1. A MILLIAN VIEW OF PATERNALISM

John Stuart Mill's On Liberty is often read as the classic text on paternalism, but its comments on paternalism are part of its argument for individual liberty rather than an analysis or theory of paternalism. Mill does not use the term ‘paternalism’ in On Liberty or in any of his other writings.Footnote ¹ The closest he gets terminologically is when he speaks of ‘the system of despotic, or what is called paternal, government’ (1978 [1859]: 101). As this phrase suggests, Mill is concerned about the power that government and social norms exert over individuals, and he objects to the use of that power except where directed toward the protection of other people. He thus condemns restrictions on a person's freedom either for the person's own good or because the person's private behaviour (such as his own unusual relationship with Harriet Taylor) offends others. There are other social objectives that Mill does not mention, such as avoiding cruelty to animals, preserving national monuments, or protecting places of natural beauty and interest. As Mill would, I believe, concede, these may also justify limiting liberty. His concern is to defend liberty from paternalistic and moralistic restrictions. Whether a society restricts homosexual behaviour on the grounds that it is condemned by the Bible, or whether it does so on the (spurious) grounds that homosexual relations are harmful to the individuals involved, what is of concern to Mill is the limitation on individual freedom. Mill is not concerned about whether there might be paternalistic actions that do not limit liberty. Instead of offering a theory of what we now call ‘paternalism’, Mill condemns intrusive limits on liberty that have either a moralistic or a paternalist basis.

Paternalism might turn out to be nothing other than the paternalistic coercion Mill was concerned about. For example, Gerald Dworkin writes (Reference Dworkin1972: 65) ‘By paternalism I shall understand roughly the interference with a person's liberty of action justified by reasons referring exclusively to the welfare, good, happiness, needs, interest or values of the person being coerced’. Richard Arneson agrees that paternalistic policies restrict liberty with the objective of benefitting the individual or individuals whose liberty is restricted and adds that to count as paternalistic, the policy must be carried out against the will of these individuals (1980: 471). In his recent (2016) revision of the ‘Paternalism’ entry in the Stanford Encyclopedia of Philosophy, Dworkin requires that to count as paternalistic an action or omission must interfere with the liberty or autonomy of the person whom the action aims to benefit. Depending on how one understands autonomy, this condition, despite harkening back to Mill, could represent a retreat from the view that paternalism limits liberty. The Millian view is roughly that paternalist actions directed to Charlie limit Charlie's liberty or autonomy against his will with the objective of benefitting him.

To simplify the discussion, I shall focus on government paternalism, which is bound to be more problematic, because government knows so much less about what is good for me than do my friends and family. However, much of the discussion applies equally to actions of individuals and of non-governmental groups. Limits to liberty may be implemented via constraints, punishments, costs, or deception, including withholding information. Confining someone who is suicidal so as to make it impossible for her to harm herself exemplifies physical constraint. Fining drivers who do not use seat belts limits liberty via punishment. Taxing harmful products coerces via costs. Informing would-be masturbators that they risk blindness limits liberty by means of deception, unless the homophobes spreading this false information are themselves misinformed, in which case conveying this falsehood would constitute a misguided effort at rational persuasion.

Millian paternalist policies that coerce individuals for their own benefit often have other consequences and other justifications. For example, requiring motor cyclists to wear helmets may reduce health-care costs that are borne by others, and taxing tobacco raises revenue. In some cases, the non-paternalistic grounds for a policy are far-fetched, while in others it may be uncertain whether the policy is truly paternalistic or merely happens to benefit those whom it coerces. One might suggest a counterfactual test. One possibility is to maintain that a policy is paternalistic only if it would not have been carried out if the policymaker had not believed that it would benefit the individuals whose freedom it limits. This test seems too restrictive. I suggest instead that a policy is paternalistic only if it would still have been carried out if there were no non-paternalistic reasons for it.Footnote ²

2. EXCEPTIONS AND ‘SOFT’ PATERNALISM

Although Mill in On Liberty apparently rules out all legal and social limits on freedom that are not grounded in the protection of others, he makes two much-discussed exceptions. He believes that contracts whereby an individual agrees to be a slave, no matter how freely they are entered into, should not be enforced. In that way, we protect the freedom of the individual who is attempting to surrender it (1978 [1859]: 101). Although preventing someone from selling themselves into slavery protects their future freedom, it limits their current freedom and appears to be an instance of paternalism that Mill approves of.

