The Power to Nudge

Abstract

Nudging policies rely on behavioral science to improve people's decisions through small changes in the environments within which people make choices. This article first seeks to rebut a prominent objection to this approach: furnishing governments with the power to nudge leads to relations of alien control, that is, relations in which some people can impose their will on others—a concern which resonates with republican, Kantian, and Rousseauvian theories of freedom and relational theories of autonomy. I respond that alien control can be avoided, if nudging is suitably transparent and democratically controlled. Moreover, such transparency and democratic control are institutionally feasible. Building on this response, I then provide a novel and surprising argument for more nudging: democratically controlled public policy nudging can often contain the power of private companies to nudge in uncontrolled and opaque ways. Therefore, reducing alien control often requires more rather than less nudging in public policy.

INTRODUCTION

Nudging has recently proven popular with governments around the world. Its idea is simple yet powerful. Relying on insights from behavioral science, nudging seeks to improve people's decisions by changing the way options are presented to them. It does not change the options themselves nor the costs and benefits associated with these options. As we now learn more about how people actually make decisions, proponents of nudging suggest using this knowledge to tweak choice environments in a way that nudges agents into choosing options that are good for them. For example, agents often go along with an option, if it is the default. We can harness such status quo bias, by making desirable options the default. To increase the rate of organ donors, for example, a country might adopt an opt-out system such that being an organ donor is the default. Or to increase people's retirement saving, we can make it the default that employees are enrolled in a retirement plan unless they opt out.

Nudging has been met with both enthusiasm and sustained critical attention from all sides of the political spectrum. In this article, I focus on one important and common objection to nudging which I call the objection from alien control. The worry is that furnishing the government with far-reaching powers to nudge its citizens leads to problematic power relations, relations of alien control. An agent A holds alien control over someone else B, when A has the power to influence B’s life and such power is not controlled by B (B is an individual, but A can include individual and collective agents). If the government holds alien control over its citizens, it can impose its will on them. Critics now argue that nudges are more inconspicuous and insidious than traditional interventions, such as taxes and mandates, which makes it hard to place them under suitable individual and democratic control. And because they are unsuitable to be controlled, far-reaching governmental power to nudge increases the extent to which the government holds uncontrolled power—or alien control—over citizens (Dunt 2014; Goodwin 2012; Grüne-Yanoff 2012; Hausman and Welch 2010). I will spell out the objection from alien control in republican terms. But the underlying normative concern will be shown to strike a chord with a broad range of normative views, including Kantian and Rousseauvian theories of freedom as well as relational theories of autonomy.

The objection from alien control has not been satisfactorily addressed by proponents of nudging so far. When challenged on whether nudging is compatible with freedom and autonomy, proponents typically argue, first, that nudging preserves freedom of choice and, second, that nudging is compatible with volitional autonomy (Mills 2015; Saghai 2013; Sunstein 2014, 2015b). But these defenses alone fail to dispel worries about power relations that allow government officials to impose their will on citizens. Moreover, in their responses, proponents of nudging typically rely on traditional liberal views of freedom and volitional views of autonomy. A detailed engagement with other prominent traditions, such as republican, Kantian, and Rousseauvian views of freedom and relational accounts of autonomy, has not been provided yet. This article fills these gaps in the literature.

The first contribution of this article is to show that the objection from alien control fails. I argue that if systematic public policy nudging is done right, it does not lead to relations of alien control. The main idea is that it is feasible to make nudging suitably transparent and democratically controlled. In a next step, I build on this response to provide a novel and somewhat surprising argument in favor of more systematic public policy nudging. When turning my attention to nudging techniques in the private sector, I argue that systematic public policy nudging might contain the power of private actors, particularly private companies, to use nudges opaquely and without democratic control. The upshot is that a normative worry about alien control speaks for, rather than against, systematic public policy nudging.

These arguments provided here have significant implications. First, a common reason to reject nudging does not apply, if nudging is suitably transparent and under democratic control. What is more, rather than less nudging, those who worry about power relations should typically welcome more systematic democratically controlled nudging. Second, I lay out how to achieve such transparency and democratic control, which gives us some guidance for how nudging should be institutionally implemented.

I proceed as follows. In the next section, I introduce some of the main ideas behind nudging. In the section thereafter, I outline the objection from alien control. I then argue that the objection from alien control fails. The power to nudge is not a form of alien control, if it is both suitably transparent and under suitable democratic control. In the section entitled Transparency, I discuss how nudging should and can be suitably transparent. In the section thereafter, I then discuss how nudging should and can be suitably democratically controlled. In the penultimate section, I provide the novel positive argument for more systematic public policy nudging: suitably transparent and democratically controlled nudging can contain the uncontrolled power of private companies to nudge and thereby help democratize control over choice environments. I briefly summarize my arguments in the final section.

NUDGING

Relying on heuristics, mental shortcuts, and intuitive decision making, real-life agents often systematically diverge from the traditional, abstract ideal of rational choice. Such divergence is a necessity and often works remarkably well. However, as proponents of nudging argue, real-life decision making is also rife with cognitive biases resulting in decisions that leave people worse off by their own lights. Here are some examples: in their decision making, people often rely on anchoring; often rely on the availability and representativeness heuristics; are often too optimistic and overconfident; display loss aversion; give undue importance to the status quo; are subject to framing; often make their decisions conditional on what others are doing rather than evaluating options on their own merit; are often short-sighted and suffer from weak will; and real-life agents often and systematically go wrong when dealing with probabilities, conditional probabilities in particular (Kahneman 2011; Kahneman and Tversky 2000; Thaler and Sunstein 2008, 23–40).

Proponents of nudging, most famously Thaler and Sunstein, suggest a range of public policies that cleverly use people's biases to improve their decisions, namely by changing the environment within which people make decisions (their choice architecture).¹ For example:

Retirement: when confronted with many different retirement plan options, employees often end up choosing no plan at all. Making enrolment in a plan the default can increase coverage rates drastically, as people often stick with the default.

Elective Surgery: by presenting risk information differently, we can change people's decisions. For example, saying that “of one hundred patients who undergo this surgery, ninety are alive after five years,” rather than “of one hundred patients who undergo this surgery, ten are dead after five years,” can make it more likely that a patient chooses to have the surgery.

