How Change Happens

by Cass R Sunstein

  Although choice architecture and nudging are inevitable, some nudges are certainly avoidable. A government might decide not to embark on a campaign to discourage smoking or unhealthy eating. It might refrain from nudging people toward certain investment behavior. To that extent, it is reasonable to wonder whether government should minimize nudging. If we distrust the motives of public officials or believe that their judgments are likely to go wrong, we will favor such minimization. Some of the ethical objections to choice architecture and to nudging are best understood as a plea for minimization, not elimination. In fact the plea might well be more precise than that. At least in some cases, it is a claim that government should avoid particular interventions that are taken to be manipulative and hence to compromise both autonomy and dignity.

  Seven Objections

  Here is a catalog of potential objections to nudges as such: (1) Nudges are paternalistic. (2) Some nudges intrude on people’s autonomy. (3) Some nudges might be seen as coercive, even if they preserve freedom of choice as a technical matter. (4) Some nudges insult people’s dignity; they might be infantilizing; they might treat people as children. The idea of the “nanny state” captures this objection. (5) Some nudges could count as forms of manipulation. It is relevant in this regard that nudging is not always transparent. Consider, for example, negative option marketing, by which people who purchase certain products find themselves enrolled in programs for which they pay a monthly fee. Nudges of this kind might be said to operate behind people’s backs. (6) Some nudges impede or at least do not promote learning. (7) Choice architects may err, especially when they work for government, and for that reason, it is best to avoid nudging (to the extent that this is possible).

  It is important not to take these concerns as all-purpose objections to efforts to improve choice architecture. Does any of these objections make sense as applied to initiatives to promote active choosing? To inform consumers of the caloric content of food, to remind people that a bill is due, or to ask people whether they want to enroll in a retirement plan? We might be skeptical about the force of these concerns as applied to the overwhelming majority of real-world nudges. But let us take the objections in sequence.

  Paternalism

  If paternalism is objectionable, it is because it runs afoul of some kind of foundational commitment or principle. For example, it might undermine autonomy. Although the term is often used as a freestanding objection, the real complaint is that paternalism, in general or in particular circumstances, violates a principle that people rightly value.

  Choice architecture may or may not be paternalistic. But it is true that nudges can be seen as a form of “libertarian paternalism” insofar as they attempt to use choice architecture to steer choosers in directions that will promote their welfare (as judged by choosers themselves). A GPS device can be so understood, and the same is true for a reminder, a warning, a use of social norms, and a default rule.

  This is a distinctive form of paternalism in the sense that it is (a) soft and (b) means-oriented. It is soft insofar as it avoids coercion or material incentives, and thus fully maintains freedom of choice. It is means-oriented insofar as it does not attempt to question or alter people’s ends. Like a GPS device, it respects those ends. To those who object to paternalism, the most serious concerns arise in the face of coercion (where freedom of choice is blocked) and when social planners, or choice architects, do not respect people’s ends. To this extent, nudges aspire to avoid some of the standard ethical objections to paternalism.

  Nonetheless, some skeptics object to paternalism as such. Many of those objections point to individual welfare and to the risk that planners or choice architects will compromise it. Perhaps people are the best judges not only of their ends but also of the best means to achieve those ends, given their own tastes and values. People might reject the route suggested by a GPS device on the ground that they prefer the scenic alternative; the GPS device might not easily capture or serve their ends. A coercive GPS device would, in some cases, intrude on people’s ends.

  Moreover, the distinction between means and ends is not always simple and straightforward. One question is the level of abstraction at which we describe people’s ends. If we describe people’s ends at a level of great specificity—eating that brownie, having that cigarette, texting while driving—then people’s means effectively are their ends. The brownie is exactly what they want; it is not a means to anything at all (except the experience of eating it).

  If we describe people’s ends at a level of high abstraction—“having a good life”—then nearly everything is a means to those ends. But if we do that, then we will not be capturing people’s actual concerns; we will be disregarding what really matters to them. These points do raise some problems for those who favor a solely means-oriented form of paternalism. They must be careful to ensure that they are not describing people’s ends at a sufficiently high level of abstraction as to misconceive what people care about.

  But insofar as a GPS device is a guiding analogy, and insofar as freedom of choice is fully maintained, it is not easy to see nudges as objectionably paternalistic. At least some nudges are entirely focused on means. Consider cases in which people are mistaken about facts (with respect to the characteristics of, say, a consumer product or an investment). If a nudge informs them, then it is respecting their ends. Or suppose that people suffer from a behavioral bias—perhaps because they use the availability heuristic, perhaps because of unrealistic optimism. A nudge that corrects their mistake can help them to achieve their ends.

