deductive valid arguments in numbered premise form

Hanna-LibertarianPaternalismManipulation-2015.pdf

Home >Social Science homework help >Philosophy homework help >deductive valid arguments in numbered premise form

Libertarian Paternalism, Manipulation, and the Shaping of Preferences

Author(s): Jason Hanna

Source: Social Theory and Practice , October 2015, Vol. 41, No. 4, Special Issue: Preference, Choice, and (Libertarian) Paternalism (October 2015), pp. 618-643

Published by: Florida State University Department of Philosophy

Stable URL: https://www.jstor.org/stable/24575752

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at https://about.jstor.org/terms

is collaborating with JSTOR to digitize, preserve and extend access to Social Theory and Practice

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

https://www.jstor.org/stable/24575752

Libertarian Paternalism, Manipulation, and the Shaping of Preferences

Jason Hanna

Abstract: "Libertarian paternalism" aims to harness cognitive biases in order to improve prudential decision-making. Some critics have objected that libertarian paternalism is wrongly manipulative. I argue that this objection is mostly unsuccessful. First, I point out that some strategies endorsed by libertarian paternalists can help people to better appre ciate reasons. Second, I develop an account of manipulation according to which an agent manipulates her target by worsening the target's deliberative position. The means of in fluence defended by libertarian paternalists—for instance, the judicious use of default rules—are not manipulative in this way.

Keywords: paternalism; manipulation; libertarian paternalism; nudge; Cass Sunstein; Richard Thaler

A number of philosophers, social scientists, and legal scholars—most famously Richard Thaler and Cass Sunstein—have defended policy measures that they classify as instances of "libertarian" paternalism.1 De fenders of libertarian paternalism observe that there are ways of influen cing decision-making that do not limit liberty but also do not quite fit the mold of ordinary persuasion. Because we are susceptible to various cog nitive biases, our decision-making is often affected by "choice architec ture," or the way in which options are structured and presented. Liberta rian paternalism aims to encourage more prudent decision-making through the design of choice architecture. As Thaler and Sunstein point out, a company might increase employee participation in a retirement savings plan by making enrollment the default, so that employees must actively opt out if they do not wish to save. To take a second example, a

'Thaler and Sunstein have elaborated and defended their position in a number of books and articles. See Cass R. Sunstein, Why Nudge? The Politics of Libertarian Pater nalism (New Haven: Yale University Press, 2014); Richard H. Thaler and Cass R. Suns tein, Nudge: Improving Decisions about Health, Wealth, and Happiness (New Haven: Yale University Press, 2008); Sunstein and Thaler, "Libertarian Paternalism is Not an Oxymoron," The University of Chicago Law Review 70 (2003): 1159-202; Thaler and Sunstein, "Libertarian Paternalism," American Economic Review 93 (2003): 175-79; and Sunstein and Thaler, "Preferences, Paternalism, and Liberty," Royal Institute of Philoso phy Supplement 59 (2006): 233-64.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

Libertarian Paternalism, Manipulation, and Preferences 619

medical practitioner may influence patient decision-making by delibe rately framing or presenting information in a way that is likely to get pa tients to choose more prudently.2 Proponents of such measures have claimed that they ought to be acceptable even to liberals who oppose more coercive forms of paternalism.3

Perhaps not surprisingly, however, libertarian paternalism has been widely criticized on at least two grounds. First, some critics have pressed objections that are essentially consequentialist. Even if each of us some times falls prey to cognitive biases and judgment-distorting heuristics, who is to say that "choice architects"—viz., those responsible for pre senting and structuring options—could do better?4 Moreover, if the gov ernment begins exploiting heuristics and biases, one might worry that it will use such strategies for nefarious ends. A second, distinct, line of crit icism is nonconsequentialist. It maintains that even if libertarian paternal ist strategies were effective and used only for worthy ends, they would still be objectionable. One commonly expressed concern is that liberta rian paternalism is objectionably manipulative.5 Call this the manipula tion objection. The manipulation objection is probably more daunting than the consequentialist objections described above. It is hard to believe that the use of libertarian paternalist strategies does more harm than good in each and every case. Consequentialist considerations are thus unlikely to pose a general objection to libertarian paternalist interventions as a class, for many of the same reasons that consequentialist considerations are unlikely to pose a general objection to coercive paternalist interven

2Shlomo Cohen, "Nudging and Informed Consent," The American Journal of Bioeth ics 13, no. 6 (2013): 3-11; and Jason Hanna, "Consent and the Problem of Framing Effects," Ethical Theory and Moral Practice 14 (2011): 517-31.

3In addition to Thaler and Sunstein, see J.S. Blumenthal-Barby, "Choice Architec ture: A Mechanism for Improving Decisions While Preserving Liberty?" in Christian Coons and Michael Weber (eds.), Paternalism: Theory and Practice (New York: Oxford University Press, 2013), pp. 178-96; J.D. Trout, "Paternalism and Cognitive Bias," Law and Philosophy 24 (2005): 393-434; and Cohen, "Nudging and Informed Consent." Some of these other writers (wisely, in my view) avoid the "libertarian paternalist" label.

4Edward Glaeser, "Paternalism and Psychology," University of Chicago Law Review 73 (2006): 133-56.

5Daniel Hausman and Brynn Welch, "Debate: To Nudge or Not to Nudge," Journal of Political Philosophy 18 (2010): 123-36, p. 128; Till Griine-Yanoff, "Old Wine in New Casks: Libertarian Paternalism Still Violates Liberal Principles," Social Choice and Wel fare 38 (2012): 635-45; Christian Coons and Michael Weber, "Introduction: Paternal ism—Issues and Trends," in Coons and Weber (eds.), Paternalism, pp. 1-24, esp. p. 23 ; and Luc Bovens, "The Ethics of Nudge," in Till Griine-Yanoff and S.O. Hansson (eds.), Preference Change: Approaches from Philosophy, Economics and Psychology (New York: Springer, 2009), pp. 207-19. Even some defenders of paternalism believe that li bertarian paternalism is manipulative; see Sarah Conly, Against Autonomy (New York: Cambridge University Press, 2013), pp. 39-40.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

620 Jason Hanna

tions as a class.6 By contrast, the manipulation objection, if sound, would provide at least a pro tanto reason to reject each use of the supposedly problematic strategies of influence favored by libertarian paternalists. Moreover, many philosophers have thought that the charge of manipula tion poses a special problem for libertarian paternalism (as opposed to other, more coercive, forms of intervention).7

In this essay, I argue that the manipulation objection is unsuccessful. It appears that a response to the manipulation objection could take one of three forms.8 First, one might argue that libertarian paternalism is not generally manipulative. Second, one might argue that although liberta rian paternalism might be manipulative, it is usually not objectionably manipulative. Finally, one might argue that although libertarian paternal ism is usually objectionably manipulative, and thus pro tanto wrong, it is often permissible all things considered. The distinction among these three possible responses is complicated by the fact that philosophers dis agree about whether manipulation is a moralized concept—that is, whether manipulation is (at least pro tanto) wrong as a matter of defini tion.9 If the concept of manipulation is moralized in this way, then the first and second responses above are not genuine alternatives to each other. Given my aims here, I need not take a decisive stand on this issue. My goal is to determine whether libertarian paternalism is (at least pro tanto) objectionable in the way characteristic of other behavior that ap pears to be wrongly manipulative. My argumentative strategy is thus best captured by the second alternative above. When I claim that libertarian paternalism is not manipulative, I mean to claim that it is not objectiona bly manipulative. Even if everything I say is correct, it may turn out that libertarian paternalism is manipulative in some sense that holds little

6For relevant discussion, see Jeremy A. Blumenthal, "A Psychological Defense of Paternalism," in Coons and Weber (eds.), Paternalism, pp. 197-215; and Gerald Dwor kin, "Paternalism," in Richard A. Wasserstrom (ed.), Morality and the Law (Belmont, Calif.: Wadsworth, 1971), pp. 107-26, esp. pp. 113-17.