Second, Mill discusses the following case:

Mill claims that the limitation on liberty is only apparent, not real, on the grounds that the interference conforms to the true desires of the bridge-crosser. Joel Feinberg and others have interpreted Mill as maintaining that interfering with an action that an agent did not voluntarily undertake is either not paternalistic at all (Feinberg Reference Feinberg1971, esp. p. 112; Reference Feinberg1986: 15) or a different defensible ‘soft’ variety, in which coercion isn't really coercion when it only prevents someone from doing something involuntarily. As Feinberg puts it,

Arneson copes with cases such as these by maintaining that even though the crosser is ‘seized’ and ‘turned back’ these actions are not against the crosser's will (or are not known to be against his will) (Arneson Reference Arneson1980: 487). If preventing an action is against someone's will if and only if the action is voluntary, then soft-paternalistic actions turn out not to be paternalist at all. On this view, ‘hard’ or genuine paternalism interferes only with actions that the policymaker knows to be voluntary.Footnote ³

3. PATERNALISTIC POLICIES THAT DO NOT LIMIT LIBERTY

The term, ‘paternalism’ has only been in existence for a century and a half, and during that time it has changed its connotation from positive to negative. Given Mill's influence, it has been associated with limits on liberty (for example, Feinberg Reference Feinberg1986: 7). On the other hand, some contemporary philosophers (for example Shiffrin (Reference Shiffrin2000) and Le Grand and New (Reference Le Grand and New2014)) maintain that paternalism need not restrict liberty. How should one decide whether they are right? Is the task to uncover the ‘essence’ of paternalism? Shiffrin suggests that it is not enough to test definitions against intuitions (whose authority over the usage of philosophical jargon is, I suggest, questionable). In addition, she argues that ‘it is worthwhile to assess what is central in our normative reactions to paternalism and to employ a conception of paternalism that complements and makes intelligible our sense of paternalism's normative significance’ (Shiffrin Reference Shiffrin2000: 212).

The ‘normative significance’ of paternalism she has in mind is whether the government (as our agent) or we as individuals treat individuals with proper respect.

From such a perspective, as Shiffrin points out, refusals to aid as well as intrusive actions are sometimes intuitively paternalistic. For example, in refusing to help my son with his math homework, I am not violating his rights or limiting his liberty, but if I think he would be better off working out the difficulties on his own, I am attempting to benefit him by acting against his will.Footnote ⁴ Should we count this as paternalistic on the grounds of intuition, its partial satisfaction of the conditions in the Millian view, and the disrespect it shows toward my son's autonomy, or should we deny that it is paternalistic, on the grounds that there is no restriction on liberty. If what matters to us about paternalism are questions about the justification for limiting liberty, then we might deny that my refusal to help my son is paternalistic. If, on the other hand, we find that ‘what is distinctive and worrisome about paternalism’ is ‘this substitution of judgment and the circumvention of an agent's will’ (Shiffrin Reference Shiffrin2000: 213), then we should seek a definition of paternalism that counts my refusal to help my son as paternalistic. As Dworkin puts it, ‘What we must ascertain in each case is whether the act in question constitutes an attempt to substitute one person's judgment for another's, to promote the latter's benefit’ (1988: 123).

What should such a definition of paternalism say? Shiffrin argues that it need not require that the action be contrary to the will of the person whom it aims to benefit. Suppose, for example, that I want to protect my elderly father from the many fraudulent offers that would probably arrive in his email in-box daily. So I install spam filters on his computer, knowing that he has no idea how to access his ‘junk’ mail.Footnote ⁵ My action is not contrary to his will, but it is independent of it and circumvents it. I am substituting my judgement for his.

In the light of examples such as these, Shiffrin takes it as essential to paternalism that ‘the motive behind [the paternalistic action] is simply that the (putative) paternalist knows better than the agent, or may better implement, what the agent has authority for doing herself’ (Shiffrin Reference Shiffrin2000: 216).Footnote ⁶ As this statement suggests, Shiffrin counts an action as paternalist if it is directed toward assisting an agent to pursue an objective, even if achieving the objective does not contribute to the agent's well-being. Since she is concerned about the substitution of judgement and circumvention of an agent's will, it is not, from her perspective, important whether the intervention is directed toward what benefits the agent or toward the achievement of some other aim of the individual.

But if one is concerned about the limits on government action and one takes for granted that government has an obligation to contribute to the well-being of its citizens, not to the accomplishment of projects unrelated to their well-being, then one needs to modify Shiffrin's conception of paternalism. Some economists have maintained that welfare economics should be focused on enabling people to achieve their objectives rather than on promoting well-being (Sugden Reference Sugden2004; Vining and Weimer Reference Vining and Weimer2010), but for the purposes of this paper, I shall assume the more conventional view that the central goal of normative economics is to enable policy to enhance individual's welfare.