  To be sure, some behavioral biases are not easy to analyze in these terms. If people suffer from present bias, is a nudge a form of paternalism about means? Suppose that people eat high-calorie food or drink a great deal or fail to exercise because they value today and tomorrow, and not so much next year or next decade. If a nudge succeeds in getting people to focus on their long-term interests, it might increase aggregate (intrapersonal) welfare over time. But is such a nudge focused solely on means? If a person is seen as a series of selves extending over time, the choice architect is effectively redistributing welfare from earlier selves to later ones (and by hypothesis maximizing welfare as well). But it is not clear that we can speak, in such cases, of means paternalism. And if a person is seen as continuous over time, and not a series of selves, efforts to counteract present bias are, by hypothesis, undermining the ends of the chooser at the time of choice. It is hard to question whether the relevant ends are those that the chooser has at that time or at a later time or whether it is best (as I tend to think) to identify the chooser’s ends by focusing on some aggregation or index of selves over a lifetime.

  Let us bracket the most difficult issues and acknowledge that some forms of choice architecture count as paternalistic. Is that a problem? As we have seen, one reason to reject paternalism involves welfare: Perhaps people are the best judges of what will promote their interests, and perhaps outsiders will blunder (as Mill believed). Consider Hayek’s remarkable suggestion that “the awareness of our irremediable ignorance of most of what is known to somebody [who is a chooser] is the chief basis of the argument for liberty.”5 Hayek was speaking of welfare, not autonomy. He was suggesting that the “somebody,” acting as chooser, would inevitably know more than “we” know. Consider someone who is choosing what to eat for lunch, where to live, who to date, what job to take. A form of paternalism that maintains liberty and that is focused on means is less likely to be objectionable on welfare grounds. An understanding of behavioral biases suggests that people might err with respect to means; but an appreciation of the special knowledge of choosers suggests that we should respect their judgments about their ends.

  A possible response is that even means-oriented nudges will be inadequately informed, at least if they come from government. When public officials produce default rules, they might steer people in bad directions; when they inform people, they might tell them something that is false, or that it is not useful to know. If we are de
eply skeptical of the good faith and competence of public officials, we will want to minimize official nudges and we will prefer choice architecture that comes from invisible hands and spontaneous orders.

  This view cannot be ruled out in the abstract. It depends on certain assumptions about the risk of government error; whether or not it is right, it is not clear that it should be counted as a distinctly ethical objection. Nor it is clear that it should count as an objection to efforts to increase navigability, unless we think that those efforts are themselves likely to be misconceived.

  In the face of ignorance of fact and behavioral biases, some welfarists are drawn to coercive paternalism.6 When paternalism would improve welfare, welfarists should support it. For welfarists, there is a good argument that paternalism, hard or soft, should be evaluated on a case-by-case basis—unless there is some systematic, or rule-welfarist, justification for a principle or presumption against paternalism.

  There may be good reason for such a presumption, rooted in a judgment that choosers are likely to have better information than choice architects. But in some cases, that judgment is incorrect, because choosers lack, and choice architects have, accurate knowledge of relevant facts. There are serious risks in using this point as a reason for coercion, at least when the interests of third parties are not involved. The advantage of nudges is that they reduce those risks, because people are allowed to go their own way (see chapter 10 for discussion).

  In the face of missing information, information-providing nudges are a natural corrective. But in some cases, a good default rule—say, automatic enrollment in pension programs—is hard to reject on welfarist grounds. To be sure, active choosing might be better, but that conclusion is not obvious; much depends on the costs of decisions and the costs of errors. Welfarists might well be inclined to favor choice-preserving approaches, on the theory that individuals usually know well what best fits their circumstances, but default rules preserve choice; the fact that they have a paternalistic dimension should not be decisive against them.

  Another reason to reject paternalism involves autonomy and the idea of respect for persons. Stephen Darwell writes that the “objectionable character of paternalism of this sort is not that those who seek to benefit us against our wishes are likely to be wrong about what really benefits us.… It is, rather, primarily a failure of respect, a failure to recognize the authority that persons have to demand, within certain limits, that they be allowed to make their own choices for themselves.”7 This brings us to the next objection.

  Autonomy

  Do nudges intrude on autonomy? Recall that nudges are inevitable, so in a sense the question is confusing. It appears to be premised on a judgment that existing nudges do not compromise autonomy and that a new nudge, proposed or actual, would introduce nudging where it did not exist before. That is a mistake. The real question is whether particular nudges intrude on autonomy.

  In any case, autonomy requires informed choices, and many nudges are specifically designed to ensure that choices are informed. In the face of a lack of information, a behavioral bias, or some kind of systematic mistake (by the actor’s own reflective lights), it is hardly clear that a nudge, taken as such, infringes on autonomy, rightly understood. And when social contexts are not navigable, a nudge that improves navigability increases autonomy rather than undermining it. A GPS device does not create a problem from the standpoint of autonomy, nor does a user-friendly computer, nor does an effort to simplify a mortgage or a credit card agreement.

  It is also important to see that autonomy does not require choices everywhere; it does not justify an insistence on active choosing in all contexts. If people have to make choices everywhere, their autonomy is reduced, if only because they cannot devote attention to those activities that seem to them most worthy of their attention. There is a close relationship between time management and autonomy. To the extent that nudges reduce the difficulty of time management, they increase an important form of autonomy.