7See Coons and Weber, "Introduction: Paternalism—Issues and Trends," p. 23. Even if libertarian paternalism is not manipulative, a critic might worry that it compromises autonomy by giving others (perhaps especially the state) too much control over one's own life. This objection would presumably also apply to coercive paternalism. Regard less, I cannot discuss all possible nonconsequentialist objections to libertarian paternal ism here.

8I am indebted to Danny Scoccia for encouraging me to distinguish among these different possibilities.

9For relevant discussion, see Allen Wood, "Coercion, Manipulation, Exploitation," in Christian Coons and Michael Weber (eds.), Manipulation: Theory and Practice (New York: Oxford University Press, 2014), pp. 17-50, esp. pp. 18-21; Marcia Baron, "Manipula tiveness," Proceedings and Addresses of the American Philosophical Association 77, no. 2 (2003): 37-54, esp. pp. 39-40; and Marcia Baron, "The Mens Rea and Moral Status of Ma nipulation," in Coons and Weber (eds.), Manipulation, pp. 98-120, esp. pp. 114-16.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

Libertarian Paternalism, Manipulation, and Preferences 621

moral interest: for instance, perhaps choice architects can be said to ma nipulate choice architecture in the same morally innocuous way that re search scientists manipulate experimental conditions.

My argument proceeds as follows. In section 1, I distinguish what I shall call "preference-shaping" libertarian paternalism from other forms of intervention before briefly fleshing out the manipulation objection. I offer a preliminary response to this objection in section 2. There, I argue that libertarian paternalism sometimes helps people to appreciate reasons relevant to their behavior. In section 3,1 develop a fuller account of ma nipulation. I proceed to show in section 4 that this account implies that libertarian paternalism is not generally manipulative. Finally, in section 5,1 respond to the objection that my view could permit subliminal mes saging. I conclude that concerns about manipulation do not provide a sound objection to libertarian paternalist measures of the sort endorsed by Thaler and Sunstein.

1. Libertarian Paternalism and the Manipulation Objection

Before turning to the manipulation objection, it will be useful to have at least a rough characterization of libertarian paternalism. As Thaler and Sunstein describe it, libertarian paternalism differs from other forms of paternalism in at least three respects. First, and most importantly, liberta rian paternalism supposedly does not limit liberty: it improves prudential decision-making without removing options or making it costly for people "to obtain their preferred outcomes."10 Second, Thaler and Sunstein claim that libertarian paternalism is "inevitable," since people's choices sometimes depend on how options are presented.11 This claim is some what misleading, since it is surely false to claim that libertarian paternal ism is inevitable. After all, someone could exploit choice architecture without any concern for the target's own good.12 The more charitable interpretation of Thaler and Sunstein's claim is that influence is often inevitable: in many cases, there is no "neutral" choice architecture.13 Third, it is commonly claimed that the strategies favored by libertarian paternalists "harness" cognitive biases for good ends.14 By contrast, other

10Sunstein and Thaler, "Libertarian Paternalism is Not an Oxymoron," p. 1185. "ibid., pp. 1171 ff. 12Jamie Kelly, "Libertarian Paternalism, Utilitarianism, and Justice," in Coons and

Weber (eds.), Paternalism, pp. 216-30. l3Kalle Grill, "Expanding the Nudge: Designing Choice Contexts and Choice Con

tents," Rationality, Markets and Morals 5 (2014): 139-62, pp. 142-45; Hanna, "Consent and the Problem of Framing Effects"; and Blumenthal-Barby, "Choice Architecture," p. 194.

14This language is used by Thaler and Sunstein (Nudge, p. 8) and Trout ("Paternalism and Cognitive Bias," p. 432).

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

622 Jason Hanna

forms of paternalism—for instance, fining people in order to deter self harmful behavior—need not rely on susceptibility to cognitive bias: even a perfectly rational agent who is immune to biases might respond to pe nalties or incentives. The sorts of paternalistic influence that I will examine here share one further characteristic: they do not involve deception or the deliberate withholding of information that one has a duty to disclose.

As we have seen, several of Thaler and Sunstein's examples involve the judicious use of framing effects or default rules. In some cases, dif ferent but equally informative ways of framing or describing options might elicit different choices. Famously, people seem more favorably disposed to a potentially risky medical procedure when the risks are framed in terms of the survival rate ("90% of patients survive the proce dure") rather than the mortality rate ("10% die on the operating table").15 Default rules can be significant, since defaults tend to be "sticky": when an option is selected as the default, more people will end up with it. In addition to the use of frames and defaults, Thaler and Sunstein tout a motley array of putatively libertarian paternalist measures, including mandatory "cooling-off' periods,16 graphic warnings,17 and policies un der which gambling addicts can place themselves on a list of people to be denied entry to casinos.18

In considering such cases, it is useful to distinguish between what I will call preference-shaping intervention and non-preference-shaping interven tion}9 Preference-shaping intervention aims to influence behavior by oper ating on a person "from the inside." As a result of preference-shaping intervention, a certain option will appear more (or less) attractive than it otherwise would, even though the objective features of this option have not significantly changed.20 Of course, one way to alter a person's preferences is to simply provide her with information, but I will assume (for now, at least) that preference shaping must involve the use of nonrational means. Non-preference-shaping intervention, by contrast, aims to influence beha vior by making certain options more (or less) costly, or more (or less) dif ficult to obtain, without in other ways changing the target's preferences.

15Barbara J. McNeil, Stephen G. Pauker, Harold C. Sox, Jr., and Amos Tversky, "On the Elicitation of Preferences for Alternative Therapies," New England Journal of Medi cine 306, no. 21 (27 May 1982): 1259-62.

l6Thaler and Sunstein, Nudge, pp. 250-51. l7Sunstein, Why Nudge? p. 48. l8Thaler and Sunstein, Nudge, p. 233. l9The "shaping" terminology is due to Hausman and Welch, "Debate," p. 128. 20For a similar distinction, see Claudia Mills, "Politics and Manipulation," Social

Theory and Practice 21 (1995): 97-112, p. 99. See also the discussion of the difference between manipulation and coercion in Christian Coons and Michael Weber, "Introduc tion: Investigating the Core Concept and Its Moral Status," in Coons and Weber (eds.), Manipulation, pp. 1-16, at p. 15.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

Libertarian Paternalism, Manipulation, and Preferences 623

The distinction between preference-shaping and non-preference shaping intervention suggests that there are potentially important differ ences among the examples Thaler and Sunstein discuss. Many of their examples involve non-preference-shaping influence that does not appear to be manipulative at all. Suppose that most (but not all) of a company's employees prefer to be enrolled in a particular health insurance plan but do not want to go through the hassle of filling out an enrollment form. If the company sets the relevant plan as the default for the sake of these employees, it is respecting their preferences, not shaping them.21 Al though other employees may be annoyed that they have to fill out paper work to opt out, they cannot plausibly claim that they are being manipu lated because their preferred plan was not set as the default. Other cases discussed by Thaler and Sunstein do not share the features of libertarian paternalism described above and do not appear to involve preference shaping. Mandatory cooling-off periods give a person the opportunity to change her mind but may not be aimed at changing her preferences. They also seem to limit liberty, since they in effect require someone to wait to do something that she might want to do now. Pre-commitment strategies, such as gambling self-bans, may not shape anyone's preferences; instead, they restrict (future) liberty, even if such restriction is justified by the subject's past consent.

The most likely target of the manipulation objection is preference shaping paternalism. Consider the healthcare professional who frames the risks of a medical procedure so as to encourage the patient's consent. Or consider Thaler and Sunstein's well-known cafeteria example: given that customers are influenced by the order in which food is displayed, a cafeteria owner may place vegetables near the front of the line to encour age healthier eating.22 These cases, unlike others discussed above, meet the criteria for libertarian paternalism: no one's liberty is limited, some influence is inevitable (since the risks must be presented in some way and the food displayed in some order), and the targets' cognitive biases are exploited. When a choice architect relies on our susceptibility to framing or ordering effects, she may be trying to influence our prefe rences or behavior without providing us with any new reasons to make the desired choice. Though I have argued that the use of defaults does not always involve preference shaping, it sometimes might. Thaler and Suns tein suggest that the default may sometimes establish a baseline for eval uation, relative to which the disadvantages of other options loom espe

21Sunstein and Thaler suggest that a choice architect may set a default by considering the choice that most people would make if required to choose ("Libertarian Paternalism is Not an Oxymoron," p. 1194).

