In his thought-provoking article “Should evolutionary economists embrace libertarian paternalism?” (Journal of Evolutionary Economics 24(3), 2014, 515–539), Martin Binder contributes to the discussion on an innovative policy program that engages policymakers and the wider public on a global scale. “Libertarian Paternalism”, as popularized by Richard Thaler and Cass Sunstein (2003, 2008) – henceforth TSLP)Footnote 1 – continues to make a huge impact in policy and scientific circles. Its key instrument, “nudges”, involves the purposeful design of choice environments (“choice architectures” in TSLP lingo) in ways that would look irrelevant from the point of view of Homo Economicus. Real-world agents, however, are influenced by details of the situational context such as the framing of options or the setting of defaults. This fact may then be used, e.g., to stimulate healthy food consumption, higher savings rates or higher willingness to donate organs, while preserving people’s freedom of choice.

Due to the sometimes poor way Thaler and Sunstein defend their program, though, it has turned out to be easy to criticize from a variety of angles. Most importantly, critics fear a loss of individual autonomy: after all, most nudges operate through intentionally exploiting people’s cognitive biases.Footnote 2 Binder adds an important twist to the critical literature by examining this new policy tool from a perspective that sees the economy as an evolving, novelty-generating system and individuals as adaptive learners. Specifically, he wants to analyze the “extent to which subscribing to an evolutionary methodology and worldview makes libertarian paternalism a plausible policy tool or not” (523) From that perspective, he identifies three arguments in favor of LP and six against it.

The aim of this comment is fourfold. First, I argue that Binder starts his discussion from an implicit model of the institutional status quo that is incomplete in that it neglects the empirical prevalence of “private nudges”; second, he neglects the full implications of an evolutionary worldview for the policy problem at hand; third, Binder overlooks a promising constitutional way to refine Thaler & Sunstein’s LP program (let’s call it constitutional libertarian paternalism, or CLP); fourth and finally, I show that in light of a possible CLP variant, four of Binder’s six arguments against LP lose much of their force, tilting the argumentative balance in favor of such a refined variant of LP.Footnote 3

As to the first point, Binder’s discussion of the pros and cons of TSLP seems to start from the premise that the institutional status quo is entirely free of any nudging. Common sense tells us otherwise. As Julian Reiss (2013) has pointed out, “humans with bounded rationality and willpower are subject to myriad influences anyway, and most of them do not aim to improve consumer well-being” (ibid.: 299, see also Kirchgässner 2013). Acknowledging that fact is important, if only for indicating a potentially useful role any LP program might play: Many real-world agents (i.e. those with bounded rationality and willpower) could reasonably request governmental help in countering the multitude of private nudges they are subject to on a daily basis—in other words, a case could be made for a widespread demand for counter-nudges. These may include both measures to enable agents to recognize and neutralize the effect of nudges on their own and governmental counter-nudges. As to the latter, even if one can argue that market-generated nudges tend to be less harmful (in terms of welfare or autonomy) than governmental ones,Footnote 4 denying sovereign citizens the constitutional right to nonetheless demand such counter-nudges seems a non-sequitur (Kirchgässner 2013; Vanberg 2014).

My second point is about what it means to adopt an “evolutionary” viewpoint. Binder focuses on the problematic dynamic side-effects of nudging only, overlooking the full implications of such a viewpoint for the agents populating an evolving economy: in a novelty-generating economy, agents are continuously faced with the problem of coping with hitherto unknown goods and services with a tendency to gain in complexity over time, giving rise to unfamiliar decision contexts. In other words, agents are vulnerable to being manipulated by private nudges on a regular basis, which makes the call for corrective counter-nudges ever more urgent.

