Judging Nudging: Answering the Manipulation Objection

Introducing Nudging

The terms ‘nudging’ and ‘libertarian paternalism’ have been among us for more than a decade now (Sunstein and Thaler, 2003). Since the publication of Richard Thaler and Cass Sunstein’s (2008) bestselling book Nudge, they have spurred a normative and conceptual debate within academia and have instructed real-life policies in, for example, the United States and the United Kingdom (Department of Health, 2010). Recently, Sunstein (2015) has rehearsed some of the original issues, and this will probably rekindle the debate for years to come.

The subtitle of Thaler and Sunstein’s book is ‘Improving Decisions about Health, Wealth and Happiness’. Despite this wide focus, a prominent strand of the discussion focuses on how nudging can and should be made relevant to public health issues. Can and should policy-makers nudge people towards their own health? Given that paternalism in public health is the topic of a long and fierce debate, one can interpret nudging either as a suitable and soothing solution to the charges of paternalism or as an outright re-instalment and justification of paternalist zeal in public health care (Ménard, 2010; Skipper, 2012).

An important objection against nudging – which often remains insufficiently spelled out – is that it is problematic not because it is coercive (it is not) but because it is manipulative. In this article, we critically analyse this argument and assess its implications for the normative question whether nudging is a justified technique to improve public health. First, we critically discuss the claims that nudges are inherently manipulative and therefore undermine people’s autonomy. Second, even if a nudge is manipulative, this in itself does not necessarily undermine its legitimacy: nudges can be justified instances of manipulation. In order to clarify when this is the case, we refer to a number of aspects that are underdeveloped in the literature. In some cases, nudges can be understood as promoting rather than undermining people’s autonomy. Also, and especially in public health issues, nudging can be interpreted as democratic self-discipline, in which the ‘demos’ uses certain tactics to ‘bind itself’, that is, help itself make choices that are beneficial in terms of both health and autonomy. We will conclude that even if some people reject such strategies (nudging) and such goals (health), certain health-promoting nudges can be perfectly warranted in those circumstances where there are significant health benefits that cannot be achieved with other strategies. While nudging is and should be the topic of (democratic) controversy, we therefore reject the idea that nudging is always unjustified. Nudges belong to the legitimate toolbox of democratic governments.

Before we go into the normative question when nudging is or is not permissible and justified, we should clarify what nudging is exactly (section ‘Understanding Nudging’) and when it can be called manipulative (sections ‘Nudging and Manipulation: Definition’, ‘Nudging and Manipulation: Success, Goals’ and ‘Nudging and Manipulation: Techniques, Means, Choice Perversion’). Different elements in the definitions of manipulation will allow us to consider and answer objections that centre around nudges being manipulative. This will finally lead us to analyse democratic and other reasons relevant to assessing the legitimacy of nudges (sections ‘Nudging as Democratically Legitimate Manipulation’ and ‘Nudging as Legitimate Manipulation for Other Reasons’). For the sake of clarity and focus, we will consistently use Thaler and Sunstein’s (2008) famous cafeteria example (pp. 1–4) because it is paradigmatic of the entire project, it is clearly health related and it is remarkably instructive for the issues at stake. ¹

Understanding Nudging

According to Thaler and Sunstein (2008), a nudge is ‘any aspect of the choice architecture that alters people’s behavior in a predictable way without forbidding any options or significantly changing their economic incentives’ (p. 6). This definition has five elements: nudges are (1) interventions at the level of ‘choice architecture’ (2) in order to change people’s behaviour (3) in a way that is predictable, (4) without eliminating options and (5) without changing incentives.

In order to distinguish nudging from rational persuasion, ² we need to add a sixth element, namely that (6) nudging makes use of psychological mechanisms, cognitive biases and heuristics that cause people to make decisions that often go against their own interests (Bovens, 2009: 208; Grill, 2013: 38; Reiss, 2013: 291). Thaler and Sunstein start from behavioural research showing the predictable irrational responses caused by these mechanisms, biases and heuristics. Nudging strategies use these very same mechanisms in order to secure better outcomes in terms of ‘health, wealth and happiness’. Changing the ‘choice architecture’ – not the mechanisms! – is used to improve people’s decisions and better serve the welfare or interests of the people who are being nudged (the ‘nudgees’). In the field of public health, it is therefore hailed as ‘a new approach that empowers individuals to make healthy choices’ (Department of Health, 2010: 2).