22Thaler and Sunstein, Nudge, pp. 1-3.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

624 Jason Hanna

cially large.23 The default may also come to represent a norm from which people are reluctant to depart.24 If the selection of a default influences people in these ways, it may affect their preferences without providing them with any new information relevant to their choice. Indeed, many noncoercive strategies of influence may rely at least in part on their pre ference-shaping effects. Nonetheless, since means of influence that rely on framing or ordering effects and default rules appear to be the most like

ly targets of the manipulation objection, I will focus throughout on these strategies. In the interest of clarity, I will continue to refer to them as in stances of preference-shaping paternalism and eschew the more expan sive (and potentially misleading) "libertarian paternalist" label.

Having clarified the nature of preference-shaping paternalism, we can now go on to consider whether it is manipulative. In arguing that it is, many critics contrast manipulation with rational persuasion. Their thought appears to be that preference-shaping paternalism is objectiona ble insofar as it influences a person's choices through means that do not appeal to his reason. The following criticisms are representative:

The reason why nudges such as setting defaults seem ... to be paternalist, is that in addi tion to or apart from rational persuasion, they may "push" individuals to make one choice rather than another. Their freedom, in the sense of what alternatives can be chosen, is virtually unaffected, but when this "pushing" does not take the form of rational persua sion, their autonomy—the extent to which they have control over their own evaluations and deliberation—is diminished.25

Such steering policies [of the sort recommended by Thaler and Sunstein] are manipula tive, first because the government employs them with the intention of affecting people's choices. Secondly because they deliberately circumvent people's rational reasoning and deliberating faculties, and instead seek to influence their choices through knowledge of the biases to which they are susceptible.26

[Libertarian paternalism] may seem worse [than more traditional forms of paternalism] because the target's view of what is to be done is no longer merely ignored or thwarted, but is rather viewed as something one may attempt to extinguish, supplant, or "manage."27

As the first passage makes clear, the problem cannot be that preference

23Sunstein and Thaler, "Libertarian Paternalism is Not an Oxymoron," p. 1181. 24Consider default rules regarding organ donation. If donation is presumed to be the

default, actively opting out may have the flavor of an explicit refusal to do one's part in supporting a system from which one would profit if the need were to arise. If non donation is presumed to be the default, not actively opting in may have more the flavor of a decision not to do something supererogatory for the potential benefit of others.

25Hausman and Welch, "Debate," p. 128. 26Griine-Yanoff, "Old Wine in New Casks," p. 636. The language of "circumvention"

is also used by Hausman and Welch, "Debate," p. 130. 27Coons and Weber, "Introduction: Paternalism—Issues and Trends," p. 23, emphasis

in original.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

Libertarian Paternalism, Manipulation, and Preferences 625

shaping paternalism necessarily precludes attempts at rational persua sion. After all, one could both adjust the choice architecture and also at tempt to persuade a person to choose what one regards as the best option.

The manipulation objection is perhaps best interpreted as claiming that preference-shaping paternalism is problematic because its practition ers do not rely exclusively on the techniques of rational persuasion.28 When a person is susceptible to framing, for instance, the factor that makes a difference to her choice is the way in which the options are de scribed. And in framing the options in one way rather than another, the choice architect seems to appeal to her target's irrational foibles, not her reason. Utilizing these foibles, even in the service of a worthy goal, may seem manipulative. The manipulation objection to preference-shaping paternalism might thus be summarized as follows:

(1) If a person makes a certain choice, but would have made a different choice if the options had been presented or structured in a different but equally informative way, then her rational faculties have been bypassed or circumvented.

(2) It is objectionably manipulative to intentionally bypass or circum vent someone's rational capacities.

(3) Therefore, it is objectionably manipulative to intentionally shape choice through the design of choice architecture.

Note that this argument maintains only that there is something especially problematic about intentional shaping.29 Therefore, it is not jeopardized by the claim that preference-shaping influence is sometimes inevitable. The argument also leaves open the possibility that manipulation is some times permissible all things considered; it maintains only that there is a (presumably nontrivial) pro tanto objection to manipulative influence.

2. Preference Shaping: Rational Persuasion or Manipulation?

As we have seen, most defenders of the manipulation objection draw a contrast between rational persuasion and preference shaping. The in tended contrast is not as stark as it may initially appear, however. Con sider an illustration. In their critique of libertarian paternalism, Daniel Hausman and Brynn Welch argue that it is not paternalistic to place pla cards containing nutritional information in a cafeteria line; such behavior

28Mark D. White, The Manipulation of Choice: Ethics and Libertarian Paternalism (New York: Palgrave Macmillan, 2013), p. 90.

29See J.S. Blumenthal-Barby, "A Framework for Assessing the Moral Status of 'Ma nipulation' in Coons and Weber (eds.), Manipulation, pp. 121-34, at pp. 125-26.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

626 Jason Hanna

"would not be aiming to shape how the [diners] choose or to limit what can be chosen."30 But of course there are different ways in which a cafe teria could provide nutritional information. Each placard might include the nutritional information for just one entrée, or placards could include information for multiple entrées, perhaps to encourage direct compari son. The placards could be placed directly in front of each item, or they could be displayed only at the beginning of the line. A small symbol (perhaps a heart) might be placed on some placards to identify items that are especially healthy. Diners may pay more attention to the placards if they are printed on brightly colored paper, or placed in aesthetically ap pealing frames, and so forth. Is it manipulative for the cafeteria owner to design and arrange the nutritional placards in the way that is most likely to encourage customers to make healthy choices?

Suppose first that the answer is "no." One might argue that if it is not objectionably manipulative for the cafeteria owner to provide her custom ers with nutritional information, then it is not objectionably manipulative for her to design the nutritional labels with an eye to encouraging healthy choices. That may be right. The problem for defenders of the manipula tion objection, however, is that we might just as well claim that if it is not objectionably manipulative for the cafeteria owner to display food for sale, it is not objectionably manipulative for her to arrange the food with an eye to encouraging healthy choices. Likewise, we might claim that if it is permissible for someone to provide others with information or to set a default, it is permissible for her to frame the information or set the de fault in a way that will promote desirable outcomes. In short, answering "no" to the above question seems to condone the sorts of preference shaping measures that the manipulation objection was designed to rule out.

Suppose then that we answer "yes": while it is not manipulative for the cafeteria owner to provide her customers with nutritional informa tion, it is manipulative for her to deliberately choose to display the in formation in ways that will encourage healthier eating.31 While this re sponse is consistent with the manipulation objection, it seems implausi ble. That is, it seems implausible to claim that although it is commenda ble to inform people, it is objectionable to consider their interests in de ciding precisely how to inform them. Consider in this regard govern

30Hausman and Welch, "Debate," p. 129. Although providing information can ob viously influence choice, Hausman and Welch do not believe that it shapes choice in the supposedly problematic way they have in mind.

3ISunstein has distinguished between disclosure policies that provide "just the facts" and disclosure policies that go "beyond the facts" to persuade people (Why Nudge? p. 139). He gives graphic cigarette warning labels as an example of the latter. While Suns tein believes that both types of disclosure policy can be permissible, the reply we are considering rejects "beyond the facts" disclosure.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

Libertarian Paternalism, Manipulation, and Preferences 627

ment-mandated warning labels on cigarette packaging. Some critics have argued that labels including gruesome images, such as diseased lungs, are objectionably manipulative.32 But even if we think that warnings should include only text, we still have to ask how large the labels should be, how the warnings should be worded, whether the warnings should be printed on each package or merely displayed at the point of sale, and, if the former, where the labels should be placed on the cigarette packaging. And so on. While each set of labeling requirements may succeed in pro viding smokers with information, they may have differential effects on smoking behavior. On the face of it, the most satisfying basis for decid ing among the different sets of requirements would appeal to the interests of smokers and prospective smokers. There is a more general lesson to be drawn here. It is tempting to suppose that one person manipulates another when she uses nonrational means to influence the other's beha

vior. On closer inspection, however, this view is too simple (at least if manipulation is thought to be objectionable). The size or color of a warn ing label does not provide one with reasons to refrain from smoking. In a rather clear sense, then, a policy that requires very large or brightly co lored labels relies on nonrational means of influence. Nonetheless, if large or brightly colored warning labels would deter some people from smoking, it seems odd to think that we should not consider this fact in deciding on a set of labeling requirements.33 Why is this?