Tax Compliance: we can influence people's behavior by giving them information about how others are behaving. For example, including information about how many others have already paid their taxes, or highlighting the loss of public service, can make it more likely that people pay their taxes.²

Thaler and Sunstein sometimes call their program libertarian paternalism: it is libertarian, because it does not affect people's freedom of choice in any significant way. It is often paternalist, because it is aimed at increasing people's welfare by trying to get individuals to choose a certain way (even if they may have been disposed to choose differently). Thaler and Sunstein hold that the welfare to be promoted is subjective and thus constituted by the interests and preferences of the person being nudged. It is not a perfectionist or otherwise externally imposed idea of welfare (Thaler and Sunstein 2008).

Note that when defending systematic public policy nudging, I will not use the label libertarian paternalism. First, many nudging policies, such as tax compliance, are not paternalist, because their aim is to effect behavior with benefits for others or for society overall. My arguments apply to both paternalist and nonpaternalist nudges—although I mostly focus on paternalist nudges, as they typically meet with stronger ethical objections. Second, the label libertarian has left some people with the false impression that a defense of a nudge program comes with a commitment to libertarianism. My defense of nudging in no way commits me to a libertarian disapproval of other, traditional roles of the state, such as taxation, mandates, welfare state provisions, and so on. Nor, of course, does it commit me to the opposite view that we should be as interventionist as possible. Finally and relatedly, my aim is to defend a system of public policy nudging. I do not intend to defend the (somewhat hyperbolic) claim that nudging is a third way for government, that is, a new political framework that transcends the traditional divide between statists and libertarians.³

THE OBJECTION FROM ALIEN CONTROL

Over recent years, many objections have been raised to the nudge program.⁴ In this article, I respond to one important set of objections that, I think, proponents have failed to adequately address so far. (While this response might go some way towards also alleviating some of the other worries, my aim is of course not to answer all possible and actual objections to nudging.)

Many might intuitively object to systematic nudging, because they take issue with the types of social relations it might lead to. To illustrate this point, consider a popular argument used by proponents of nudging, what I call the inevitability argument: objections to nudging fail, in a sense necessarily, because choice architecture is inevitable (Sunstein 2015a, 420–2). Choices always happen within choice environments. If we do not nudge persons towards good decisions, alternative choice environments might nudge them towards bad ones. Because nudges are inevitable, let us go for the good ones.

However, critics respond that the inevitability argument misses an important distinction:

There remains an important difference between choices that are intentionally shaped and choices that are not. Even when unshaped choices would have been just as strongly influenced by deliberative flaws, calculated shaping of choices still imposes the will of one agent on another. (Hausman and Welch 2010, 133)⁵

The difference between intentional and unintentional nudges is important. In drawing said distinction, I think what worries most commentators is the social relation between those that nudge (henceforth the nudgers) and those that are nudged (henceforth the nudgees). A range of philosophers and political theorists—including Hausman and Welch in the above quotation—take issue with social relations and institutions that allow one set of agents to impose their will on others. And it is only of intentional nudges, rather than nonintentional ones, that we can say that someone might impose his or her will on another.

The worry now is that granting the state the power to systematically nudge its citizens indeed puts some people in a position to impose their will on others. I call this the objection from alien control: equipping nudgers with the power to nudge makes nudgees subject to alien control. An agent A holds alien control over person B, if A has the power to influence B’s life in a way that B does not control (where A can be an individual or a collective agent). Nudgees are subject to alien control, because they are subject to power they do not control. Nudgees cannot properly control the power to nudge, or so a common argument runs, because nudges are hard to detect. Hausman and Welch, for example, argue that

[T]here may be something more insidious about shaping choices than about open constraint. For example, suppose. . .that subliminal messages were highly effective in influencing behavior. [. . .] Influencing behavior in this way may be a greater threat to liberty, broadly conceived, than punishing drivers who do not wear seat belts, because it threatens people's control over their own evaluations and deliberation and is so open to abuse. The unhappily coerced driver wearing her seat belt has chosen to do so, albeit from a limited choice set, unlike the hypothetical case of a person who brushes his teeth under the influence of a subliminal message. . . .[T]o the extent that it lessens the control agents have over their own evaluations, shaping people's choices for their own benefit seems to us to be alarmingly intrusive. (Hausman and Welch 2010, 131)

Hausman and Welch do not explicitly write about alien control. But inasmuch as we are concerned with the moral quality of social relations, their worry about nudging applies here too: the insidiousness of nudging means that nudge interventions seem largely beyond an individual's control and allow others to impose their will on one's deliberations and decisions.

Invoking Pettit's account of freedom, Grüne-Yanoff spells out this worry in explicitly republican terms:

A government employing such [nudging] policies therefore increases its arbitrary power over its citizens. An increase in arbitrary power, implying an increase in citizens’ “defenceless susceptibility to interference”. . .is a sufficient condition for a decrease in these citizens’ liberty. . .Thus, the mere mandate to enact these policies decreases liberty. . .(Grüne-Yanoff 2012, 638)⁶

To assess Grüne-Yanoff's claims, let us spell out Pettit's idea of republican freedom. To be free in the republican sense, one needs to be free from alien control or free from domination or arbitrary power—all of which mean the same thing.⁷ According to Pettit, “someone, A, will be dominated in a certain choice by another agent or agency, B, to the extent that B has a power of interfering in the choice that is not itself controlled by A.” (Pettit 2012, 50) To unpack this notion, consider two of its central components.

First, on a republican conception, the mere uncontrolled power to interfere suffices to render someone unfree, even if such power is never exercised. Compare this with liberal, negative views of freedom, according to which freedom is the absence of actual interpersonal constraints (Berlin 1969; Carter 1999; Kramer 2003; Kristjánsson 1996; Miller 1983; Schmidt 2016; Shnayderman 2012, 2013; Steiner 1974). On the negative view, I am not unfree if someone has the uncontrolled power to interfere with my choice but does not exercise his or her power. Uncontrolled but unexercised power does not render me unfree. This is different for republicans for whom uncontrolled power, even if unexercised, renders me unfree.