In his more fine-grained analysis of Thaler and Sunstein’s definition of nudges, Yashar Saghai (2013) includes this sixth element: ‘A nudges B when A makes it more likely that B will φ, by triggering B’s automatic cognitive processes, while preserving B’s freedom of choice’ (p. 487). The key term here is ‘automatic’. Informing and persuading people do not work in this automatic way: we need to process (e.g. read), consider and evaluate the information in order to consciously change our behaviour. In contrast, nudges operate via the unconscious, uncontrolled, fast and easy processes – commonly called System 1 or the Automatic System – instead of the reflective, conscious, controlled, slow and more demanding processes of System 2 or the Reflective System (Sunstein, 2015: 26–27; Thaler and Sunstein, 2008: 19–22). It is exactly this aspect that makes nudges apt for ‘Humans’ like ourselves who are clearly no ‘Econs’ with ‘complete information, unlimited cognitive capacities, and no lack of will power’ (Sunstein and Thaler, 2003: 173).

Conrad Heilmann (2014: 77–83) uses nifty abbreviations to describe what goes on in nudging. In his view, (ideal) nudges are interventions aimed at people whose Reflective Systems endorse a prudent option (Rp) but whose Automatic Systems endorse another option (Aq). The nudge moves the nudgee’s Automatic System to endorse the prudent option (Ap) – while still allowing her Reflective System to reflect on and correct for this – ultimately leading to the nudgee voluntarily choosing the prudent option (so ultimately, AqRp → ApRp).

The cafeteria example is clear in this regard. Putting apples (the prudent and healthy option) and not Twinkies (the nonprudent and unhealthy option) at eye level increases the probability that people will pick an apple, thereby increasing overall apple consumption and contributing to public health. The process that leads people to do this is largely automatic: they tend to take whatever is at eye level or what requires the least effort. It is the same process that leads them to pick unhealthy items that also makes them pick the healthy alternatives. We simply need to arrange things differently. It is not that we have no other choice but to pick up the apple or that this choice is completely unconscious. It is just that not much thought goes into the decision anyway. However, and this is crucial, to the extent that people care about their health more than about the sheer satisfaction of a momentary sugar boost, they have good reason to prefer the apple over the Twinkie. So, their Reflective System endorses the apple as the better option.

Thaler and Sunstein (2008: 5) defend policies that are both ‘libertarian’ (‘people should be free to do what they like’) and ‘paternalistic’ (‘a policy is “paternalistic” if it tries to influence choices in a way that will make choosers better off, as judged by themselves’). Two remarks are in order to clarify the relation between nudges (according to the definition above) and libertarian paternalism. First, while nudges can be paternalistic, whenever they are aimed at the benefit of the nudgee, they are not necessarily so. If the Twinkie Company puts its products at eye level, it clearly nudges us but not to secure our interests. On our account, nudging refers to techniques used to influence people’s behaviour. ³ While Thaler and Sunstein obviously focus on nudging people towards their own interests, there is no reason why people cannot be nudged for the sake of other people’s interests. For instance, a nudge towards post-mortem organ donation (Thaler and Sunstein, 2008: 177–178) obviously does not benefit individuals’ own health but that of others. Second, some strategies that rely on people’s rational capacities do belong to ‘libertarian paternalism’ without counting as nudging. Think of labelling Twinkie wrappings with nutritional information. In what follows, however, we will focus exclusively on paternalistic nudges or ‘nudge paternalism’ (Anderson, 2010: 375).

Whenever people have worries about nudge paternalism, Thaler and Sunstein systematically (and in their view crucially) contend that it is liberty-preserving: options are neither removed nor made more costly. People are still free to choose and decide as they see fit. But we can question whether this is enough to convince the critics. An important argument against nudging is that it interferes not so much with people’s liberty but with their autonomy (Blumenthal-Barby, 2012: 352; Hausman and Welch, 2010: 128–130; Reiss, 2013: 294). What is at stake is not the amount of options that people have, but the level of control over deciding which option to pursue. Nudges, so the argument goes, are problematic because they are manipulative. Whenever we are nudged, our behaviour is steered by someone else in directions and in ways we did not necessarily approve of. This is why such manipulation is said to undermine our autonomy. To properly assess this claim, we first need a better understanding of what manipulation entails.

Nudging and Manipulation: Definition

In the literature, the term ‘manipulation’ is used both as a value-neutral, technical concept ⁴ and as a value-laden, normative term. While the first aims to understand manipulation as a specific type of influence (putting aside the question whether it is good or bad), the second assumes that there is something inherently wrong with manipulation. In an attempt to spell out what exactly is morally objectionable about it, Martin Wilkinson (2013b) asks, for instance, ‘if policy makers know your quirks and tailor their policies to fit your psychology, are they not manipulating you, treating you as a puppet?’ (p. 486). While there clearly is something wrong with being treated as a puppet, Wilkinson (2013a: 345–346) wants to clarify what exactly this is. In his view, manipulation is primarily wrong because of the implied violation of autonomy. Like puppets, nudgees are not in full control of their behaviour. Somebody else is in charge, causing their movements. They are not doing the choosing themselves, there is no real self-determination and, hence, their autonomy is undermined.