In my view, a particular means of influence does not objectionably manipulate a person if it encourages her to appreciate and act on the ba sis of good reasons. Appreciating a reason involves more than merely knowing some fact.34 Someone might know that smoking increases her risk of lung cancer or that regular consumption of trans fats is associated with heart disease. But she might not fully appreciate the reasons pro vided by these facts: she may fail to apply the risks appropriately to her own case, she may ignore the risks or try not to think much about them, and she may fail to vividly imagine the severity of the risked outcomes.35 More generally, an agent fully appreciates a reason only when she assigns

32Grüne-Yanoff, "Old Wine in New Casks," p. 637. For the opposite view, see Joel Feinberg, "Legal Paternalism," reprinted in Rolf Sartorius (ed.), Paternalism (Minneapo lis: University of Minnesota Press, 1983), pp. 3-18, at p. 11.

33For a similar point, see Grill, "Expanding the Nudge," p. 144. 34Joel Feinberg, Harm to Self (New York: Oxford University Press, 1986), p. 278;

and Jason Hanna, "Paternalism and Impairment," Social Theory and Practice 37 (2011): 434-60.

35Sunstein, Why Nudge? chap. 1; and Julian Savulescu, "Rational Desires and the Limitation of Life-Sustaining Treatment," Bioethics 8 (1994): 191-222, esp. pp. 194-99. For discussion of this point as applied to smoking in particular, see Robert Goodin, "No Smoking," Ethics 99 (1989): 574-624, pp. 582-83.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

628 Jason Hanna

it the appropriate weight in her deliberations. Some philosophers would argue that a reason's appropriate weight is determined by the agent's own goals, desires, and values. Others would argue that it is determined in some more objective way. In many of the cases that concern us, how ever, these views will converge. Consider someone who greatly values longevity and desires to be active into old age. If brightly colored ciga rette warning labels or images of diseased lungs would encourage such a person to better attend to the life-shortening effects of smoking, then the use of these means of influence encourages her to appreciate the reasons she has not to smoke.36 It is not manipulative to encourage a person to better appreciate reasons relevant to her behavior.

Once the notion of appreciation is spelled out, however, it seems like ly that many preference-shaping measures can be plausibly claimed to help people appreciate reasons. As the discussion of warning labels sug gests, an agent might encourage a target to better respond to reasons by utilizing an apparently arbitrary feature of the target's psychology. In defending the use of framing in our medical treatment example, for in stance, someone might claim that the mortality rate description scares people and inhibits their ability to rationally consider their options, or that the survival rate description helps them better attend to the proce dure's long-term benefits. In some cases, preference-shaping paternalism is aimed at noncoercively encouraging decision-makers to consider and take to heart the consequences of their behavior.

Likewise, I suspect that defaults at least sometimes encourage people to appreciate reasons. When defaults shape preferences, they may do so by making the reasons in favor of the default option more salient, or by conveying the impression that decision-makers should depart from the default only if they can identify good reasons to do so. If the selection of a default influences people in these ways, then it may work by getting people to appreciate the reasons to choose the default: if the default really is a good option for most people, then there is nothing objectionable about placing it in a light in which it strikes most people as a good option.

Return to the simple two-premise statement of the manipulation ob jection, described at the end of the previous section. The discussion to this point shows that we should reject premise (1)—the claim that influ ence attributable to choice architecture necessarily bypasses or circum vents the decision-maker's rational faculties. In some cases, apparently arbitrary features of the choice situation can engage the agent's rational capacities. Indeed, this point appears to be borne out by common expe rience. For instance, while it does seem manipulative for an agent to speak in a slow, measured voice in order to lull her target into childlike

36Cohen, "Nudging and Informed Consent," p. 8.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

Libertarian Paternalism, Manipulation, and Preferences 629

compliance, it does not seem manipulative for someone to use the same tone of voice in order to help an overly excited decision-maker calm down and think more clearly.37 In the latter case, an apparently nonra tional means of influence (namely, tone of voice) actually encourages reasoned deliberation. As I have argued, the same point can be made about many uses of preference-shaping paternalism. In order to know whether a means of influence is manipulative, we need to know not merely whether it is non-rational, but also how it affects the target's deli beration.

3. An Account of Manipulation

I have so far argued that many uses of preference-shaping paternalism are not manipulative, insofar as they encourage the target to appreciate reasons. This argument only partially rebuts the manipulation objection, however, since it fails to vindicate some preference-shaping paternalist measures supported by Thaler, Sunstein, and others. Return to the cafete ria example. By displaying healthier foods at the beginning of the line, the cafeteria owner does not encourage her customers to appreciate rea sons. Instead, the customers' susceptibility to her influence is perhaps best explained by their reliance on a decision-making strategy that is less than fully rational. Instead of surveying all available items in the line and then selecting what they regard as the best options, the customers proba bly consider on a case-by-case basis whether each item down the line is sufficiently good and stop taking food once their trays are full.38 Instead of attempting to rid her customers of their irrational bias in favor of foods displayed earlier in the line, the cafeteria owner exploits this bias in order to benefit them. Even if everything I have said so far is correct, the cafeteria owner may manipulate her customers. Moreover, if the cafe teria owner does manipulate her customers, one might suspect that many other forms of preference shaping are likewise manipulative, insofar as they too make use of irrational biases.

I do not believe that the cafeteria owner is guilty of manipulation. In order to explain why, however, it will be useful to have at least a rough account of manipulation. I develop such an account in the remainder of this section. In the next section, I apply the account to the cafeteria ex ample and similar cases. To be clear, I will not attempt to offer a concep tual analysis that captures all of the ways in which we use the term "ma nipulation." There may be many different types of manipulation, a full

37I am indebted to Danny Scoccia for this example. 38I am indebted to Eric Chwang for helpfully describing the cafeteria case in this way.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

630 Jason Hanna

survey of which lies outside the scope of this paper. Moreover, as I have indicated, my concern is only whether there is a type of manipulation that grounds an objection to preference-shaping paternalism.

Given my aims, one sensible way to proceed is to consider several paradigm cases of noncoercive, nondeceptive manipulation that appears to make use of targets' psychological foibles or irrational quirks. If we can provide an account of the way in which the behavior at issue in these cases is manipulative, then we can fruitfully apply the account to the pre ference-shaping measures that are our concern. Of course, this argumen tative tactic has its limits: even if the account thus developed does not imply that preference-shaping paternalism is manipulative, there may be an alternative or broader account that does. Nonetheless, if there is a way to accommodate our intuitions about certain paradigm cases of manipula tion without condemning preference-shaping paternalism, the manipula tion objection is likely to lose much of its force.

Let us begin, then, by considering three paradigm cases of manipulative behavior that appeals to the target's psychological foibles or irrational bi ases. One was discussed in the previous section: imagine that an agent uses a certain tone of voice because she knows that it will deter the target from thinking on her own. As a second case, consider Eric Cave's example in which a reluctant theatergoer gets his partner to miss a play "by engaging [his partner's] sublimated compulsive tendency to check the stove when [they] are halfway to the theatre."39 Finally, imagine that an agent at tempts to get her target to do what she wants by appealing to the target's "mother complex"—for instance, by telling the target that his mother will be ashamed unless he complies with the agent's request.40 Each of these cases involves manipulation. What do they have in common?

On the account I favor, manipulation—or at least the type of manipu lation at issue in these cases—involves the intentional use of nonrational

means of influence, where these means affect the target's deliberation for the worse.41 As a result of (successful) manipulation, the target acts on the basis of bad reasons, or fails adequately to reflect on reasons, or attends to

39Eric Cave, "What's Wrong with Motive Manipulation?" Ethical Theory and Moral Practice 10 (2007): 129-44, p. 132.