Second, and conversely, not every interference is a source of unfreedom on the republican view. If someone interferes with me but I hold suitable control over his or her power to interfere, then such interference does not render me unfree. For example, imagine I want to reduce my sugar consumption and give my partner the key to my candy cabinet (I retain the power to get the key back on 24-hour notice) (Pettit 2012, 75, 171). My partner interferes with my freedom of choice by locking the cabinet. But I am not made unfree, because I still control his or her power to interfere with my food choices. Accordingly, he or she does not impose his or her will on me. Contrast this with uncontrolled power of interference:

. . .when I say that that power of interfering is not controlled by A, I mean that it is not exercised on terms imposed by A: it is not exercised in a direction or according to a pattern that A has the influence to determine. In that respect, it is unlike the interference that someone invites when they hire an agent, for example, to make certain decisions for them. (Pettit 2012, 50)

Note two things about control:

First, when I control another person's power to interfere, my control can be indirect. For example, for the state's traffic regulations to be suitably controlled, I do not need to personally oversee or be involved with every traffic law decision and enforcement. But there should be other individuals or collective agents that, acting on my behalf, exercise control over traffic legislation. To suitably control state intervention requires democratic control (which I discuss below in the section Democratic Control).

Second, control need not always be active. I can also exercise virtual or reserve control. Consider an analogy. When riding a horse, I exercise active control over where to ride by shortening the reins on one side. But if the horse already runs in the direction I desire, I can exercise virtual control and not do anything. I still exercise control. For if the horse went into a direction I do not desire, I would act to get the horse back in my desired direction by shortening the reins. Finally, I can also exercise reserve control. If I have no desire to go in a particular direction, I can just let the horse go wherever. I am still in control. For if I start desiring a particular direction, I can exercise active or virtual control. Going back to human relations, to suitably control another person's power to interfere, one need not always exercise active control. Often, virtual or reserve control will do (Pettit 2014, 1–2, 211).

Applied to nudging, we can say that the power to nudge is a form of alien control, if the nudgers hold power over nudgees that is not suitably controlled by the nudgees. Grüne-Yanoff and others think that the power to nudge is a form of alien control. The main argument is that nudging escapes suitable control, because nudges are insufficiently transparent. Let us call this the transparency argument:

1. A. By giving the nudgers the power to nudge others (the nudgees), we grant nudgers the power to influence nudgees’ lives.

2. B. For power to influence other people's lives not to be a source of alien control, it needs to be suitably controlled—either directly or indirectly—by those subject to this power.

3. C. For power to influence other people's lives to be suitably controlled, such power and its exercise need to be suitably transparent.

4. D. By its very nature, the power to nudge and the exercise thereof cannot be suitably transparent.⁸

5. E. Therefore, the power to nudge and the exercise thereof are not suitably controlled by those being nudged.

6. F. Therefore, the power to nudge is a source of alien control.

In this article, I will rely on this recent republican framework to spell out and, ultimately, dismiss the objection from alien control (and the transparency argument in its favor). Note, however, that the objection from alien control has traction beyond contemporary republicans and should strike a chord with a broad range of authors and theoretical frameworks. For example, consider views of freedom sometimes called “freedom as independence” or as belonging to the Franco-German republican tradition (as opposed to the Italian-Atlantic tradition Pettit identifies with (Pettit 2013)). For Rousseau, being free requires not being subjected to the will of others. As Neuhouser writes on Rousseau, “freedom. . .always refers to a relation between one will and another: To be unfree is to obey a foreign will, and freedom is always being free of the will of another” (Neuhouser 1993, 381). Similarly, Kant—and Kantians—also hold that freedom is about not being subject to another person's will (Hodgson 2010; Kant 1996; Ripstein 2010; Rostbøll 2016).

Finally, consider authors who defend relational views of personal autonomy. Relational theories of autonomy move away from the traditional, individualized picture of autonomy and hold that, necessarily, being autonomous implies not being in certain types of social relations with other people. As Garnett argues, “self-rule” might not only—or not even primarily—be about ruling oneself. Self-rule is also about not being ruled by others (Garnett 2013). Relational autonomy implies not being subject to other people's dominating power (Friedman 2003; Garnett 2013, 2014; MacKenzie and Stoljar 2000; Oshana 2006).

All these theories have in common that they pick out as objectionable social relations that give some persons the power to impose their will on others. Moreover, unlike liberal theories of freedom, they view such power as a source of unfreedom, even if it remains unexercised. However, these theories also have important differences whose exploration is beyond the scope of this article. In discussing the objection from alien control, I thus henceforth focus on contemporary republicanism, particularly as defended by Pettit.⁹

With the theoretical framework laid out, we can now appreciate how previous defenses of nudging fail to answer the objection from alien control.

First, when responding to worries about personal freedom, proponents of nudging typically argue that nudging leaves freedom of choice, and in extension liberal freedom, intact. For example, in retirement, it is up to individual employees whether they want to save more for retirement—a default does not remove the freedom to choose. However, this answer does not disarm the objection from alien control. Unlike liberal theories of freedom, the theories surveyed above hold that the absence of external constraint alone is insufficient to guarantee freedom. Alien control can be a source of unfreedom even in the absence of actual external constraints.

Second, when responding to worries about personal autonomy, proponents of nudging often argue that, not only does nudging leave freedom of choice intact, it can also help people further their own subjective interests. If nudgees are nudged away from decision making that runs counter to their own interests, nudging might even enhance autonomy (Felsen and Reiner 2015; Mills 2015; Sunstein 2015b).¹⁰ For example, many employees would indeed like to save more for retirement. However, when confronted with too much choice, they often fail to sign up for any scheme at all. A default nudge as in retirement might help people fulfill their long-term preferences and thus increase their autonomy. However, this answer does not disarm the objection from alien control, because it assumes an individualized, volitional account of autonomy. Relational theories of autonomy, on the other hand, worry about social relations that allow one group of people to impose their will on others. Moreover, this worry remains even if the imposed will happens to coincide with what nudgees want. Relational theories of autonomy hold that being subject to other people's will raises different, and typically stronger, normative worries than being subject to faults in one's own decision making.

Therefore, an appeal to freedom of choice and nonrelational, volitional autonomy does not defuse the objection from alien control. Instead, we need to show that the power to nudge can be suitably transparent and controlled such that it is not a form of alien control. The transparency argument above states that such control is impossible, because nudges are insufficiently transparent to be suitably controlled. So, to disarm the objection from alien control, my first order of business is to address the transparency argument.