In his fuller characterisation of manipulation, Wilkinson (2013a) immediately includes its potentially wrong-making qualities:

Manipulation is intentionally and successfully influencing someone using methods that pervert choice […] Nudging clearly could satisfy the condition of having an intentional actor who successfully influences someone’s decisions. In principle, nudging could also pervert someone’s decision-making process and thereby infringe upon his or her autonomy. But if and when nudging does not use a perverting method, it is not manipulation. If and when nudging is not a form of intentional influence, it is not manipulation. If and when nudging does not succeed in altering behaviour, it is not manipulation (p. 347).

Let us follow up on this definition and consider, as we do in ordinary language as well, manipulation as prima facie wrong because it impairs the manipulee’s autonomy (Wilkinson, 2013a: 345). This is not to claim that all nudges are manipulative. After all, as Wilkinson’s quote makes clear, a nudge can fail to meet one of these necessary conditions. In turn, this is not to say that a non-manipulative nudge is perfectly justified: it can also be wrong for other reasons. In addition – and we will come back to this later on – even if a nudge meets all the criteria and is rightly considered manipulative, it is still only prima facie wrong. In other words, there can be instances of justified manipulation. Wilkinson’s definition of manipulation thus enables us (1) to clarify what exactly critics mean when they call nudges manipulative and (2) to assess under which conditions manipulation may be justified. Let us therefore look at the separate elements of manipulation that Wilkinson distinguishes.

Nudging and Manipulation: Success, Goals

First, in order to count as manipulation, nudges need to be successful and exert genuine influence on people’s choices. In the cafeteria case, one can distinguish three categories: (1) the apple-lovers – people with a settled preference for apples; (2) the undecided – people who have no settled preference for either apples or Twinkies (or any other food item); (3) the Twinkie-lovers – people with a settled preference for Twinkies (or any other food item, except apples). Now, suppose these people walk into a cafeteria in which apples are placed at eye level. When would this nudge be successful?

This is obviously the case when the nudge alters people’s decisions. This is when things go wrong, critics argue, because it means that the nudger is imposing her own goals on people who would have chosen differently without being nudged (White, 2013: 92). Such instances of value-substitution are widely criticised in the literature on manipulation (Baron, 2003: 41; Greenspan, 2003: 155–156) and on nudging. While proper or ideal nudges help people choose ‘what they would regard as the all-things-considered best choice, were they to deliberate about it’ (Heilmann, 2014: 75), manipulations have no such intention and aim to make the victim do whatever the manipulator wants her to.

In the cafeteria case, it seems to be the Twinkie-lovers (category 3) that fall victim to manipulation: the nudge influences their Automatic System to prefer an apple (Ap), while their Reflective System prefers a Twinkie (Rq). In contrast, the nudge helps the apple-lovers choose what they really want. It aligns their Automatic System (which otherwise would have been seduced to prefer a Twinkie) with their Reflective System (which prefers an apple; Rp). So if we stress the idea that manipulation involves a substitution of goals or values, then the same nudge can be manipulative for some people (the Twinkie-lovers) but not for other people (the apple-lovers).

Now, if people are nudged in a direction that they do not reflectively endorse, Bovens (2009) speaks of ‘exception’ nudges (p. 212). As Heilmann (2014) clarifies, the nudgee here ‘maintains that she is an “exception” or has exceptional preferences, such as really having no interest in a healthy diet’ (p. 84). Nudging her to go against her own overall judgement is wrong because it ‘is trying to make the decision-maker achieve a goal that she did not have in the first place’. Not only Bovens and Heilmann but also Hausman and Welch and Wilkinson consider these cases worrisome. Nudges are said to insufficiently respect the nudgee’s decision-making process, bypassing her Reflective System and leading her to ‘choose’ a healthy option without ‘really’ wanting it.

Our question is whether there really are people who are steered by the cafeteria’s choice architecture to go against their reflective preference. One way to find out is to confront them afterwards (ex post). ‘You were being nudged. You picked the apple. Is that really what you wanted?’ It seems unlikely that many people would object. At least, the deliberate choice architecture in supermarkets has not yielded mass protest on this account (‘Honey, I don’t know what just happened. I wanted to pick up some fruit, but I only have Twinkies in my bag. Darn these folk at Walmart!’). But, more importantly, this is a bad test for discovering manipulation. If people say they were not manipulated, this might just prove the strength and sneakiness of the manipulation (see Wilkinson, 2013a: 349).