4. PATERNALISM FOR BEHAVIOURAL ECONOMICS

Although Le Grand and New follow Shiffrin in jettisoning the requirement that paternalist policies restrict liberty and in emphasizing the substitution of judgement, they insist that paternalist policies be directed toward the benefit of those whose judgement they challenge. Here is their definition:

• • address a failure of judgment by that individual

• • further the individual's own good. (Le Grand and New Reference Le Grand and New2014: 23)

Although this only states a sufficient condition, they mean the condition to be both necessary and sufficient.Footnote ⁷ The definition concerns only government interventions, because government actions are Le Grand's and New's concern; and it restricts paternalistic policies to those directed toward the benefit of the individual whose judgement fails. Le Grand and New require that there be a ‘failure of judgement’ on the part of the agent toward whom the paternalistic action is directed, while Shiffrin requires only that the would-be paternalist knows better (or thinks she knows better). By ‘failures of judgement’ I take Le Grand and New to refer to the foibles that behavioural economics investigates. In support of this interpretation, notice that they deny that mistakes based on false beliefs count as failures of judgement. In apparent contrast to Le Grand and New, Shiffrin allows that a paternalist intervention may respond to people's inability to do what they correctly judge to be best for themselves, rather than to a failure of judgement. In fact, Le Grand and New agree that paternalistic policy may aim to correct weakness of will, but they regard weakness of will as one kind of failure of reasoning or judgement.

For Shiffrin any advantage the would-be paternalist has over the individual subject to the intervention in terms of knowledge or ability to put that knowledge to work could serve to justify a paternalist policy that aims to facilitate the individual's projects. Le Grand and New, in contrast, insist repeatedly that interventions are paternalist only if the subject whom they aim to benefit manifests a ‘failure of judgement’. So, for example, Shiffrin's definition would classify as paternalist the policy of an atheist government that closes the nation's monasteries to improve the lives of those who have up to now been monks. The rulers believe that they know better than the monks what makes for a good life. By closing the monasteries, the government aims to ‘take over or control what is properly within the agent's own legitimate domain of judgment or action’ (Shiffrin Reference Shiffrin2000: 216). On Le Grand and New's definition, in contrast, if the commitment to a monastic life is not due to a failure in judgement but to a difference in values, then the closing of the monasteries would not only be unjustified: it would not be paternalistic. Although an implication of their definition, I am confident that this is not what Le Grand and New intend, because they regard policies that limit people's behaviour on the grounds that their ultimate objectives are unworthy as unjustified paternalism.

In addition, people often make bad choices because of false beliefs. Whether or not there may have been reasoning failures in the acquisition of those beliefs, reasoning from false beliefs may be impeccable. ‘Having little information in itself does not impair our ability to reason’ (Le Grand and New Reference Le Grand and New2014: 15). Le Grand and New go on to point out, correctly, that providing information to correct a false belief does not substitute the policymaker's judgement for the agent's judgement. As Mill himself argued, engaging in rational discussion with an individual is not paternalistic; it neither limits liberty nor shows disrespect for the individual's judgement. Interventions that merely supply information are thus ordinarily not paternalistic, and they are, let us suppose, the ideal way to benefit individuals who are doing a poor job of deciding for themselves, owing to their false beliefs.Footnote ⁸ But there are many other ways to intervene in people's lives to save them from making bad choices that result from their false beliefs. And there seems to be no good reason to maintain, as Le Grand and New's definition implies, that an intervention cannot be paternalistic if it is designed to benefit an individual who without any failure of judgement has embarked on a harmful course of action owing to a false belief.

The notion of a ‘failure in judgement’ or a ‘reasoning failure’ is crucial to Le Grand and New's view of paternalism, and it is this notion that ties their work to recent advances in psychology and behavioural economics. Their crucial insight is that if an intervention aims to correct what behavioural economics identifies as a reasoning failure (in order to benefit the person whose reasoning fails), then there is a case to be made in defence of the intervention. Justified paternalistic interventions depend on demonstrable reasoning failures rather than disagreements between ‘big brother’ and individual citizens concerning what objectives are best.

However, they get into trouble by building judgement failures into the definition of paternalism. It is a tempting move to make, because it enables them to give a persuasive account of the sense in which so-called ‘libertarian paternalism’ or nudging directed toward the benefit of those nudged is truly paternalism. Even though libertarian paternalism that avoids deception does not limit liberty and, on Mill's view, would fail to be paternalism at all, it is, on Le Grand and New's definition paternalist. Directed as it often is toward benefitting individuals who suffer from a reasoning failure,Footnote ⁹ it is the very paradigm of paternalist policies.