TRANSPARENCY

In more recent writings, Sunstein explicitly holds that nudges should be transparent, “[if] officials issue a regulation. . ., they should disclose what they are doing, and. . .why they are doing it (with reference to any behavioral bias, if that is the reason they are doing it)” (Sunstein 2015a, 428–9). Some critics, however, argue that nudging cannot be suitably transparent (premise D of the transparency argument):

1. D. By its very nature, the power to nudge and the exercise thereof cannot be suitably transparent.

In this section, I argue that the transparency argument fails, because premise D is false.

Does Transparency Make Nudging Ineffective?

The first argument for premise D is that systematic nudging faces a dilemma: either nudgers nudge secretly, in which case the power to nudge is a form of alien control. Or they nudge transparently, in which case nudging might just stop working (Bovens 2009, 217; Grüne-Yanoff 2012, 638). Nudging exploits people's cognitive biases. If people are made aware of their biases, they become immune to the seductive power of the nudge.

However, there are several problems with this argument:

First, neither Bovens nor Grüne-Yanoff provides direct empirical support for their claim that transparency will make nudging ineffective. Some preliminary evidence suggests that, at the very least, making nudging transparent does not make nudging ineffective. Loewenstein et al., for example, argue that informing people about the use of default nudges for decisions about advance directives does not lower their effectiveness (Loewenstein et al. 2015). Wansink shows that the size of food containers and serving bowls has a bearing on how much one eats even if one is made aware of such effects (Wansink 2007, 60–70). Of course, these results do not settle the empirical question. But, at the very least, they give us reason to think that transparency is likely not going to make them completely ineffective. Sometimes it might not even affect their effectiveness at all.

Second, and more systematically, let us chart the different possible reactions people might have when they become aware of a nudge. How should we expect them to respond when they find out about a nudge?

Consider first cases in which a person recognizes a nudge whose direction runs counter to the person's considered preference. For example, you enter a restaurant really desiring a burger with fries. The default combination is a burger with a salad, which you immediately recognize as a health nudge. You decide to have fries. Now, such an outcome does not present a systematic problem for proponents of nudging, if they are earnest about respecting people's subjective interests.

Consider next cases in which someone recognizes a nudge whose direction accords with his or her preferences. In such situations, we might have two possible effects. The first is the preference effect: upon discovering he or she is being nudged, a person will act according to the nudge, because the person is aware that doing so accords with his or her preferences. For example, my employer uses opt-out nudges for retirement plans (see retirement above). Upon recognizing the nudge, my reaction was to reflect a bit more about the reasons behind this nudge and, upon acknowledging the importance of saving for old age, to go along with it. The second, less felicitous effect is the reactance effect: upon discovering he or she is being nudged, a person might act contrary to the nudge even though going along with it would be better given his or her preferences. Some preliminary evidence suggests that some people report they would display reactance and reject a nudge, even if the nudge lined up with their preferences (Jachimowicz, Duncan, and Weber 2016).

Which effect should we expect to carry the day in real life? First, the evidence so far suggests that the preference effect is significantly stronger than the reactance effect (Loewenstein et al. 2015; Sunstein 2016, sec. 2; 4). Second, the reactance effect is not inevitable and can be reduced by adjusting the type and presentation of a nudge (Jachimowicz, Duncan, and Weber 2016). Third, when it comes to the reactance effect, we should be careful not to extrapolate too hastily. While some reactant individuals report in surveys that they would reject a nudge, they might still act very differently in real life when important interests are at stake (Sunstein 2016, 31). Finally, nudges overall meet with high acceptability, typically higher acceptability than more interventionist methods such as taxation or mandates. Acceptability for nudges mainly depends on whether people agree with the intended purpose of a nudge, on whether a nudge coheres with their values and interests, and on whether a nudge is perceived as effective (Petrescu et al. 2016; Sunstein 2016). This provides further prima facie reason to believe that if fully transparent nudges successfully track people's interests and convictions, the preference effect should outweigh the reactance effect. Overall, while it is certainly possible that transparency might make nudging a little less effective, we can reasonably assume that transparency is not going to make nudging completely ineffective.

Finally, let us assume, for the sake of argument, that transparency would make nudging less effective but not fully ineffective. If so, I suggest we should simply accept some tradeoffs. Inasmuch as we disvalue relations that allow persons to impose their will on others, we should accept that we lose some effectiveness by making nudging policies transparent. This might well be a price worth paying and in no way calls into question the rationale for nudging policies more generally.

In conclusion, the first argument for premise D—that transparency would make nudging ineffective—fails.

Just How Transparent Should Nudging Be?

The second argument for premise D is this: to prevent alien control requires a demanding notion of transparency—so demanding in fact, that it is impossible to implement.

How transparent does nudging have to be to avoid alien control? Is it sufficient if the general types of policies are transparent? Or should each token, i.e., each individual application, of these policies be transparent too (Bovens 2009, 217)? Thaler and Sunstein's example of subliminal messaging illustrates that mere type-transparency is insufficient:

What if the government openly announces that it will be relying on subliminal advertising in order, for example, to combat violent crime, excessive drinking, and the failure to pay one's taxes? Is disclosure enough? We tend to think that it is not—that manipulation of this kind is objectionable precisely because it is invisible and thus impossible to monitor. (Thaler and Sunstein 2008, 245–6)

Grüne-Yanoff thinks that type transparency is insufficient and that we need token transparency: people should know in each instance whether they are being nudged. However, such a demanding notion of token transparency seems impossible to implement:

1. A. For a power to enact certain policies to be suitably transparent, each token of these policies needs to be recognized by those affected by them.

2. B. Given the very nature of nudging, it is impossible to ensure that each token of nudging policies is recognized by those affected.

3. C. Therefore, by its very nature, the power to nudge and its exercise cannot be suitably transparent.

However, this notion of token transparency is implausibly demanding. Imagine a person who fails to recognize an easily recognizable token of a nudging policy, because the person is just too absentminded in that moment. Requiring every token of nudging policies to be recognized seems too strong. Instead, we might hold that each token should be recognizable. But how recognizable? If we stipulate that each token should be easily recognizable, i.e., without any effort or research, the transparency argument against nudging would presumably still hold:

1. A. For a power to enact certain policies to be suitably transparent, each token of these policies needs to be easily recognizable by those affected by them.