So we need to look at people’s initial state of mind (ex ante). If we ask people before they enter the cafeteria what they would like to have for lunch, however, most will reply they do not know yet and that they will decide on the spot, depending on what is on offer (category 2). This lack of outspoken preferences is largely why people are susceptible to nudges. While ‘browsing’ for lunch, they will be more likely to choose the apples, but even if they do, we still do not know whether they did so because of the nudge. We may say that ‘what’ has made them do it is the nudge, but this might not answer the question of ‘why’ they did it. It is possible that they pick the apple, but not only because of the nudge. While the preference for an apple might be triggered by the nudge, it does not necessarily go against what people (really or generally) want. The crucial issue is not so much what exactly causes a person’s preferences and behaviour (or the difficulty in figuring this out, see Wilkinson, 2013a: 346–347) but to what extent the resulting preferences and behaviour are in line with that person’s ‘overall preference structure’ (Bovens, 2009: 213).

Let us therefore return to the Twinkie-lovers who have a clear preference that is supposedly subverted by the nudge (category 3). ⁵ Again, are there really such ‘exceptions’ who have no desire for apples but who buy the apples nonetheless? This seems a highly mysterious category. After all, Thaler and Sunstein (2008) stress that resisting nudges such as these should be ‘easy and cheap’ (p. 6). Yashar Saghai (2013) further develops this condition of easy resistibility: nudgees should be able to easily become aware of the nudge and inhibit the propensity it triggers. ⁶ This effectively solves the problem with ‘exception’ nudges: ‘If there is sufficient conflict between the influencer’s and the influencee’s aims, the attention-bringing capacities are likely to be activated, and the influencee may be in a position to resist the attempt to counter her behavior’ (Saghai, 2013: 490). The more a nudge goes against a person’s settled preferences, the more it makes her aware of the influence (ceteris paribus). If you are an apple-hater but find yourself picking up an apple in a cafeteria, you are likely to become aware that something weird and potentially autonomy-undermining is going on. This also explains why a lot of people still pick Twinkies: they are able to inhibit the propensity this nudge triggers.

So while the concern about value-substitution is a legitimate one, it does not pose an objection to nudges such as the one used in the cafeteria. After all, these not only leave the choice set intact but are also easily resistible (as they should if they are to be libertarian). Note, however, that such nudges can still be successful and meet this criterion for manipulation. After all, they can induce the undecided to opt for apples (category 2) and give a leg-up to weak-willed apple-lovers. ⁷

Nudging and Manipulation: Techniques, Means, Choice Perversion

Let us therefore move from the first element (success and influence) to the second element of manipulation as Wilkinson sees it, namely, as a ‘perversion of the decision-making process’. Most authors take this to be the crucial aspect that distinguishes manipulation from other kinds of influence, such as rational persuasion (Buss, 2005: 208), coercion or the use of economic incentives. According to Jennifer Blumenthal-Barby and Hadley Burroughs (2012), ‘manipulation occurs when one influences another by bypassing their capacity for reason, either by exploiting nonrational elements of psychological makeup or by influencing choices in a way that is not obvious to the subject’ (p. 5). On this definition, nudges are clearly manipulative since they do make use of non-rational psychological mechanisms in order to change people’s behaviour. According to Hausman and Welch (2010: 136), this is why nudges worryingly threaten people’s autonomy.

Note, however, the disjunction in the definition of Blumenthal-Barby and Burroughs. The second part – influencing people behind their back – is often considered to be the most problematic aspect of nudging because it is not apparent to the nudgee what is happening (Grüne-Yanoff, 2012: 636–637). Nevertheless, Sunstein (2015) offers a way out by means of the publicity and transparency condition, according to which nudges ‘should be visible, scrutinized and monitored’ (pp. 147–148). If manipulation means influencing someone ‘without making it apparent that this is occurring’ (Aggarwal et al., 2014: 33), then this would effectively rule out the possibility of governments manipulating their citizens behind their backs.

While the publicity and transparency conditions are clearly important, they are not enough to answer the manipulation objection. Being open about nudges does not make them less manipulative. Even if the influence of a nudge is obvious to the nudgee, it can still pervert her decision-making process by bypassing her capacity to reason. So the first part of the disjunctive is still applicable. Here, the idea is that nudges ‘try to “make the person do something that she has not herself (actively) chosen—for reasons she is not fully aware of”’ (Tengland, 2012: 144). This is why nudges are considered manipulative, violating people’s autonomy, even if they are fully transparent. The wrongness of nudging, and of manipulation in general, the argument goes, lies not so much in what it gets people to do (the discussion about goals) but in how it works (the techniques that pervert the decision-making process).