My third point builds on that observation: It concerns a potential refinement of Thaler & Sunstein’s LP program that Binder does not address: as we will see below, most of the objections he adduces can be neutralized by adopting a constitutionally constrained variant of LP (CLP). Such a variant would address the citizens directly, rather than some social planner. Nudges would then be legitimate to the extent that they reflect the common constitutional interests of the citizens (as the principals of policy-makers and bureaucrats).Footnote 5 Citizens, aware of their own cognitive biases and lack of self-control, may wish to implement governmental counter-nudges as part of wider collective self-commitment strategies. The rationale is easy to see: Assume (i) that individuals have reflective second-order preferences (i.e., preferences over preferences); they benefit if their own first-order preferences get better aligned with their own second-order preferences: To illustrate, if they don’t want to crave for smoking, they are better off if they actually don’t want to smoke. Assume (ii) that individual first-order preferences are “adaptive”, i.e., partly shaped by the set of opportunities provided in the marketplace. A given opportunity set can either cause both sets of preferences to be better aligned (“improved” from the individual’s own point of view) or not. The former case clearly involves a positive externality, which the market likely underprovides (George 2004). Hence the need for collective agreements – that increase the likelihood of alignment – at the constitutional level. To be sure, apart from counter-nudges, CLP may also encompass debiasing tools that support agents in immunizing themselves against being tricked by any kind of nudges, private or governmental (Rebonato 2012; Binder and Lades 2013). Importantly, CLP would need to be much more transparent than TSLP, even if this would reduce its instrumental effectiveness (see below). In any case, exploring the outlook of a constitutional libertarian paternalism is a promising agenda for future research.

Fourth, let’s take a brief look at the objections Binder discusses against TSLP—I will argue that at least four of them lose much of their force if applied to CLP. Let’s pick the entries on his list in turn. In the way he presents it, Binder’s first point (“inevitability objection”) is a rebuttal of a specific defense of TSLP rather than an actual objection against the latter’s substance. Binder seems to confound two different claims here: The first one states that given the ubiquity of framing effects, paternalism is inevitableFootnote 6; the second states that given the ubiquity of framing effects, “a planner cannot avoid framing the options and influencing decision-makers” (526). Binder goes on to state that this is “plain wrong” (ibid.). In practically all the choice problems discussed in the literature on LP, however, the second claim is valid: No choice setting is perfectly “neutral”. Binder could have turned this rebuttal into an actual objection against TSLP by pointing out that, while no choice setting is neutral, some settings are nonetheless more conducive to rational choice (or autonomy, for that matter) than others (see, e.g., Rebonato 2012: 135). Moreover, all this does of course not make paternalism itself “inevitable”—any private choice architect may design and frame the options in such a way as to maximize her private profits, say, rather than the well-being of her customers (or she may influence her customers’ decisions unintentionally). If we follow Dworkin (2014) influential definition of paternalism as any kind of purposeful interference into an agent’s choices against her wishes, then it seems that CLP would only partially qualify as paternalistic: At least at the level of constitutional preferences, individuals would agree to be nudged. Hence, while paternalism may very well not be inevitable, CLP may avoid the objectionable kind of paternalism that interferes against the agents’ wishes all the way down.

Binder’s second point (“rationality objection”) in fact contains three distinct arguments: (a) about the notion of rationality applied in TSLP in general being inadequate; (b) about the specific definition of welfare Thaler & Sunstein propose being inadequate; and (c) about the risk that in practice, TSLP may end up falling back on pragmatic ad hoc notions of welfare.

Binder is of course right in maintaining that the normative benchmarks suggested in TSLP are highly deficient, both conceptually and in terms of practicality. It’s inconsistent to uphold Homo Economicus as a normative role model, while dismissing it in the domain of positive theorizing (Berg 2003). And the “full information account” of welfare Thaler & Sunstein more or less explicitly use is a non-starter, as Qizilbash (2012) and Grüne-Yanoff (2012) demonstrate. The important point to note is this: Libertarian Paternalism is not necessarily based on such problematic concepts of rationality and welfare. Sunstein, for one, seems to distance himself from the “full information” account of welfare in recent writings (such as Sunstein 2013). As most critics – Binder included – concede, those standards can hardly be applied in practice anyway, which is why we can expect TSLP to resort, in practice, to pragmatic notions such as a country’s macro aggregates (its savings rate, say). This practice raises a bunch of new problems, of course. But it’s at least conceivable that a macro policy goal, such as “increase people’s savings rate” may be made the object of a general agreement at the constitutional level, provided dissenting individuals can indeed easily opt-out of the corresponding savings schemes, if they wish to do so.

