointless punishment?

In the context of punish-

ment, moral intuitions are sometimes disconnected from

the consequences of punishment, suggesting that a moral

heuristic may well be at work (see Darley et al. 2000). Sup-

pose, for example, that a corporation has engaged in seri-

ous wrongdoing. People are likely to want to punish the cor-

poration as if it were a person (see Kahneman et al. 1998;

Sunstein et al. 2002). They are unlikely to inquire into the

possibility that the consequences of serious punishment

(say, a stiff fine) will not be to “hurt” corporate wrongdoers,

but instead to decrease wages, increase prices, or produce

lost jobs. Punishment judgments are rooted in a simple

heuristic, to the effect that penalties should be a propor-

tional response to the outrageousness of the act. In think-

ing about punishment, people use an outrage heuristic (see

Kahneman & Frederick 2002, pp. 49, 63). According to this

heuristic, people’s punishment judgments are a product of

their outrage. This heuristic may produce reasonable re-

sults much of the time, but in some cases, it seems to lead

to systematic errors – at least if we are willing to embrace

weak consequentialism.

Consider, for example, an intriguing study of people’s

judgments about penalties in cases involving harms from

vaccines and birth control pills (Baron & Ritov 1993). In

one case, subjects were told that the result of a higher

penalty would be to make companies try harder to make

safer products. In an adjacent case, subjects were told that

the consequence of a higher penalty would be to make the

company more likely to stop making the product, with the

result that less safe products would be on the market. Most

subjects, including a group of judges, gave the same penal-

ties in both cases. “Most of the respondents did not seem to

notice the incentive issue” (see Baron 1993a, pp. 108, 123).

In another study, people said that they would give the same

punishment to a company that would respond with safer

products and one that would be unaffected because the

penalty would be secret and covered by insurance (whose

price would not increase) (Baron 1993a). Here, too, the ef-

fects of the punishment did not affect judgments by a ma-

jority of respondents.

A similar result emerged from a test of punishment judg-

ments that asked subjects, including judges and legislators,

to choose penalties for dumping hazardous waste (Baron et

al. 1993). In one case, the penalty would make companies

try harder to avoid waste. In another, the penalty would lead

companies to cease making a beneficial product. Most peo-

ple did not penalize companies differently in the two cases.

Most strikingly, people preferred to require companies to

clean up their own waste, even if the waste did not threaten

anyone, instead of spending the same amount to clean up

far more dangerous waste produced by another, now-de-

funct company.

How could this preference make sense? Why should a

company be asked to engage in a course of action that costs

the same but that does much less good? In these cases, it is

most sensible to think that people are operating under a

heuristic, mandating punishment that is proportional to

outrageousness, and requiring that punishment be based

not at all on consequential considerations. As a general rule,

of course, it is plausible to think that penalties should be

proportional to the outrageousness of the act; utilitarians

will accept the point as a first approximation, and retribu-

tivists will insist on it. But it seems excessively rigid to adopt

this principle whether or not the consequence would be to

make human beings safer and healthier. Weak consequen-

tialists, while refusing to reject retributivism, will condemn

this excessive rigidity. Those who seek proportional pun-

ishments might well disagree in principle. But it would be

worthwhile for them to consider the possibility that they

have been tricked by a heuristic – and that their reluctance

to acknowledge the point is a product of the insistent voice

of their own homunculus.

5.2.2. Probability of detection. Now we turn to some

closely related examples from the domain of punishment.

On the economic account, the state’s goal, when imposing

penalties for misconduct, is to ensure optimal deterrence

(for this point and some complexities, see Polinsky &

Shavell 1998). To increase deterrence, the law might in-

crease the severity of punishment, or instead increase the

likelihood of punishment. A government that lacks sub-

stantial enforcement resources might impose high penal-

ties, thinking that it will produce the right deterrent “signal”

in light of the fact that many people will escape punishment

altogether. A government that has sufficient resources

might impose a lower penalty, but enforce the law against

all or almost all violators. These ideas lead to a simple the-

ory in the context of punitive damages for wrongdoing: The

purpose of such damages is to make up for the shortfall in

enforcement. If injured people are 100% likely to receive

compensation, there is no need for punitive damages. If in-

jured people are 50% likely to receive compensation, those

who bring suit should receive a punitive award that is twice

the amount of the compensatory award. The simple exer-

cise in multiplication will ensure optimal deterrence.

But there is a serious question of whether people accept

this account, and if not, why not. (For the moment, let us put

to one side the question of whether they should accept it in

principle.) Experiments suggest that people reject optimal

deterrence and that they do not believe that the probability

of detection is relevant to punishment. The reason is that

they use the outrage heuristic. I participated in two experi-

ments designed to cast light on this question (Sunstein et al.

2000). In the first experiment, subjects were given cases of

Sunstein: Moral heuristics

538 BEHAVIORAL AND BRAIN SCIENCES (2005) 28:4

wrongdoing, arguably calling for punitive damages, and also

were provided with explicit information about the probabil-

ity of detection. Different subjects saw the same case, with

only one difference: the probability of detection was sub-

stantially varied. Subjects were asked about the amount of

punitive damages that they would choose to award. The goal

was to see if subjects would impose higher punishments

when the probability of detection was low. In the second ex-

periment, subjects were asked to evaluate judicial and exec-

utive decisions made to reduce penalties when the probabil-

ity of detection was high, and to increase penalties when the

probability of detection was low. Subjects were questioned

about whether they approved or disapproved of varying the

penalty with the probability of detection.