40I owe this example to Danny Scoccia. 41There is an interesting question about precisely how nonrational means of influence

are to be distinguished from rational means of influence. A full discussion of this issue would take us too far afield, so I will largely rely on our intuitive judgments. In general, if an agent provides someone with information or states reasons, and the agent intends that the other be influenced merely through her awareness of this information or these reasons, then the agent has not intentionally used nonrational means. If the agent intends that her target be influenced by other factors—e.g., the way in which the information is presented—then she has intentionally used nonrational means.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

Libertarian Paternalism, Manipulation, and Preferences 631

irrelevant considerations, or places more (or less) weight on certain con siderations than those considerations properly merit.42 Thus, in the first case above, the agent intentionally takes steps to short-circuit the target's independent deliberation. In the second case, the agent mobilizes the tar get's irrational worry. The third case involves manipulation because the target's feelings of filial guilt do not provide him with good reasons to do what the agent wants.43 The common feature of these cases, in short, is that the agent deliberately influences the target, with the result that the target's deliberative process is worsened. This claim raises a complica tion, however. There are some cases in which an agent might worsen her target's deliberation in some respect without thereby worsening her tar get's deliberation on balance. For instance, suppose that an agent encou rages the target to act for bad reason Rb when that target otherwise would have acted for equally bad reason R2. While such an agent may not worsen the target's deliberation all things considered, she does wor sen the target's deliberation in some respect—namely, by encouraging the target to give more weight to Ri than she otherwise would have. I am inclined to say that such behavior is manipulative. If an agent does not worsen the target's deliberation in any respect, however, then on my ac count she has not manipulated the target.

Aside from its ability to classify certain paradigm cases of manipula tion in the intuitively correct way, the rough account just described has at least two virtues. First, it accommodates the fact that there is often no unqualified answer to the question of whether some particular type of influence is manipulative. Consider in this regard appeals to emotion. It is usually manipulative to play on a person's irrational fear. But this does not mean that emotional appeals are generally manipulative. If an agent attempts to engage someone's appropriate feelings of sympathy in order to encourage benevolent behavior, the agent may improve her target's deliberative position by rendering him more responsive to reasons.44

Second, my account appears to provide an explanation of why ma nipulation is wrong. Plausibly, people have an interest in deliberating under conditions in which they are likely to respond appropriately to rea sons. If an agent prevents someone from deliberating under such condi tions, it would appear that the agent has frustrated one of that person's interests. This point can perhaps also be cast in terms of autonomy,

42For a similar account, see Robert Noggle, "Manipulative Actions," American Philo sophical Quarterly 33 (1996): 43-55. I express uncertainty about some details of Nog gle's view below, however.

43Marcia Baron notes, in a similar vein, that it can be manipulative to offer induce ments when doing so gets the target to agree to one's proposal for "the wrong sort of reason." See Baron, "Manipulativeness," p. 41.

"^Noggle, "Manipulative Actions," p. 49.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

632 Jason Hanna

depending on how we understand that concept. Some philosophers argue that autonomy involves responsiveness to reasons.45 It might thus be plausible to claim that manipulation offends against autonomy, insofar as it evinces a "failure of respect for [the target] as a person capable of res ponding appropriately to identifiable reasons."46

I have argued that manipulation involves the intentional use of non rational means that affect the target's deliberation for the worse. While this account seems promising, it raises an important question: what is the baseline relative to which manipulation worsens the target's deliberative situation? In his illuminating discussion of manipulation, Robert Noggle characterizes manipulation as "the attempt to 'adjust someone's internal levers'—that is, someone's belief, emotion, or desire—away from their ideal settings."47 The relevant settings or standards, according to Noggle, are those that the manipulator regards as ideal. By adopting this view, Noggle hopes to preserve a connection between manipulation and decep tion: whether someone has lied depends on what she believes, and whether someone has manipulated her target, on Noggle's view, depends on the deliberative conditions that she regards as ideal. This view can be contrasted with two alternative views. One alternative maintains that the

manipulator attempts to distort the target's deliberation relative to the target's own ideals—or perhaps better yet, relative to what the manipula tor believes to be the target's ideals.48 A second alternative maintains that the relevant standards are objective, or independent of both the manipu lator's and the target's beliefs.

As I will argue in the next section, each of these three proposals—the manipulator-relative baseline, the target-relative baseline, and the objec tive baseline—will end up supporting my claim that the cafeteria exam ple and similar cases do not involve manipulation. Nonetheless, it is worth briefly considering which proposal is most plausible, partly be cause the appeal of the general account of manipulation described above depends on whether an adequate baseline can be found, and partly be cause different baselines will yield different implications about at least some cases. In the remainder of this section, I will provide (somewhat

45George Sher, Beyond Neutrality (New York: Cambridge University Press, 1997), pp. 48-56.

46Patricia Greenspan, "The Problem with Manipulation," American Philosophical Quarterly 40 (2003): 155-64, p. 162.

47Noggle, "Manipulative Actions," p. 47. 48Noggle (ibid.) considers but rejects a similar view. One of Noggle's concerns is that

if we define the baseline relative to the target of manipulation, an agent might manipulate someone without knowing it. In order to avoid this result, however, we need not relativ ize the baseline to the manipulator's ideals; instead, we could relativize it to the manipu lator's beliefs about the target's ideals.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

Libertarian Paternalism, Manipulation, and Preferences 633

tentative) support for the objective baseline. One problem for Noggle's manipulator-relative baseline is that it

seems underinclusive. Consider the following example:

Racist Candidate: In order to gain an edge over his opponent, a political candidate runs a set of campaign ads designed to subtly play on voters' subconscious racist fears. The candidate is himself thoroughly racist, and he believes racist fear to be wholly appropriate.

Noggle claims that the candidate has not manipulated the voters, since he has not moved them away from the ideals that he regards as appropriate.49 I am not so sure. Perhaps the candidate has not attempted to manipulate the voters, but there is a strong case that the voters have nonetheless been manipulated.50 In case this is not clear, consider a further example:

Sincere Cult Leader. A religious cult leader recruits followers through the sorts of apparently manipulative (but not quite coercive) tactics often employed by such groups. The cult leader sincerely believes that his tac tics improve the recruits' capacity to see the truth and make prudent de cisions about whether to join the group.

The cult leader directs his recruits toward the deliberative standards that

he regards as ideal. But this fact alone seems insufficient to acquit him of the charge of manipulation.

Consider now the view according to which the baseline is set relative to the target's own beliefs about the appropriate deliberative standards. This view also appears to be underinclusive. Consider the following example:

Unacknowledged Compulsive Behavior. Albert has a compulsive tenden cy to return home from outings to ensure that his stove is turned off. He does not, however, regard this tendency as compulsive or irrational; in stead, he identifies with it and regards it as an indication of his prudence. (For instance, we can imagine that if Albert were offered a pill that would rid him of his obsessive compulsive behavior, he would decline to take it.) While Albert is driving Betty to a play that she does not particu larly want to see, she reminds him that he has not yet returned to check the stove. Albert returns home and they miss the play.51

49Noggle, "Manipulative Actions," p. 50. 50Compare the case of manipulative seduction carried out by a thoroughgoing miso

gynist, as discussed in Eric Cave, "Unsavory Seduction and Manipulation," in Coons and Weber (eds.), Manipulation, pp. 176-200, at p. 199.

5'This is a variant on the case discussed in Cave, "What's Wrong with Motive Mani pulation?" p. 132.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

634 Jason Hanna

By deliberately appealing to Albert's obsession with his stove, Betty ac tually leads him closer to his own deliberative standards (and we can suppose that she knows this). The target-relative baseline would thus imply that she has not manipulated Albert. But it seems that she has.