  It is nonetheless true that on grounds of autonomy and welfare, the best choice architecture often calls for active choosing. Default rules might intrude on autonomy, especially if they do not track people’s likely (informed) choices. It is important to remember that because of the force of inertia, people might not reject harmful defaults. If so, there is an intrusion on their autonomy because they will end up with outcomes that they did not specifically select and do not or will not like.

  Whether the interest in autonomy calls for active choosing, as opposed to reliance on a default rule, depends on the circumstances. Often active choosing is best and should be built into the relevant choice architecture—but along some dimensions, default rules can be superior to active choosing on autonomy grounds. If people choose not to choose, or if they would make that choice if asked, it is an insult to their autonomy to force them to choose. And if people would like to choose, a default rule does not deprive them of that choice; they can reject the default. Even in the face of inertia, many people will do so. Preservation of freedom of choice goes a long way toward respecting autonomy. With respect to autonomy, the real problem lies in the possibility of manipulation; I will turn to that problem later.

  Coercion

  If choice architects coerce people, they are no longer merely nudging. But skeptics might again emphasize that because of the power of inertia, people might (passively) accept a default rule even though they have no enthusiasm for the outcome that it produces and would reject that outcome if they focused on the issue involved.

  We should doubt whether such situations are properly described as involving coercion; no one is being forced to do anything. But there is certainly a risk that a default rule will produce harmful results even though people have not affirmatively consented to the actions that led to them. Choice architects need to account for that risk. But so long as freedom of choice is maintained and is real rather than formal, coercion is not involved.

  Dignity

  The antonym of coercion is freedom; the antonym of dignity is humiliation. As we shall discuss, this may be the most interesting objection of all, especially when it is combined with a concern about manipulation.

  There are large questions about the place of dignity in ethics. On one (less than conventional) view, dignity is properly part of an assessment of welfare. If people feel humiliated or think that they have been treated disrespectfully, they suffer a welfare loss. That loss might be extremely serious. In any assessment of welfare consequences, such a loss must be considered.

  A good welfarist should also acknowledge that an offense to dignity is qualitatively distinct; it is a different kind of loss from the loss of (say) money or an opportunity to visit a beach. But on the welfarist view, a dignity loss is just one kind of loss, to be weighed against the other goods that are at stake. Suppose, for purposes of argument, that a graphic and highly emotional appeal, triggering strong emotions (System 1) to discourage people from smoking, is plausibly seen as an offense of dignity—as a way of treating smokers disrespectfully. (Some smokers might so regard such an appeal and object for that reason.) A welfarist might be willing to support the emotional appeal, notwithstanding the relevant loss, if it saves a significant number of lives.

  On another view, an insult to dignity is not merely part of a welfarist calculus. Such an insult does not depend on people’s subjective feelings and it is a grave act, perhaps especially if it comes from government. It should not be permitted unless (perhaps) it has an overwhelmingly strong justification. If we endorse this view, it is especially important to ask whether nudges offend human dignity.

  To return to my general plea: the force of the objection depends on the particular nudge. A GPS device insults no one’s dignity. Disclosure of factual information can hardly be seen as an offense to dignity—certainly not if the information is useful and not based on a false and demeaning belief that people need it. Nudges that increase navigability do not offend dignity.

  True, it might be an insult to dignity, and
a form of infantilization, if the government constantly reminds people of things that they already know. Every child, and everyone who was once a child, can recall this form of infantilization, and it can be found in adult life as well. If people are informed of the same thing every hour or even every day, they might legitimately feel that their dignity is not being respected. If people are constantly reminded that a due date is coming, they might feel as if they are being treated like children.

  If choice architects refer to social norms, to let people know what most people do, they are not likely to be humiliating anyone. In some cases, however, repeated invocations of social norms might run into a concern about dignity. Warnings can run into the same concern insofar as they are repetitive or condescending or (are meant to) trigger strong emotions instead of merely giving people a sense of factual realities.

  Here as well, there is no objection to the relevant nudges in the abstract, but there is an objection to imaginable nudging. At the same time, it must be emphasized that outside of exotic hypotheticals, the relevant offense to dignity—coming from unwelcome and numerous reminders—is relatively minor, and from the standpoint of the concerns that have produced the focus on dignity in the Western political tradition, it is laughably modest.

  What is the relationship between dignity and default rules? If an employer automatically enrolls employees into retirement and health care plans, dignity is hardly at risk. If a cell phone company adopts a series of defaults for the phone and the contractual arrangement, nothing need be amiss in terms of dignity.

  But we could imagine harder cases. Suppose that the government insisted on “default meals” in various restaurants so that people would be given certain healthy choices unless they specifically chose otherwise. The reasonable response is: Why shouldn’t free people be asked to choose what they want? Why should they be subject to a default rule at all? Or suppose that a government specified a “default exercise plan” for adults so that they would be assumed to engage in certain activities unless they expressly opted out. People might offer the same reasonable response, perhaps with considerable agitation. A more modest and more realistic proposal is that doctors should enroll smokers into some kind of smoking-cessation plan by default (subject to opt out). Might automatic enrollment offend dignity?