2. B. Given the very nature of nudging, it is impossible to ensure that each token of nudging policies is easily recognizable by those affected

3. C. Therefore, by its very nature, the power to nudge and its exercise cannot be suitably transparent.

Let us grant, arguendo, that B is true and that it is unfeasible to make each token of nudging easily recognizable. But is premise A plausible? Not very, I think. Instead, we should follow Bovens. He holds that instead of each token being either recognized or easily recognizable, it is enough if tokens of nudging policies are detectable in principle by a “watchful person” (Bovens 2009, 217). A watchful person is someone who, with not unreasonable effort and understanding, would be able to detect a token of a nudge and would comprehend the intention behind it. It should not be unreasonably costly to seek and find relevant information about the general types of nudging. Moreover, information about types should make it possible to recognize tokens of these policies through inference and without requiring unreasonable effort. Let us call this notion “reasonable token inference transparency” or “RTI transparency” for short. What RTI transparency implies practically depends on how it can best facilitate individual and democratic control (though, of course, no precise formula can be devised from the armchair).¹¹ A detailed outline of democratic control has to wait until the next section. But let me already anticipate two points here, as they are relevant for RTI transparency.

First, exercising control over a power to interfere is different from giving consent to such a power. In the candy cabinet example, I retain control over my partner's power to lock the cabinet. This is different from giving consent to someone's power to interfere and then losing all further control over it. To suitably control the power to nudge also requires that the citizenry have continuing control over the government's power to nudge. RTI transparency thus requires not just transparency before a nudge is enacted but also afterwards such that continued control is feasible.

Second, a richer notion of transparency requires making transparent not only which nudge policy is to be enacted—and has been enacted—but also how and why. As we will see below, republican notions of democratic control set great store by deliberation and contestation. For the citizenry to exercise deliberative and contestatory power requires being in a position to assess a policy before and after it is in place. This, in turn, requires transparency about policies and their intended ends.

I argue now that we should typically expect nudging to fulfill RTI transparency. The novel argument I provide is an argument from comparison.

When Hausman and Welch argue that nudges are insidious, they contrast nudges with a mandate to wear seatbelts. However, a great many interferences are much less straightforward and visible than seatbelt laws. In which sense, do we require of those policies and regulations that they be transparent? For example, when you buy a car, the car is the result of a number of complex regulations. Many of those regulations are meant to protect you, others are meant to protect other people and the environment. Or consider that many drugs require a prescription. A fairly complicated process precedes official decisions as to whether a drug requires a prescription or is available over the counter. Moreover, to have a drug approved by a body such as the FDA is very complex to begin with. Or consider food consumption choices in the European Union. A number of regulations, subsidies, international tariffs, and so on, determine what products are available to consumers and at what cost. For a typical European consumer, it is a complex matter to determine how precisely that affects his or her vegetable and fruit options.

As these examples illustrate, the tokens of many public policies influence our freedom of choice in ways that are far from obvious to the average or even the informed consumer. However, with some effort, such tokens can in principle be recognized and understood. To recognize how one's freedom of choice is being influenced, a person needs, first, to be in a position to find out about the overall type of regulations (car manufacturers need to abide by safety standards, drugs are regulated, and so on) and, ideally, why these regulations are in place; second, he or she needs to be, in principle, able to recognize individual tokens of these regulations (“this is a product that is subject to safety legislation and such legislation influences the choices I have now”). Finally, the person needs to be in a position to find out more about how the regulation is carried out and how it affects his or her choice in that instance.

Now, we can imagine that the same points apply to nudging too. First, we can imagine open debates and announcements about what types of nudges are being employed and to what end. Second, to identify instances of nudging and to understand the mechanisms through which they work is typically not more difficult, and often easier, than many other instances of public policy. It is probably easier to understand cognitive biases and nudging policies than to understand the intricacies of complex pharmaceutical and car safety regulations or the EU's Common Agricultural Policy. Consider, for example, moving from an opt-in policy to an opt-out or mandated choice policy. For most persons, such a change is easy to comprehend. Moreover, comprehending the intention and the mechanism through which this nudge works should not strike us as more complex than most other cases of public policy. In other cases, it is even hard to miss a nudge. Consider the mundane example of the fly sticker used in many Dutch urinals to reduce spillage. Such a sticker and its purpose are easily recognized. Finally, some nudges are easy to comprehend, but recognizing each token is not straightforward. For example, if cafeterias change the order of their food or use different container sizes to nudge people towards healthier eating, we can easily understand such nudges, but we might not always spot them easily. Nonetheless, it would not be unreasonably costly or difficult to spot such nudges for those who are watchful.¹² So, compared with other types of public policies, it seems there is nothing so special about nudges that would prevent suitable transparency.

Overall, it is feasible to make nudging policies suitably transparent. Therefore, the transparency argument fails.

DEMOCRATIC CONTROL

Furnishing government agencies with the power to nudge is not a source of alien control, if such power is suitably transparent and democratic. In the preceding section, I argued that it is institutionally feasible to make nudging suitably transparent. This already goes a long way towards establishing that nudging can also be democratically controlled, because it defuses the transparency argument. Nonetheless, transparency is merely a necessary but not a sufficient condition for the absence of alien control. We could imagine nudging policies that are suitably transparent but not suitably controlled. I therefore now argue that it is feasible to place nudge policies under suitable democratic control.

We can easily imagine how, on the individual level, one person suitably controls another person's power to nudge. Modifying our candy cabinet example, imagine person A asks person B to nudge him or her into a healthier diet by serving dinner on smaller plates and by placing fruit in easily accessible snack locations. Because B’s nudging is still under A’s control, it is not an instance of alien control. But while suitable control is easily imagined on the individual level, how precisely does such control work for public policy nudges? Republicans typically care both about preventing alien control between persons (the horizontal dimension) as well as preventing the state from exercising alien control over its citizens (the vertical dimension). To prevent alien control on the latter, vertical dimension requires democratic control (or public control as republicans sometimes call it).