So far, I have argued that each of the manipulator-relative baseline and target-relative baseline is underinclusive. If this were the only prob lem, someone might still hope to accommodate the counterexamples by adopting a disjunctive account: for instance, perhaps manipulation in volves the attempt to move the target away from standards that either the manipulator or the target regards as ideal, or perhaps manipulation moves the target away from the target-relative or objective standards. One problem for such disjunctive accounts, however, is that the baselines considered so far seem overinclusive as well as underinclusive. Consider

some variations on Unacknowledged Compulsive Behavior. Suppose that Albert is at a movie with his friend Cora when he begins to worry about his stove. Cora knows that the only way to assuage Albert's worry and get him to stay for the rest of the movie is to tell him, in a calm but stern voice calculated to have the desired effect, that the stove will be fine. Because Albert identifies with his compulsive tendency to check the stove, the target-relative baseline (and any account that includes this baseline as a disjunct) will imply that Cora has manipulated Albert. Fur ther, if Cora believes that Albert's compulsive tendencies are rational— perhaps she also worries about stoves but simply wants to see the rest of the movie—then the manipulator-relative baseline will imply that she has manipulated Albert. Whatever we might think of Cora's behavior, how ever, it does not seem that she manipulates Albert by subduing his obses sive compulsive tendencies, particularly if she does so out of concern for him. It seems odd to think, on the one hand, that it can be manipulative to deliberately activate a person's compulsive worry (as in our original ex ample) and, on the other hand, that it can be manipulative to deliberately reduce a person's compulsive worry.

We are thus left with a final view, according to which the manipulator worsens the target's deliberative situation relative to some objective standard.52 This view implies that Betty manipulates Albert (and that Co ra does not), since groundless worry is not ordinarily a good basis for choice. The objective baseline view also yields the intuitively correct judgments about Racist Candidate and Sincere Cult Leader, since the agents in these cases induce their targets to act for bad reasons. The ob jective baseline view gets the right result in these cases because of a key advantage it has over its competitors: it does not make the truth of a ma

52A version of the objective baseline view is endorsed in Anne Barnhill, "What is Manipulation?" in Coons and Weber (eds.), Manipulation, pp. 51-72, esp. pp. 66-68.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

Libertarian Paternalism, Manipulation, and Preferences 635

nipulation claim depend on anyone's (perhaps bizarre or idiosyncratic) beliefs about the appropriate deliberative standards. Taken together, the examples we have considered suggest that an agent can manipulate her target by moving him toward deliberative standards that she (or the tar get) regards as ideal, and that an agent can fail to manipulate a target even though she moves the target away from deliberative standards that she (or the target) regards as ideal. Only the objective baseline view easi ly accommodates the judgment that an agent (or target, or both) can be mistaken about whether a particular behavior exerts inappropriate or ma nipulative influence.

As we have seen, Noggle endorses the manipulator-relative baseline to preserve the parallel between deception and manipulation: whether an agent has lied depends on what she believes. Yet the application of many other moral notions does not depend so neatly on anyone's beliefs. For instance, we do not believe that a transaction is exploitative merely be cause one party to the transaction believes that it is unfair. Likewise, to determine whether a proposal is coercive, we would want to know whether the proposer has a right to make the proposal or whether the re cipient has a reasonable alternative to acceptance; it is not sufficient merely to consult either party's beliefs about these matters.53 The objec tive baseline view holds that something similar is true of manipulation. To be sure, when an agent leads her target away from deliberative stan dards that the agent mistakenly believes to be ideal but toward delibera tive standards that actually are ideal, the agent may attempt to manipulate her target, even if she does not actually manipulate her target. (For this reason, it may be accurate to characterize the agent's behavior as in some sense manipulative.) Conversely, when an agent leads her target away from deliberative standards that actually are ideal but toward deliberative standards that the agent mistakenly believes to be ideal, the agent may actually manipulate her target even if, under some circumstances, she is not culpable for doing so.54 According to the objective baseline view,

53See Alan Wertheimer, Coercion (Princeton: Princeton University Press, 1987). 54It may appear that proponents of the objective baseline view are committed to a

form of externalism about reasons. This appearance is misleading. To see why, we should distinguish two questions. First, can a person have a reason to do something that he can not be motivated to do? Second, can we assess the deliberative standards a person accepts by comparing them to some objective ideal? Externalism is an attempt to answer the first question, while the objective baseline is an attempt to answer the second. To see that these questions are indeed distinct, notice that someone might accept a set of deliberative standards that is not geared toward maximally satisfying his (informed, ideal) desires. For instance, someone might make decisions by consulting a book that he believes to be holy, even if this procedure often prevents him from doing what would best satisfy his (in formed, ideal) desires. I am grateful to Danny Scoccia for pointing out the need to clarify my view on this issue.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

636 Jason Hanna

there is good reason to separate the question of whether manipulation has occurred from questions of culpability.55

Finally, it is worth noting that proponents of the manipulation objec tion ought to be sympathetic toward the objective baseline view. Paterna listic choice architects are unlikely to believe that they are manipulating their targets or leading their targets astray from the appropriate delibera tive standards. If anything, they are likely to believe that they improve their targets' deliberation (and even that they improve their targets' deli beration relative to the targets' own standards).56 Proponents of the ma nipulation objection would presumably claim that such beliefs by them selves are insufficient to show that preference-shaping paternalism does not involve manipulative influence. The objective baseline, unlike the manipulator-relative baseline, can accommodate this judgment.

Much more could be said about the nature of manipulation, and I do not pretend to have offered a decisive argument in favor of the objective baseline view. In defending this view, I hope mainly to have shown that there is a promising way of developing my rough account of manipula tion, according to which manipulation involves the intentional use of nonrational means of influence that worsen the target's deliberation.

4. Preference-Shaping Paternalism Reconsidered

The question now is what this account implies about the moral status of preference-shaping paternalism. I argued in section 2 that preference shaping paternalism sometimes encourages people to appreciate reasons and that, when it does, it is not objectionably manipulative. This claim seems to follow from the more general account of manipulation I have offered: in encouraging someone to appreciate reasons, the agent brings her closer to the ideal deliberative standards. As I noted at the beginning of section 3, however, the cafeteria case shows that some preference shaping measures championed by Thaler and Sunstein do not encourage people to better appreciate reasons. Instead, the cafeteria owner simply relies on pre-existing biases. Does this sort of behavior constitute mani pulation, according to my account?

It does not. While the cafeteria owner may not attempt to rationally persuade her customers, she does not in any respect worsen their deliber ative position, either. Regardless of how the food in the cafeteria line is arranged, customers will be biased in favor of items displayed closer to

55For the view that agents can manipulate others without knowing it, see Kate Manne, "Non-Machiavellian Manipulation and the Opacity of Motive," in Coons and Weber (eds.), Manipulation, pp. 221-45, esp. pp. 228-31.

56See Thaler and Sunstein, Nudge, p. 5.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

Libertarian Paternalism, Manipulation, and Preferences 637

the front. Displaying vegetables rather than cake first in line will not af fect customers' susceptibility to this bias, nor will it cause them to attend to bad reasons or irrelevant considerations that they otherwise would have ignored. It is not as though diners who go through the "healthy" cafeteria line make rational or deliberative errors that their counterparts in the "unhealthy" cafeteria line avoid. Similar claims can be made about the use of defaults. After all, the status quo bias will likely rear its head regardless of whether the default is set paternalistically (rather than in some other way). More generally, when an agent adjusts the "external" choice architecture without changing the target's "internal" motives or deliberative strategy, the agent does not lead the target away from any deliberative standard to which he otherwise would have conformed. The

agent's reliance on her target's psychological bias is not manipulative so long as the bias would have been operative in the same way even absent the agent's influence. Notice that this conclusion follows regardless of whether we accept the objective, manipulator-relative, or target-relative baseline. To be sure, some adjustments to the choice architecture may constitute manipulation. If I know that you have difficulty thinking clear ly while loud music is playing, it would be manipulative for me to turn on the stereo, at least if doing so is designed to influence you. But such behavior is manipulative precisely because it reduces your ability to en gage in clear-headed deliberation. The same cannot be said of cases in which an agent merely utilizes biases that would have been operative in any case.

At this point, however, a critic might object. Perhaps my account of manipulation implies that it is not manipulative to exploit a person's psy chological biases.57 But if so, the critic may argue, so much the worse for my account. Instead of concluding that my account shows that the cafete ria owner does not manipulate her customers, perhaps we should con clude that the manipulative character of the cafeteria owner's behavior shows that my account is mistaken.