This is why nudging has been widely criticised for thwarting and therefore failing to respect people’s capacity for rationality and agency (Hausman and Welch, 2010: 128; Noggle, ⁸ 1996: 52; Tengland, 2012: 146). As Jeremy Waldron (2014) puts it, nudges are ‘an affront to human dignity: I mean dignity in the sense of self-respect, an individual’s awareness of her own worth as a chooser […] My capacities for thought and for figuring things out are not really being taken seriously’. However transparent it might be, and even if it helps me in reaching my true goals (like a healthier life), such manipulation is considered wrong because it perverts my decision-making capacities. Because it treats me as a child, it is arrogant and insulting and it constitutes ‘a failure to view others as rational beings’ (Baron, 2003: 50). Waldron does not want to live in a ‘nudge-world’ full of manipulating marketers and policy-makers but wants the government to respect us and rather let us err autonomously than rally us up like sheep into the pen of health or happiness.

Is it really a sign of disrespect to use these non-rational, reason-bypassing techniques of influencing people? Not necessarily. Suppose we are rationally convinced that it would be better for us to eat apples rather than Twinkies (category 1). Here, the cafeteria’s choice architecture simply uses non-rational influences in order to help us do what we consider best. Rationality – our capacity to determine our goals and successfully pursue them – sometimes requires the use of mechanisms that do not operate via rationality (Reiss, 2013: 296). To say that nudging is manipulative because it bypasses our rational capacities ignores that such bypassing can be required exactly to enhance our rationality. In these cases, nudges do not bypass but scaffold our rationality in situations where it is under pressure.

Now, the critic might insist. Even if nudges cause us to choose the things we would choose upon reflection, they make us ‘choose’ for the wrong reasons, or even worse, for no good reason at all. The psychological mechanisms that nudgers tap into do not count as reasons, and that is what makes the resulting behaviour irrational (Bovens, 2009: 209). What causes nudgees to pick the apple is not their considered judgement that apples are better than Twinkies but the combination of an irrational psychological mechanism and some clever choice architecture. And that is disrespectful. Instead, people would like to achieve their goals on their own strength. However, and we go into this argument more fully in the next section, (the influence of) some kind of choice architecture is often inevitable. To claim that we would be completely free, autonomous and self-reliant without nudges is therefore unconvincing. In the cafeteria, a nudge towards apples simply swaps the outcome of one unreflective process with another. If I put whatever is at eye level on my tray (apples or Twinkies), I retain the same level of reflection and voluntariness in different choice architectures. In these cases, nudging does not make me any less rational. It does not ‘bypass’ my rational decision-making capacities because these were simply never really in charge. ⁹

The concern about ‘bypassing’ only comes up when we compare nudging to other strategies (e.g. giving nutritional information). Nudging does not make as much use of people’s deliberative processes as informing and persuading would do. But, of course, the whole idea behind nudging is that such alternatives often do not work. The belief that people will resist temptation as soon as they have all the information is a mirage. The idea of ‘Ulysses unbound’ – an autonomous and rational person who is free from psychological mechanisms and biases – does not take the call of the Sirens seriously and is, as we will argue below, extremely unrealistic.

Let us summarise our analysis so far. When people enter a cafeteria, they can (1) have a settled preference for apples, (2) lack a settled preference or (3) have one for Twinkies (or any other food item). In our view, a nudge towards apples poses no problem for category (1), because these people identify with the preference that the nudge brings about, which is how nudges ideally work (Heilmann, 2014: 79–80). ¹⁰ With respect to category (3), there is a concern that people are manipulated towards an option they really do not like. Heilmann (2014) rightly stresses that this ‘gets it wrong’ because the nudge tries to make these people act against their reflective preference (p. 84). ¹¹ We, however, believe such attempts will hardly ever be successful, at least as long as nudges are easily resistible. People who claim to be in this situation will have a hard time arguing why they do not succeed in picking a Twinkie.

This brings us to category (2). It seems that nudges are rather successful in influencing these people. Because they have no strong incentive to resist the (easily resistible) nudge, their behaviour is likely to be influenced by the nudge. While we believe that the cafeteria nudge can be said to manipulate these people into buying apples – they are intentionally and successfully influenced through non-rational methods – we believe the critics overstate their case. Arguing that their decision-making processes are ‘perverted’ or ‘distorted’ and that their rational capacities are ‘bypassed’ wrongly assumes that these are in place before or without installing nudges. ¹²

In sum, there are two possible answers to the question whether nudges manipulate the undecided people in category (2). No, if one defines manipulation as ‘choice perversion’: their initial decision-making process was simply not as rational or autonomous as critics assume (and hence could not be perverted). Yes, if one defines manipulation simply as ‘non-rational influence’ (as Blumenthal-Barby and Burroughs do). In both cases, however, the crucial question is how nudging compares to its alternatives. Instead of intentionally designing choice architecture towards people’s health, one can steer them in other directions or install a more ‘random’ choice architecture. What is often not possible, however, is to return to some kind of pristine autonomy or ideal rationality.