Le Grand and New (Reference Le Grand and New2014: 83f) distinguish four types of reasoning failure: limited technical ability, limited imagination or experience, limited objectivity, and limited will power, the last of which, as mentioned before, is arguably a failure to carry out the conclusions of one's reasoning than a failure of reasoning itself. Because reasoning failures can be objectively established, the boundaries of justified paternalist policies can be drawn clearly. Although the state is substituting its judgement for the judgement of individuals, it escapes Shiffrin's charge (Reference Shiffrin2000: 220) that ‘it directly expresses insufficient respect for the underlying valuable capacities, powers, and entitlements of the autonomous agent’, because it purports only to correct demonstrable reasoning failures. It is thus unlike policies that aim to discourage or prevent pursuit of individual goals that policymakers regard as harmful to their possessors. The reasoning failures that behavioural economists identify are not, of course, badges of honour; but these failures to live up to a rational ideal are pervasive human foibles rather than grounds for contempt and disrespect. Intervening to contravene the pursuit of some end, in contrast, is arrogant and unjustified. ‘There is no objective means of establishing an ends failure, and thus passing judgment on it should not form a part of any government's paternalistic policies’ (Le Grand and New Reference Le Grand and New2014: 103).

What distinguishes reasoning failures and makes them a defensible basis for state policy? Le Grand and New write:

Should one accept the necessary condition on paternalist interventions that this account suggests? What bearing does a failure to satisfy these conditions have on whether a policy counts as paternalistic? For example, suppose that it is not the case that Cindy should have known ‘the correct answer or procedure to find it’. Why should one then conclude that the substitution of public judgement for Cindy's (with the objective of benefitting her) is not paternalistic? It seems to me that in their desire to link their account of paternalism to a class of findings in behavioural economics, Le Grand and New have mistakenly attempted to build the findings of behavioural economics into the definition of paternalism.

5. FURTHER QUALMS

The discussion thus far of Le Grand and New's definition of paternalism suggests that it is mistaken to insist that a policy can be paternalistic only if it addresses a reasoning or judgement failure. Allowing that paternalistic actions could be directed toward many putative mistakes in addition to the cognitive failings identified by behavioural economics loosens the connection between the understanding of paternalism and the findings of behavioural economics. Although Le Grand and New's definition needs to be modified, their account of when paternalism is justified might still rest on the distinction between on the one hand interventions that correct reasoning failures or other mistakes agents make concerning the means of achieving their ends and on the other hand interventions that attempt to impose the values of policymakers on members of the population.

Part of what motivates Le Grand and New's definition is their appreciation of the range of subsidies and in-kind benefits that government supplies. These are, they argue, intuitively paternalistic:

I deny that subsidies to the arts and to museums are always paternalistic.Footnote ¹⁰ But they sometimes are, and the provision of many in-kind benefits such as food stamps seem to be at least in part paternalistic. Giving people food stamps rather than cash may reflect the judgement of the policymaker or of tax payers that the recipients of aid are subject to reasoning failures and will spend aid given in cash on less beneficial objects than food. So one might think that reflection on the paternalism of in-kind benefits supports Le Grand and New's definition of paternalism. But this is questionable. Unless one assumes that individuals are entirely self-interested, it may be that there is no cognitive failing in the choice of an action that does not promote an individual's well-being. To adapt an old example presented by T.M. Scanlon (Reference Scanlon1975), suppose that impoverished members of some sect think that it is vital to build a monument to their deity. If given aid in cash, they would use it for the monument. For that reason, policymakers might prefer to distribute aid in kind in the form of food stamps and housing vouchers. Unless one maintains that the objective of building the monument rests on a failure of judgement, Le Grand and New's definition of paternalism would mistakenly imply that the decision to provide aid in kind in this case is not paternalistic. I conclude that whether there is a reasoning failure is relevant only to the question of whether a paternalistic intervention is justified, not to whether a policy is paternalistic.

Notice that the provision of in-kind benefits, although obviously not a limitation on liberty, as compared with providing no benefits at all, does restrict liberty when it is compared to the alternative of providing individuals with cash.Footnote ¹¹ This suggests a possible variation on a Millian view of paternalism that requires as a necessary condition not a limitation on liberty but a greater limitation on liberty than some readily available eligible alternative policy.