There are two points worth making about this objection. First, insofar as the objection rests on the claim that preference-shaping paternalism "exploits" the target's psychological biases, it may seem plausible largely because it fails to distinguish two different senses of "exploitation." There is a nonmoral sense of exploitation, in which "to exploit" is simply

"This "exploitation" language is used in Hausman and Welch, "Debate," p. 130; Bovens, "The Ethics of Nudge," p. 209; Grüne-Yanoff, "Old Wine in New Casks," p. 637; and J.S. Blumenthal-Barby and Hadley Burroughs, "Seeking Better Health Care Outcomes: The Ethics of Using the 'Nudge'," The American Journal ofBioethics 12, no. 2 (2012): 1-10, esp. p. 5.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

638 Jason Hanna

"to make use of."58 It is in this sense that preference-shaping paternalism uncontroversially exploits its targets' psychological biases. There is also a moral sense of exploitation, in which to exploit someone is to take un fair advantage of her.59 Of course, one might make use of a person's psy chological biases in order to take unfair advantage of her—say, by get ting her to imprudently agree to a one-sided contract that greatly benefits oneself. But this concern does not seem to arise for preference shaping that aims to benefit its targets.

Second, while some critics object that preference shaping makes use of "imperfections in human judgment and decision-making," the same is true of a great deal of morally innocuous behavior.60 The cognitive biases exploited by preference-shaping paternalism are "imperfections" only in the sense that they are nonrational—the sorts of features that characterize Humans but not Econs, as Thaler and Sunstein would put it.61 By the same token, however, many traits of which we routinely and permissibly make use constitute imperfections. For instance, a project leader may recognize that her co-workers are likely to do better work if she adopts an enthusiastic attitude or speaks in an encouraging tone of voice; per haps enthusiasm is contagious. The co-workers' susceptibility to this form of influence is in some sense nonrational. But it does not seem that

they are manipulated (in any objectionable sense) by the project leader's deliberate attempts at enthusiasm. To take another example, the extent to which consumers find a product's logo or packaging appealing depends on nonrational foibles and biases (for instance, the positive associations arbitrarily tied to certain fonts or color schemes). Yet it seems implausi ble to think that companies objectionably manipulate consumers merely by deliberately designing appealing logos.

Before I turn to consider one final objection in section 5, let me bring together the various strands of the view I have developed. In order to determine whether A has manipulated B by using or relying on non rational means of influence, we should first ask what would have hap pened in the absence of ^'s influence. One possibility is that B would not have been subject to the same or comparable influence in the absence of A's efforts: A may, for example, activate a response or cognitive bias that otherwise would have been dormant. Whether A manipulates B in this sort of case depends on the effect of his influence. If ^4's deliberate use of

58Alan Wertheimer, Exploitation (Princeton: Princeton University Press, 1996), p. 5. Wertheimer quotes Richard Arneson's entry on exploitation in Lawrence C. Becker (ed.), Encyclopedia of Ethics (New York: Garland, 1992), p. 350.

59Wertheimer, Exploitation, esp. p. 10. 60The quoted passage is from Hausman and Welch, "Debate," p. 130. 61Thaler and Sunstein, Nudge, esp. pp. 6-8.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

Libertarian Paternalism, Manipulation, and Preferences 639

nonrational means encourages B to better appreciate reasons relevant to her behavior (or more generally if A moves B toward the ideal delibera tive standards), then A has not manipulated B. By contrast, if ^'s use of nonrational means encourages B to act on the basis of bad reasons, or attend to irrelevant considerations, or be moved by inappropriate emo tions, then A probably has manipulated B.

A further possibility, exemplified by the cafeteria case, is that B would have been subject to the same sort of influence absent A's delibe rate attempt (or any other agent's deliberate attempt) at shaping.62 In this sort of case, I do not believe that A manipulates B in any objectionable sense. This is because A's attempt at shaping does not in any respect worsen B's deliberation. The cafeteria customers are in no worse a deli

berative position, the owner having exercised her influence, than they would have been in if the owner had displayed items with no intention of influencing her customers' dietary choices (as she is surely permitted to do). Similar claims could presumably be made about the use of defaults and frames.

Finally, note that this account also implies that the cafeteria owner does not objectionably manipulate her customers if she places foods with the highest profit margin first in line. This result is significant. Retailers often arrange items and design choice architecture in ways that encour age consumers to spend more money. Confronted with this point, de fenders of the manipulation objection have two options. First, they could argue that nearly all preference shaping—whether carried out for pater nalistic reasons or not—is objectionable. A strong version of this re sponse would maintain that most commercial uses of choice architecture are impermissible. This view seems implausibly sweeping.63 A weaker version of the first response would maintain that while such commercial practices are pro tanto wrong, they are usually permissible all things con sidered. I am skeptical that it is even pro tanto objectionable for the cafe teria owner to more prominently display foods with the highest profit

62The parenthetical qualification may be necessary to deal with cases of overdetermi nation. Suppose that B is currently calm but that A stokes B's irrational fear. If A had not done so, however, C would have stoked B's irrational fear in precisely the same way. Strictly speaking, B may have been subject to the same sort of influence wielded by A even in the absence of ^'s involvement. Still, it may seem that ^'s behavior is objection ably manipulative.

63Allen Wood argues that virtually all advertisements—including those designed to encourage desirable behavior, such as quitting smoking—are objectionably manipulative ("Coercion, Manipulation, Exploitation," pp. 37-41). This conclusion seems radical and would be rejected by many critics of libertarian paternalism. For a plausible defense of commercial practices sometimes thought to be manipulative, see Robert L. Arrington, "Advertising and Behavior Control," Journal of Business Ethics 1 (1982): 3-12.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

640 Jason Hanna

margin.64 Moreover, if the pro tanto wrongness of manipulation can be easily overridden, it may turn out that preference-shaping paternalism is often permissible all things considered, even if it is manipulative. A second way of responding to common commercial uses of choice archi tecture maintains that only paternalistic preference shaping is objection ably manipulative.65 It seems odd, however, to think that paternalistic preference shaping is manipulative while nonpaternalistic preference shaping is not: after all, each involves the same means of influence. I sus pect that in order to defend an asymmetry between paternalistic and non paternalistic preference shaping, one would need to appeal to something like the putatively insulting character of paternalist motives.66 The claim that preference-shaping paternalism is insulting because it is prompted by paternalist motives, however, is distinct from the claim that preference shaping paternalism is manipulative. I have been addressing only the latter claim in this paper. But it is worth noting that if preference-shaping pater nalism is not manipulative, it may be unclear why it would be rendered objectionable by the motives that prompt it: many philosophers argue that presumptively objectionable paternalism must involve the use of coercion, manipulation, or other "unsavory" means.67 If I am correct, the sorts of measures championed by Thaler and Sunstein do not.

5. The Subliminal Messaging Objection

I turn now to a final concern about the putatively manipulative character of preference shaping. Put simply, the concern is that there is no prin cipled basis for distinguishing preference shaping from subliminal mes saging.68 Subliminal messaging seems objectionable largely because it bypasses our rational capacities without our awareness. But many critics have thought that the same point can be made about preference-shaping paternalism. After all, we are typically unaware of our sensitivity to

64Baron, "The Mens Rea and Moral Status of Manipulation," p. 114. 65White, The Manipulation of Choice, esp. chap. 6. A related view is that preference

shaping paternalism is objectionable only when carried out by governments. See Griine Yanoff, "Old Wine in New Casks," esp. p. 638.

66See Seana Shiffrin, "Paternalism, Unconscionability Doctrine, and Accommoda tion," Philosophy & Public Affairs 29 (2000): 205-50, esp. p. 220; Jonathan Quong, Libe ralism without Perfection (New York: Oxford University Press, 2011), pp. 100-106; and George Tsai, "Rational Persuasion as Paternalism," Philosophy & Public Affairs 42 (2014): 78-112, esp. p. 86.

67Danny Scoccia, "In Defense of Hard Paternalism," Law and Philosophy 27 (2008): 351-81, at p. 354.

68Hausman and Welch, "Debate," pp. 131-32; and Grüne-Yanoff, "Old Wine in New Casks," p. 638.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

Libertarian Paternalism, Manipulation, and Preferences 641

choice architecture, just as targets of subliminal messaging are unaware of the influence to which they have been subjected.