So while the undecided would have acted differently in the absence of nudging, this is not because they would have been completely free from influence. Both the choice architecture (external) and the psychological mechanisms and biases (internal) will continue to have an effect on their decisions, even if there is no intentional nudger making use of them. Some choice architecture (some default option, some framing, some options at eye level) is simply inevitable. Even without intentional nudges, people remain prey to these influences: they are like puppets whose strings are pulled one way or the other. If critics claim that nudges undermine people’s autonomy or distort their rational decision-making processes, they mistakenly assume that people’s autonomy and rationality was intact before nudges were applied. The insight that this is not the case is exactly the starting point in the whole case for nudging. To believe in a pristine and untainted type of autonomy is to deny the evidence that reveals this to be an illusion. ‘Human beings are simply without the sort of decision-making autonomy that advocates of paternalism are accused of violating’ (Skipper, 2012: 182).

Take laziness, akrasia and cognitive biases. To the extent that people like us – Humans, not Econs – tend to suffer from these proclivities, paternalist strategies can be said to do better in respecting us and our deeply held values than purely rational strategies of informing and persuading:

Pretending that we are competent in ways we are not is no foundation for respect. And because coercive paternalism not only recognizes our cognitive shortcomings, but moves us to help us where those abilities are shaky, it actually values our choices about our ultimate goals more than does the sort of paternalism that simply gives us a hint in the right direction but then keeps out of the way as we make choices that entirely undercut our aims and values (Conly, 2013: 242–243).

Sarah Conly talks about coercive paternalism, but her point carries over to nudge paternalism. Insisting on informing and persuading gets things wrong since nudges aim to help people to do what they are already convinced of (Marteau et al., 2011: 263; Reiss, 2013: 296). It is hard to see how this would be degrading, insulting or disrespectful. To insist that governments should treat us as rational beings is somewhat absurd in light of the evidence that reveals this to be an unrealistic ideal that gives rise to ineffective policy measures. Also, it is an illusion to believe that a ‘nudge-free environment’ will do better in securing people’s autonomy. Take a random choice situation. To argue that this leaves our ability to decide for ourselves intact, while intentional choice architecture does not, would assume a weird type of Providence. As if the choice architecture of the Universe – what would that even be? – would somehow miraculously track people’s reflective judgements.

Of course, one can prefer random or unintentional choice architecture because it does not steer people in a specific direction, as is the case in a government-imposed Grand Scheme of Health Nudges. The difference between intentional and unintentional nudging, critics argue, lies in the extent to which there is an agenda that is being imposed (Greenspan, 2003: 160) by an intentional agent (Hausman and Welch, 2010: 133; Wilkinson, 2013a: 343). However, not only is it inappropriate to speak of ‘imposing’ an agenda if nudges are easily resistible, it is completely unclear how being jostled about by a variety of unintentional or happy-go-lucky influences would be any better in terms of autonomy. While critics want to decrease government’s control over choice architecture, this reduction in the intentionality behind it all does not necessarily increase people’s control over their own choices.

Nudging as Democratically Legitimate Manipulation

This last aspect brings us to yet another criticism. Instead of arguing about goals (‘why’) or techniques (‘what’ and ‘how’), the claim here is that it is morally relevant ‘who’ is doing the nudging. This relates to the issue of intentionality. Surely, nudges are implemented by intentional agents who aim at changing people’s behaviour. And this is exactly what some believe to be objectionable about manipulation: even if it does not succeed, it is the attempt and the manipulative attitude behind it that is wrong. While this can be expected from profit-maximising companies, the argument goes, it would be wrong for governments to treat us in that way.

Mark White (2013) gives two reasons why it is problematic for governments to nudge us. ¹³ First, there is the informational and epistemic issue how government agencies and bureaucrats can accurately know our reflective judgements. This is why the value-substitution we analysed before seems a genuine concern when others try to figure out what is good for us. Second, while we expect such manipulation from private companies (who clearly steer us in directions that serve their own interests), we do not expect it from government and this makes us extremely vulnerable to it.

In our view, both claims are questionable. The second is largely empirical. Is it true that people are indeed more ‘vigilant’ when they enter a (private) supermarket than a (state-funded university) cafeteria? Given the transparency and publicity that governments need to show (Hausman and Welch, 2010: 135), awareness can surely be raised there as well. We are not convinced by the first claim either. Value-substitution is surely more of a concern when companies nudge us (they care about their own interests) than when governments do the nudging (at least they try to find out what is in our interest). In addition, why would it be impossible for governments (or anyone else) to know what we really want? While we have already dealt with this point about ‘exception’ nudges, the question is a legitimate one: how can governments reliably come to know what people themselves judge to be their real preferences?