The findings were simple and straightforward. The first

experiment found that varying the probability of detection

had no effect on punitive awards. Even when people’s at-

tention was explicitly directed to the probability of detec-

tion, they were indifferent to it. The second experiment

found that strong majorities of respondents rejected judi-

cial decisions to reduce penalties because of a high prob-

ability of detection – and also rejected executive decisions

to increase penalties because of a low probability of de-

tection. In other words, people did not approve of an ap-

proach to punishment that would make the level of pun-

ishment vary with the probability of detection. What

apparently concerned them was the extent of the wrong-

doing and the right degree of moral outrage – not optimal

deterrence.

Of course many people have principled reasons for em-

bracing retributivism and for rejecting utilitarian accounts

of punishment. And some such people are likely to believe,

on reflection, that the moral intuitions just described are

correct – that what matters is what the defendant did, not

whether his action was likely to be detected. But if we em-

brace weak consequentialism, we will find it implausible to

suggest that the aggregate level of misconduct is entirely ir-

relevant to punishment. We will be unwilling to ignore the

fact that if a legal system refuses to impose enhanced pun-

ishment on hard-to-detect wrongdoing, it will end up with

a great deal of wrongdoing. People’s unwillingness to take

any account of the probability of detection suggests the pos-

sibility that a moral heuristic is at work, one that leads to

real errors. Because of the contested nature of the ethical

issues involved, I cannot demonstrate this point; but those

who refuse to consider the probability of detection might

consider the possibility that System I has gotten the better

of them.

5.3. Playing God: Reproduction, nature, and sex

Issues of reproduction and sexuality are prime candidates

for the operation of moral heuristics. Consider human

cloning, which most Americans reject and believe should be

banned. Notwithstanding this consensus, the ethical and le-

gal issues here are extremely difficult. To make progress, it

is necessary to distinguish between reproductive and non-

reproductive cloning; the first is designed to produce chil-

dren, whereas the second is designed to produce cells for

therapeutic use. Are the ethical issues different in the two

cases? In any case, it is important to identify the particular

grounds for moral concern. Do we fear that cloned children

would be means to their parents’ ends, and if so, why? Do

we fear that they would suffer particular psychological

harm? Do we fear that they would suffer from especially se-

vere physical problems?

In a highly influential discussion of new reproductive tech-

nologies, above all cloning, ethicist Leon Kass (1998, pp. 17–

19) points to the “wisdom in repugnance.” Kass writes:

People are repelled by many aspects of human cloning. They re-

coil from the prospect of mass production of human beings, with

large clones of look-alikes, compromised in their individuality,

the idea of father-son or mother-daughter twins; the bizarre

prospects of a woman giving birth to and rearing a genetic copy

of herself, her spouse or even her deceased father or mother;

the grotesqueness of conceiving a child as an exact replacement

for another who has died; the utilitarian creation of embryonic

genetic duplicates of oneself, to be frozen away or created when

necessary, in case of need for homologous tissues or organs for

transplantation; the narcissism of those who would clone them-

selves and the arrogance of others who think they know who de-

serves to be cloned or which genotype any child-to-be should be

thrilled to receive; the Frankensteinian hubris to create human

life and increasingly to control its destiny; man playing God . . .

We are repelled by the prospect of cloning human beings not

because of the strangeness or novelty of the undertaking, but be-

cause we intuit and feel, immediately and without argument, the

violation of things that we rightfully hold dear.... Shallow are

the souls that have forgotten how to shudder.

Kass is correct to suggest that revulsion toward human

cloning might be grounded in legitimate concerns, and I

mean to be agnostic here on the question of whether hu-

man cloning is ethically defensible. But I want to suggest

that moral heuristics, and System I, are responsible for what

Kass seeks to celebrate as that which “we intuit and feel, im-

mediately and without argument.” In fact Kass’s catalogue

of alleged errors seems to me to be an extraordinary exer-

cise in the use of such heuristics. Availability operates in this

context, not to drive judgments about probability, but to call

up instances of morally dubious behavior (e.g., “mass pro-

duction of human beings, with large clones of look-alikes,

compromised in their individuality”). The representative-

ness heuristic plays a similar role (e.g., “the Frankenstein-

ian hubris to create human life and increasingly to control

its destiny”). But I believe that Kass gets closest to the cog-

nitive process here with three words: “man playing God.”

In fact, we might well think that “do not play God” is the

general heuristic here, with different societies specifying

what falls in that category and with significant changes over

time. Even in secular societies, a closely related heuristic

plays a large role in judgments of fact and morality: Do not

tamper with nature. This heuristic affects many moral judg-

ments, though individuals and societies often become ac-

customed to various kinds of tampering (consider in vitro

fertilization). An anti-tampering heuristic helps explain many

risk-related judgments. For example, “[h]uman interven-

tion seems to be an amplifier in judgments on food riskiness

and contamination,” even though “more lives are lost to nat-

ural than to man-made disasters in the world” (Rozin 2001,

pp. 31, 38). Studies show that people overestimate the car-

cinogenic risk from pesticides and underestimate the risks

of natural carcinogens (Rozin 2001). People also believe

that nature implies safety, so much so that they will prefer

natural water to processed water even if the two are chem-

ically identical (Rozin 2001).

The moral injunction against tampering with nature

plays a large role in public objections to genetic engineer-

ing of food, and hence legal regulation of such engineering

is sometimes driven by that heuristic rather than by a de-

Sunstein: Moral heuristics

BEHAVIORAL AND BRAIN SCIENCES (2005) 28:4 539

liberative, System II encounter with the substantive issues.

For genetic engineering, the anti-tampering heuristic dri-

ves judgments even when the evidence of risk is slim

(McHughen 2000). In fact, companies go to great lengths

to get a “natural” stamp on their products (Schlosser 2002),