I have two replies. The first is that my account provides a basis for classifying some subliminal messaging as manipulative. Suppose that a particular use of subliminal messaging activates subconscious desires that its targets would repudiate. Or suppose that subliminal messaging inhibits its targets' abilities to think clearly and instead encourages them to act on "blind" urges. Subconscious desires and blind urges do not typ ically provide one with good reasons to act. Subliminal messaging of the sort just described would thus get people to act on the basis of bad rea sons. My account implies that such subliminal messaging is a form of manipulation. To be sure, there may be some cases of subliminal mes saging that my view would not condemn: perhaps someone could use subliminal messaging to remind a target of reasons relevant to her beha vior. Apparently a department store successfully decreased shoplifting by blending its background music with reminders that theft is illegal and immoral, in such a way that shoppers did not consciously notice the mes sage.69 In this sort of case, however, it is not clear to me that subliminal messaging is objectionably manipulative. (It should be noted that discus sion of this issue is complicated by the fact that it is not entirely clear how

subliminal messaging works, or even if subliminal messaging works.70) Suppose, however, that subliminal messaging is always objectionable

to at least some extent. If so, the reason may be that subliminal messag ing is nontransparent.71 This takes me to my second reply: subliminal messaging is nontransparent in a way that preference-shaping paternal ism typically is not. To prepare the way for this reply, it will be useful to consider a transparency condition that Luc Bovens endorses in his criti cal discussion of libertarian paternalism. Bovens distinguishes between two varieties of transparency: type interference transparency and token interference transparency J2 Suppose that a government announces in advance that it will sometimes use subliminal messaging. Such a policy satisfies the requirement of type interference transparency, since people know that they will sometimes be subjected to a certain type of influ ence. Such a policy does not, however, meet the requirement of token interference transparency, since people do not know, on any given occa sion, that they have been subjected to subliminal messaging. Bovens concludes that permissible preference shaping should meet a requirement of in principle token interference transparency: each instance of influ

69Arlington, "Advertising and Behavior Control," p. 4. 70Cave, "Unsavory Seduction and Manipulation," pp. 177-78. 7'Thaler and Sunstein, Nudge, pp. 244-45. 72Bovens, "The Ethics of Nudge," p. 216.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

642 Jason Hanna

ence "should be such that it is in principle possible for everyone who is watchful to unmask the manipulation."73

Bovens's view confronts some difficult questions, however. When we are confident that few people will in fact "unmask" the influence, it is not clear that it makes much difference that everyone could have unmasked the influence in principle. Suppose that television viewers can determine when a subliminal message is being flashed on screen by adjusting the settings on their televisions during commercial breaks. If few people ac tually adjust the settings (out of laziness, say), the mere fact that they could easily have done so may not make much moral difference.

There is a further distinction, however, that clarifies the way in which subliminal messaging, unlike the preference-shaping paternalism I have been defending, is nontransparent. We can distinguish between (a) a per son's lack of awareness of some particular thing that is in fact influen cing him, and (b) a person's lack of awareness that some particular thing is influencing him. The lack of awareness described by (a) is characteris tic of subliminal messaging, in which a message is displayed so briefly that the target is not consciously aware of it. The lack of awareness de scribed by (b), in contrast, is characteristic of much ordinary human inte raction. To return to an earlier example, consumers are often unaware of the ways in which they are influenced by the appearance of a company's logo, but it seems implausible to conclude that a company behaves objec tionably unless it tells consumers its reasons for adopting a particular design. Likewise, John F. Kennedy surely was not morally required to tell viewers of the 1960 presidential debate that they might be influenced by his famous blue suit, even if he wore the suit to impress voters. If we wish to condemn subliminal messaging but not these other sorts of influ ence, we should probably conclude that the lack of awareness described by (a) is more troubling than the lack of awareness described by (b). In deed, it seems clear that (a) precludes oversight in a way that (b) does not.74 The preference-shaping strategies I have been defending, however, usually involve a lack of awareness in the sense of (b), not (a).75 (The customers in the cafeteria line are aware of the display of food.) To be

73Ibid., p. 217. 74Bovens ("The Ethics of Nudge," pp. 216-17) seems especially concerned about the

extent to which citizens can monitor the government's behavior. 75Thaler and Sunstein consider a case in which an agent uses distorting mirrors to

encourage others to lose weight. They object that such influence is nontransparent. See Thaler and Sunstein, Nudge, p. 244; and Blumenthal-Barby, "Choice Architecture," pp. 190-91. Those influenced by the presence of the mirrors are presumably aware of the mirrors, however. On my view, it would be objectionable to use the mirrors because doing so would be deceptive: it would encourage people to act on the basis of false beliefs about their weight.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

Libertarian Paternalism, Manipulation, and Preferences 643

clear, I have not offered a decisive argument that the distinction between (a) and (b) is morally relevant. Then again, for reasons indicated earlier, I have not claimed that subliminal messaging (depending on how we un derstand it) is always objectionable. My point is only that if one wishes to condemn all subliminal messaging without thereby condemning a great deal of ordinary behavior, one may have good reason to appeal to the distinction between (a) and (b).

6. Conclusion

The manipulation objection has been one of the most popular criticisms of preference-shaping paternalism. My goal has been to show that this objection is not successful against many of Thaler and Sunstein's pre ferred methods of preference shaping, including the use of defaults, frames, and graphic images. I have argued that manipulation involves the intentional use of nonrational means of influence, where these means affect the target's deliberation for the worse. Preference-shaping pater nalism is not generally manipulative in this sense, for two reasons: first, it sometimes encourages people to appreciate the force of reasons rele vant to their behavior, and, second, it usually relies on biases that would have affected a person's deliberation in any case.

As I noted at the outset, however, I have been concerned only about a common nonconsequentialist objection to preference-shaping paternal ism. I have not considered the extent to which particular preference shaping measures might be counterproductive, either because they fail to achieve their stated aims or because, taken in combination, they leave us less self-reliant or more susceptible to government influence.76 With the failure of deontological objections, questions about the consequences of various proposed policies—for instance, about the ratio of benefit to harm and the distribution of benefit and harm—become all the more

pressing. But that is as it should be.77

Department of Philosophy, Northern Illinois University [email protected]

76Bovens, "The Ethics of Nudge," pp. 214-15; and Blumenthal-Barby, "Choice Ar chitecture," p. 187.

77For helpful feedback, I am grateful to Eric Chwang, Peter de Marneffe, an anonym ous referee, and especially Kalle Grill, Danny Scoccia, and the fellow participants of the 2014 Umeâ Workshop on Respecting Context-Dependent Preferences.

This content downloaded from ��68.59.124.36 on Fri, 17 Nov 2023 17:05:56 +00:00��

All use subject to https://about.jstor.org/terms

Contents

p. [618]
p. 619
p. 620
p. 621
p. 622
p. 623
p. 624
p. 625
p. 626
p. 627
p. 628
p. 629
p. 630
p. 631
p. 632
p. 633
p. 634
p. 635
p. 636
p. 637
p. 638
p. 639
p. 640
p. 641
p. 642
p. 643

Issue Table of Contents

Social Theory and Practice, Vol. 41, No. 4 (October 2015) pp. 577-772

Front Matter
Introduction [pp. 577-578]
Looking for a Psychology for the Inner Rational Agent [pp. 579-598]
How Nudges Often Fail to Treat People According to Their Own Preferences [pp. 599-617]
Libertarian Paternalism, Manipulation, and the Shaping of Preferences [pp. 618-643]
Nudging for Liberals [pp. 644-667]
Nudge and Shove [pp. 668-691]
Respect for What? Choices, Actual Preferences, and True Preferences [pp. 692-715]
Context-Dependent Preferences and the Right to Forgo Life-Saving Treatments [pp. 716-733]
Mill's Circle(s) of Liberty [pp. 734-749]
Book Reviews

Review: untitled [pp. 750-754]
Review: untitled [pp. 755-758]
Review: untitled [pp. 759-764]
Review: untitled [pp. 764-769]

Books Received [pp. 770-771]
Index to Volume 41 [pp. 772-772]
Back Matter