Thaler and Sunstein’s easy answer is to focus on obvious cases. Of course, we can safely assume that the majority of people want to be ‘healthy, wealthy and happy’. Who really wants to die in a car crash or from obesity-related causes? Who indeed? But apart from the ‘obvious’, we want to stress the role of vital democratic processes – a role that is often dismissed.

According to Jeremy Waldron (2014), for example, nudging policies involve two radically separated parties. First, there are ordinary, biased, myopic and weak-willed people (the nudgees). Second, there are people ‘endowed with a happy combination of power and expertise’ who know how ordinary people think and can use clever choice architecture to influence their decisions (the nudgers). Waldron is thus concerned about government officials and health experts (‘them’) steering ordinary people (‘us’) towards specific goals. Why are they – the manipulators – in a better position than us – the manipulees – to know what we really want? Why should we trust ‘them’ with this kind of judgement and power?

In order to transcend Waldron’s ‘us’ versus ‘them’ dichotomy, we need to analyse the democratic idea that people can consent to government agencies and policies that are both responsive and accountable. Policy goals, such as health, need to be defined or at least endorsed by the people themselves. The issue at stake here revolves around the notion of consent. ¹⁴ If government is right in assuming that citizens really want to be healthy, there is a clear democratic justification for nudging. Even if nudges are manipulative, consent counts as a reason for justifying them: ‘manipulation can be justified only when the manipulee would endorse the process, or the means to the end attained, along with the end’ (Greenspan, 2003: 156). What exactly is the problem with government nudging us towards our health, if we are informed about and agree with its goal (we want to become healthy) and its means (we want to be nudged to become healthy)? Such a government is not so much disrespecting its citizens as taking up its responsibility to help citizens act upon their own values (Carter and Hall, 2012: 11).

This reveals Waldron’s distinction between ‘us’ and ‘them’ to be an artificial and misleading one. In democratic societies like ours, ‘we’ (the citizens) are part of ‘them’ (the government). What ‘they’ do, ‘they’ do in ‘our’ name and because ‘we’ enable ‘them’ to do it. Nudging is not about being manipulated by experts who know better, but about us collectively invoking government’s help when we know we are likely to make bad decisions. Instead of ‘they’, it is actually ‘we’ (our Reflective Systems) who are nudging ‘ourselves’ (our Automatic Systems) (Conly, 2013: 242). Proponents of nudging do not assume that some people are less rational than others but acknowledge that we are all, from time to time, less rational than we want to be.

This is exactly what we enable governments to do all the time. Think of food safety laws and the public agencies that enforce them. Instead of arguing that people should be allowed to choose between all possible products on the basis of adequate information, these laws and institutions not only save us time and effort but also prevent things from going horribly wrong. Here too, we collectively invoke and enable government to do at least part of the job for us.

But our problem returns. Since nudges are applied at the institutional level with policy-makers and cafeteria managers nudging all citizens and consumers, there will always be people who oppose being nudged because they do not endorse the goal (health) or the means involved (nudging). The concern here is with governments imposing one-size-fits-all policies, thereby infringing on the freedom of at least some. Even if a majority is fine with being nudged, some people will not consent (Wilkinson, 2013a: 353–354). Recognising the pluralist character of our society, however, should not render it impossible for democracies to formulate clear policy goals and means. If citizens have different beliefs and preferences, any policy will apply to at least some people who disagree with it. Yet a policy can be democratically legitimate if one can reasonably argue that it achieves what the ‘demos’ wants. Different conceptions of democracy have different ideas about what this entails. Aggregative democrats argue that a policy is legitimate if it is the expressed preference of a majority of voters. Deliberative democrats argue that a policy is legitimate if it is backed up by what – after proper public deliberation – turns out to be the best argument. One can, for example, use deliberative polls to find out what people really prefer (Schiavone et al., 2014: 108–111). In both conceptions, policies can be legitimate, even if they lack universal consent. The fact that not everybody agrees to nudging strategies in public health is not enough to call them undemocratic. ¹⁵ In fact, we rely on democratic procedures to deal precisely with such disagreement.

Finally, one should ask once more what the alternatives for nudging are. If choice architecture is inevitable, the alternative for government nudging us is somebody else nudging us (Reiss, 2013: 297). And granting the ‘power of nudge’ to companies and their marketing staff is even less likely to get us where we really want. Likewise random framings and luck-of-the-draw defaults. To acknowledge that it is difficult to determine what people really want does not imply that governments should not attempt to do so. To assess which nudges are ever justified, we believe the question about the direction in which they try to steer us is much more important than the question about who is doing the steering. While the nudging techniques that governments and companies use are the same, at least the government aims to steer us in a direction that we – as a collective – would endorse (or else it would be undemocratic).