Given limited space, I here provide merely the bare bones of a contemporary republican model of democratic model. But having its essential features in place allows us to see that nudging can indeed be suitably democratically controlled.

According to Pettit, “state interference will not be dominating. . . so long as it can be subjected to the effective, equally shared control of the people.” (Pettit 2014, 111) This view implies, first, that people have influence over the political process in that they make a difference to what political outcomes are chosen. Second, such influence has direction: popular influence does not make a merely wayward but a designed difference, a difference that tracks some preconceived or preferred pattern (Pettit 2012, chap. 5; 2014, 121). Third, access to a system of directed influence in a political process should be shared equally between citizens (Pettit 2014, 123).

How does this notion of democratic control play out institutionally? Republicans, of course, endorse democratic elections. But elections are not enough. Republicans also set great store by a mixed constitution and contestatory citizenry. A mixed constitution implies that an electoral system be complemented with a system of checks and balances, typically by appointing regulatory authorities, constitutional courts, and so on. A contestatory citizenry requires a citizenry that invigilates and interrogates governmental decisions, policies, and political processes. Such interrogation, Pettit argues, is not limited to individual citizens:

In a complex contemporary society, the only hope of a systematic, encompassing interrogation of government lies with public interest movements and bodies. Operating in public space, and guided by publicly accepted standards, such nongovernmental organizations can specialize in different domains and develop the expertise required for keeping tabs on the performance of government as a whole. One organization will specialize in consumer affairs, another in health policy, another in women's rights . . . The best hope for a flourishing democracy lies with the prospect of a people who are active and engaged enough to give life to such movements and of authorities who are willing—willing and electorally forced—to give recognition and attention to these, their harshest critics. (Pettit 2014, 148–9)

For control through deliberation and contestation to be feasible, political decisions need to be transparent. Moreover, public debate and the defense of political decisions need to follow the deliberative “norm of norms”: reasons need to be publicly provided and such reasons “should be relevant from the standpoint of every adult, able-minded citizen.” (Pettit 2014, 134; see also Sunstein 1988, 1548–51).

The republican model includes ample room for deliberation and participatory engagement. But laws and policies can be nondominating for particular individuals, even if they have not actively participated in their creation or consented to them. For individuals can have indirect control over their creation and continued existence through representatives, spokespersons, and public interest bodies. Moreover, individual control need not always be active but can be virtual or reserve control. Accordingly, if a political process or set of policies were to go against my relevant interests, I would have adequate opportunities to contest them in the public arena, either directly or through indirect control as exercised, for example, by public interest bodies.¹³

With this account of democratic control in hand, is it feasible to place nudging policies under suitable democratic control? I think the answer is yes.

The first obvious point goes back to the last section: the challenge of democratic control applies, of course, to all policies both behavioral and nonbehavioral. The main reason people have worried about nudges escaping control is their alleged lack of transparency. I have argued that this worry is unfounded. For that reason, it is hard to see why nudging policies should in principle escape democratic control any more than traditional nonbehavioral policies. But the second, more interesting point is that a number of features make nudging particularly amenable to be suitably controlled.

First, because nudging policies are relatively easy and cheap to implement, they are particularly suitable to be devised and implemented at various levels. If implemented at lower institutional levels, or only in particular areas, nudges might be controlled more locally, which can make democratic control more effective. Nudges might also be implemented in particular regions, in particular industries or even in particular local institutions (such as schools). Now, many of these institutions will have their own mechanisms of democratic control and their own public interest bodies or representatives. Industries might have workers’ councils or trade unions, schools might have student representatives, hospitals might have ethics review boards or might interact with patient organizations, and so on. Localization might thus often facilitate further democratic control over nudges.¹⁴

Second, while some nudges will be the same for everyone, others might be customized towards individual needs and preferences. Some of these personalized nudges allow individuals to actively choose how or whether they are nudged. For example, individuals might have a choice over what kind of default applies to them in a particular choice domain (Sunstein 2015b, chap. 6).¹⁵ Such “choices over choices” will give individuals more direct control.

Third, two more features of nudging make them suitable for deliberation and contestation. For a citizenry to hold contestatory power requires transparency, an informed citizenry, adequate public discourse, a free press, a free and functioning academic and scientific sector, and so on. To hold contestatory power with respect to particular policies also requires adequate information, analysis, and public discussion of those policies. So far, nudging initiatives in the United States and the United Kingdom have received a great deal of media attention, plenty of attention in the academic community, and reports about such nudging initiatives are publicly available. Moreover, proponents of nudging typically favor an evidence-based approach to public policy. This means that the effectiveness of nudging interventions is—typically more so than other public policies—subject to rigorous scrutiny before and after nudges are put in place. So, given that nudging is widely discussed and relevant evidence is systematically collected and available, the deliberative and contestatory conditions for nudging policies are favorable.¹⁶

This completes my response to the objection from alien control: first, granting one group the power to nudge others does not increase, or bring about, alien control, if such nudging is both suitably transparent and suitably democratically controlled; second, such transparency and control are practical and feasible. Therefore, the objection from alien control fails.

My argument is meant to show that noninstrumental worries about alien control do not give us a principled reason against nudging. But, additionally, my response might also help alleviate some instrumental worries regarding the possible content of nudges. For example, some critics worry that systematic government nudging might fail to reliably pick out the right or best option, might be exploited for perfectionist or even illicit ends, and lead us onto a slippery slope towards overbearing paternalism (Grüne-Yanoff 2012; Rebonato 2014; Rizzo and Whitman 2008; Whitman 2010; Whitman and Rizzo 2015). Of course, to some extent, such worries apply to many types of public policy and nudging has a comparative advantage, because it always leaves individuals with the option to reject the nudge. But my argument also provides an additional response. If we follow Pettit, implicit in a functioning system of democratic control is the deliberative norm of norms, roughly a commitment to publicly providing nonpartisan reasons for political decisions. Furthermore, a functioning democratic system will over time also develop more specific community-wide norms, such as that people should enjoy religious freedom, that speech should in general be free, that people have privacy rights, and so on (Pettit 2012, 258–75; 2014, 136–42). We can reasonably expect that such norms, backed up by a mixed constitution and a contestatory citizenry, will help filter out nudge proposals with illicit ends. And studies indeed suggest that people across political lines most strongly disapprove of nudges that seem to violate such community-wide, nonpartisan standards (Sunstein 2015a, 428; 2016, 15). Moreover, such norms and a contestatory citizenry should make it less likely that policymakers pursue nudges that would meet with excessive disagreement or controversy, or that pursue partisan political goals or a highly particular conception of the good. So, while my intention here is to disarm noninstrumental worries about alien control, my argument suggests that suitable transparency and democratic control might also alleviate some instrumental worries regarding the possible content of nudge policies.

HOW NUDGING CAN REDUCE ALIEN CONTROL

So far I have argued that, if done right, nudging policies do not subject individuals to alien control. In this part, I want to defend a stronger claim: if done right, systematic nudging might even reduce the extent to which individuals are subject to alien control.

The basic idea of the containment argument is this: through a system of public policy nudges, we can contain the power of private agents, particularly private companies, to influence people through uncontrolled and oftentimes opaque nudges. Unlike current private nudges, such a public nudge program would be transparent and democratically controlled. By containing insufficiently controlled private power to nudge, such a program can thus contribute to the democratization of nudges in people's choice environments and thereby reduce alien control. This can happen through different avenues. For example, sometimes we might counteract private nudges with public nudges; sometimes we might replace private with public nudges; or we might direct private nudges through regulation. Let me spell out the containment argument in more detail.

Remember the inevitability argument: as nudges are inevitable, we might as well nudge people in a way that has good rather than bad outcomes. Critics typically hold, however, that in this simple form the inevitability argument overlooks the distinction between intentional and unintentional nudges. The former raise worries about social relations, whereas the latter typically do not. Now consider more closely the class of intentional nudges.

In response to the objection from alien control, I have argued that intentional public nudging can indeed be suitably transparent and democratically controlled. Among other things, such public policy nudges would thus have to be proposed and defended openly; proper consultation and involvement of the public and relevant public interest groups would have to be sought; continued assessment of such policies should be provided; and such policies should be open to contestation.

Such public policy nudges would thus differ starkly from the intentional behavioral interventions used in private contexts, particularly as employed by private companies (henceforth “private nudges”). Private nudges are also intentional. However, private nudges typically do not fulfil RTI transparency nor are they under suitable democratic control. I do not mean for this point to be purely theoretical. Uncontrolled and opaque nudges are all too real. Nudging is nothing new for marketers and advertising companies who have been doing this for decades (Akerlof and Shiller 2015, chap. 3). In their recent book Phishing for Phools, economists Akerlof and Shiller introduce the concept of phishing which denotes forms of communication and choice architecture that trick consumers into behavior that is advantageous for the phisher but typically disadvantageous for the person being phished. Akerlof and Shiller hold that phishing occurs in nearly all areas of economic activity. Moreover, rather than phishing being an occasional aberration, Akerlof and Shiller argue that markets provide systematic incentives to phish. Consider, for example, how food companies systematically prey on our weakness of will when it comes to sticking to healthy diets, systematically exploit that consumers have false beliefs about how healthy certain products are and use salt, fat, and sugar in processed food in ways that get us hooked (Moss 2013). Or consider how the credit card system often entices consumers to overspend and borrow excessively or how health clubs prey on people's optimism bias (“this time I will definitely exercise three times a week”) by locking them into disadvantageous contracts. Finally, consider how large internet firms extract personal data by using lax privacy settings as defaults. The list of such mechanisms is long and affects virtually all areas of economic activity.

The idea of the containment argument is to use a system of nudge policies to contain the power of private companies to nudge. Such containment can be achieved through various avenues.

First, the strongest intervention is to stop a particular private nudge and replace it with a public nudge. For example, the design of cigarette packages might influence individual purchase decisions. Common measures in tobacco control include imposing plain packaging laws or removing some of the private packaging to add health warnings or images of smoking-related diseases. Second, a systematic nudging program could be used to regulate what nudges private companies should and should not use. We might regulate that companies should stop using unfavorable default nudges. For example, we might require internet firms, such as Facebook or Google, to apply stricter default privacy settings. Or we might regulate that companies use beneficial nudges, such as using smaller food plates or drink containers or change the order of food in cafeterias or supermarkets. Third, public nudges might be used to counteract private nudges. Hereby, we would leave nudges in place but try to reduce their effectiveness by adding public nudges. For example, food companies nudge through food packaging. They might use colors and pictures that people associate with health even for products high in sugar, salt, or fat. We might decide to leave such private nudges in place but nudge back by, for example, adding a color-coded traffic light rating system, as developed by the Food Standards Agency in the UK. Finally, a regulatory system overseeing private companies, or even cooperating with them in nudging activities, could exercise virtual control over private companies. The threat of regulatory power and the effective interrogation through public interest bodies might in some instances be enough to make private companies abandon bad nudges, or even adopt beneficial nudges, without the need for direct regulation.

Two reasons support containing uncontrolled private nudging through a democratically controlled public nudging program.

The first reason is instrumental. Akerlof, Shiller and others highlight that private nudges often make us worse off—more in debt, less healthy, overweight—and often run counter to our own interests as we perceive them. The instrumental argument is that we are likely to achieve better outcomes with a system that contains private power to nudge than without one. Why is that? If we do not contain the private power to nudge, we lack the necessary control to ensure that private nudges are oriented towards the interests of the nudgees or society more generally. First, private nudges typically lack transparency: a company will rarely make publicly available all aspects of their marketing strategy; nor will they give reasons for why they choose a specific strategy over another. What is more, sometimes faulty information is provided to customers. For example, making continued membership in a plan the default, a company might claim that it wants to make life more convenient for its customers. Their real motivation, however, might be to use people's status quo bias to increase profit. Overall, with such insufficient transparency, controlling private nudging is difficult. Second, private nudgers are typically not systematically pressured to align their nudges with the interests of nudgees or society more generally. In deciding whether to implement a particular marketing strategy, for example, a company will typically be more constrained by the interests of its shareholders (and maybe other stakeholders) rather than the interests of its consumers and society more generally. Now, containing the private power to nudge through a democratically controlled nudge system involves mechanisms that put pressure on nudgers to orient nudges, and the absence thereof, towards the interests of the nudgees and of society more generally. Therefore, containment would likely result in better choice environments and thus achieve better outcomes overall.