Nudging as Legitimate Manipulation for Other Reasons

So in contrast with Martin Wilkinson (2013a: 354), we grant that the kind of nudges that Thaler and Sunstein propose – such as the cafeteria one – can be manipulative, at least for some people. In contrast with Thaler and Sunstein, we believe that referring to the liberty-preserving character of nudges, together with the publicity and transparency condition, does not take away all normative worries. Whenever nudges manipulate, this raises a genuine concern and counts as a prima facie reason against them. But even when nudges are prima facie wrong because of the involved manipulation, this worry can be outweighed by other reasons, goals, goods and values that justify manipulative nudges in specific situations (Blumenthal-Barby and Burroughs, 2012: 5). Sometimes the stakes are so high and the alternatives so inadequate that nudging could be justified on objective grounds. So, in addition to recognising nudging as one of the many (and among the least intrusive) public health policies that policy-makers can legitimately employ, one can add that, in urgent and important cases like the obesity pandemic, we should err on the side of caution, that is, in favour of health-promoting measures (Conly, 2014). The reason for this is that health is a special and crucial policy goal. Not only do most people realise the value of health and support measures to promote it, it is also a precondition for autonomy itself.

Health-promoting nudges can promote people’s autonomy in two ways. First, they can help people achieve health, which in itself is a prerequisite for autonomy. Health is then not so much the object of autonomous choice but its precondition (see Daniels, 1981: 154). Health is essential in safeguarding and securing people’s capacity to decide for themselves (even for those who do not want to be healthy). This makes health-promoting nudges stand out from nudges directed at wealth and happiness.

This argument employs a broad conception of autonomy (having the general capacity to live a life of one’s own). But also on a narrower conception (being in control right now, having the capacity to fulfill one’s present desires) nudges can be said to promote and increase rather than violate or decrease people’s autonomy. Paradoxically, nudges (which rely on the Automatic System) can help people in making the choices they (their Reflective System) would make if they were rational, informed and free of predictable errors (which they often are not at the moment of choice). So if we value our rationality and autonomy and discover that these are threatened by psychological and motivational biases, we can use this knowledge and these very same biases to help us (Skipper, 2012: 186–187). Of course, while nudges can enhance some people’s autonomy by helping them secure what their Reflective System wants, this is not true for Twinkie-lovers, whose reflective preference may be to ‘live fast and die young’. As we have argued above, however, they are not the target audience of easily resistible nudges.

So nudges compensate for our lack of autonomy by making use of the very mechanisms that limit it. How can this work? Nudges typically rely on one psychological mechanism to counteract another. In the cafeteria, salience is used to go against our craving for sugar. Or take Michelle Obama’s Let’s Move! programme or other attempts to tackle obesity by stimulating people to publicly commit to healthier lifestyles. Here, conformism and the fear of social sanctions are used to help people overcome their myopia and akrasia. The idea is thus to ‘counter-manipulate individuals’ (Tengland, 2012: 145) and to pull one string to counteract the impact of another.

Conclusion

In sum, we believe that health-promoting nudges belong to the legitimate toolbox of democratic governments. The worry that they manipulate some people is valid but outweighed by their ability to serve valuable goals such as public health and people’s autonomy. In the case of health, most people even endorse more ‘extreme’ measures than nudging, such as restricting people’s liberty (e.g. to take extremely dangerous drugs) and taxing unhealthy substances (e.g. nicotine). If we really value public health and if (easily resistible) nudges prove an effective means to achieve this end, they might well be justified all things considered. We need to assess their legitimacy on a case-by-case basis. ¹⁶

In order to do this, let us list some of the key elements that should be taken into account when investigating a specific (health-promoting) nudge: its effectiveness and its (public health) benefits, the distribution of its costs and benefits across the population, the (potential loss of) freedom and autonomy it entails, its manipulativeness, the information the nudging institution has about the preferences of potential nudgees and the degree to which the latter (would) consent. In each of these aspects, nudging strategies should systematically be compared with their alternatives (persuading, informing, changing incentives, coercing, doing nothing, etc.).

In addition, it should be clear what we do not defend. First, we do not believe that nudging is a panacea for all ethical and political issues concerning public health (Ménard, 2010). Second, we do not believe that employing nudges towards health implies that one has to stop informing and persuading people. We simply stress that focusing exclusively on the latter is likely to prove ineffective because it is based on an unrealistic view of human psychology.

People have legitimate concerns about (1) the direction in which they are nudged, (2) the very fact that they are being (successfully) nudged and (3) the fact that it is their government that is nudging them. We have formulated these concerns in terms of manipulation and discussed various ways of deflating them. This has led us to conclude that, under specific conditions, governments can legitimately employ nudges as effective strategies to achieve the valid policy goal that is public health.