Now, admittedly, my list of private nudges so far has been somewhat tendentious, as I have mainly listed private nudges that were disadvantageous for consumers. But private nudges can also be beneficial. For example, while health clubs often nudge customers into unfavorable contracts, some also use nudges to encourage more exercise. Moreover, websites might help you reduce your cognitive load by using defaults based on choices you have made in the past (Sunstein 2015b, 3–5). But this nuance is, of course, consistent with the overall tenor of the instrumental argument. The claim is merely that containing the private power to nudge will likely lead to better outcomes overall, not that all private nudges are bad. Moreover, remember that containment does not mean getting rid of private nudges. Often we can leave private nudges in place but use regulation to achieve a better fit between nudges and people's interests. Sometimes we might even improve the outcomes of private nudges already by exercising merely virtual control through good oversight and the threat of possible regulation.

But next to the instrumental, there is a noninstrumental reason for containment. If left uncontained, the power of private companies to nudge is a form of alien control. Such power is neither suitably transparent nor democratically controlled. Moreover, private nudges are pervasive, which means we are often subject to alien control in many of the central decisions in our lives, for example, choices about what food to ingest, which far-reaching financial decisions to make, whom to give our personal information, and so on. The relational, noninstrumental argument is that containing the private power to nudge might not only make it more likely that nudging will track people's interests, it will also remove social relations that allow some persons to impose their will on others. Inasmuch as we disvalue relations that embody alien control, we have reason to contain uncontrolled and opaque private nudging. Containing such private power contributes towards a democratization of choice environments for some of the central choices in our lives. Such democratization is not only intuitively attractive but also strikes a chord with the broad range of philosophical theories—republicanism, relational autonomy, etc.—surveyed earlier.¹⁷

The containment argument thus is this: in many situations, containing uncontrolled private nudging through a system of suitably transparent and democratically controlled nudging is not only more likely to lead to better outcomes, it will also decrease alien control. A system of transparent and democratically controlled nudging contributes towards democratizing control over choice environments.¹⁸

This conclusion is both important and surprising. As explained above, many authors worry that systematic nudging would lead to an increase of uncontrolled power over individuals. I have here argued that, under suitable conditions, the opposite can be true. Moreover, some authors who write specifically on nudging and democracy have expressed the worry that a nudging state would diminish the role for public debate and deliberative democracy. John, Smith, and Stoker, for example, worry that instead of seeing individuals as rational and deliberative citizens, top-down behavioral policies see individuals as largely unable to change their own preferences and behavior. John, Smith, and Stoker argue that instead of nudging, we should seek to effect systematic behavior change by allowing agents to adjust their own behavior after involving them in relevant debates in deliberative fora (what they call “think” proposals) (John, Smith, and Stoker 2009). But in light of the arguments in this article, I think we have good reason to reject this presumed dichotomy and such democracy-based objections to nudging. Rather than nudging undermining democratic control, I have argued that widespread public policy nudging can itself be a vehicle for heightened democratic control, namely by increasing control over the external, social context within which we make decisions.

Before concluding, let me briefly address an important objection to the containment argument. One might object that well-functioning markets are in an important sense democratic. Consumers exercise power in markets, because they possess an “exit” option (to use Hirschman's term (Hirschman 1970)). This also means that if a consumer is unhappy about certain nudges—or about a company's marketing strategy—they can take their business elsewhere. Therefore, to control private nudging, a system of public nudging is not necessary—maybe even counterproductive—because consumers hold sufficient power in virtue of their exit option. However, there are a number of problems with this objection.

First, very rarely will consumers be in a situation to exercise control over whether and how companies employ nudges. When someone takes their business elsewhere, this will often not be information enough for the company to know what precisely they should change. Should they lower their prices? Should they have a different—maybe more aggressive—marketing strategy? The channel from consumer demand to influences on their marketing is usually very indirect. In fact, in the face of dwindling demand, a company might respond with more phishing rather than less.

Second, I have suggested that effective control over nudging interventions requires RTI transparency. But private nudges typically do not fulfil RTI transparency, which makes it hard for individual consumers to effectively control them.

Finally, and most importantly, if nudging is successfully employed to elicit purchase decisions, then consumer demand will typically reinforce private nudging rather than reduce it. If a nudge works particularly well, it will trigger purchase decisions or bind consumers to a company. Thus, rather than containing private nudges, consumer demand can often reinforce the use of such nudges.

Overall then, we might stand a better chance to effectively control private nudges with a system of public nudges in place instead of leaving it all to consumer behavior. The containment argument should be allowed to stand.

CONCLUSIONS

I started this article by outlining a common objection to nudging, i.e., the objection from alien control: giving one group of persons the power to nudge others increases the extent to which one group can exercise alien control over another. I have argued that this is not the case, if such power is suitably transparent and under proper democratic control. Moreover, it is indeed institutionally feasible to make nudge policies suitably transparent and democratically controlled. A general worry that systematic nudging would lead to problematic forms of social control is thus unfounded. I then argued that this response also provides us with a novel, positive argument for more nudging: through a system of transparent and democratically controlled public policy nudging, we can contain uncontrolled private power to nudge and thereby increase democratic control over choice environments. This is particularly so, given how frequently individuals are subject to insidious nudges from private companies in ways that escape individual or democratic control. So, if you worry about the uses of nudges in public policy, you should probably be even more worried about how similar methods are already being used by companies and marketers. We should have a say about the direction into which we are being nudged. We can do so by making nudging transparent and by placing it under democratic control.

The argument also shows that, in future work on nudging, we should pay particular attention to the kinds of institutions through which we develop and implement nudging policies. Specifically, we should make sure the process and the content of nudging policies are transparent, that their institutional implementation is subject to proper democratic control, and that such policies are rigorously assessed. For example, such institutional questions should be asked in connection with the recent part privatization of the Behavioural Insights Unit in the UK (Dunt 2014; Niker 2014). They are also relevant for behavioral policy interventions in development settings, particularly when those to be nudged are vulnerable populations. A challenge for future large-scale nudging policies will be to ensure that the relevant institutions and procedures will remain suitably transparent and democratically controlled. If they are, they might not only increase overall welfare but also facilitate democratic control over the social contexts within which we make choices.