Point three – the “compounded knowledge objection” – is usually based on the Hayekian insight that public officials (entering here as “choice architects”) are not sufficiently informed to employ nudges in an optimal way. We should however avoid a Nirvana approach in this context – truly perfect nudging would require information that is utterly inaccessible –,Footnote 7 and also refrain from one-sided generalizations: Epistemic privilege is neither always with the agents themselves nor with government officials and “experts”. Again, a constitutionally constrained LP that (i) is contestable in being responsive to the agents’ preferences and (ii) favors general over specific counter-nudges would seem less affected by this objection than TSLP.

The fourth item on Binder’s list (“conservatism objection”) is generally well-taken. As Schnellenbach (2012) shows, due to its unclear welfare foundations, TSLP can be expected to use established social norms as proxies for social welfare (see above), thereby stabilizing those norms, which may eventually slow down social change and stifle innovation. While this risk seems real in many cases, such as campaigns against obesity or drugs,Footnote 8 other cases don’t seem likely to engender problematic social norm dynamics of the sort: consider nudges for careful driving or the successful “Don’t mess with Texas” campaign against littering (Thaler and Sunstein 2008). Admittedly, though, to the extent that this fourth objection applies, it seems to apply to a CLP as well.

Argument number five is the “slippery slopes objection”, a pet issue of libertarian critics (e.g. Wilkinson 2008; Rizzo and Whitman 2009). It relies on somewhat speculative predictions about hidden psychological dynamics of political equilibria: Soft paternalism allegedly paves the way for hard paternalism, which is taken to be even more evil. For instance, it is often argued that voters may be manipulated into holding beliefs (about “undesirable” behaviors such as smoking, say) that prepare them to eventually also accept hard paternalism. But even in the context of TSLP, it seems that this argument needs much further elaboration before it should be taken seriously. For starters, one can reasonably assume that it’s the lack of transparency that makes TSLP particularly attractive for policy-makers. A switch toward – highly visible – hard paternalism would then be hard to explain in a Public Choice framework. To the extent that this argument applies, though, it may apply to CLP as well.

Finally, Binder mentions the risk that nudges may distort individuals’ preference learning in ways the individuals themselves don’t even notice (“manipulative preference learning objection”). Again, it’s crucial to see that such manipulation is already with us on a daily basis. Private firms use ever more sophisticated nudges (also inspired and informed by behavioral economics) to steer us toward spending more, borrowing too much, eating fatty stuff etc. My point is not that Binder is wrong; it’s just that his argument is incomplete. And again, I would argue that a constitutionally constrained LP would alleviate most of these concerns: It would employ generally agreed counter-nudges in a transparent, accountable and strictly controlled way. Arguably, they would be less effective under these conditions of transparency. There’s no reason, though, to assume them to become wholly ineffective: We cannot activate our reflective System 2 on a permanent basis; Footnote 9 rather, we can let ourselves be manipulated voluntarily, if (repeatedly) given the chance to critically reflect on this.

Hence, I submit that four of these six objections (to wit, 1, 2, 3, and 6) would not apply with equal force to a CLP. Among them are two (numbers 3 and 6) that seem particularly important from an evolutionary perspective. While Binder makes many valuable points about how to assess LP from an evolutionary viewpoint, I argue that he overlooks some of the normative challenges brought up in an evolving economy: Goods and services get ever more differentiated and complex, making the real threat to individual autonomy choice and information overload – resulting in cognitive overload and a growing vulnerability to being arbitrarily manipulated – rather than the restrictions of opportunity sets traditionally stressed by classic liberals. As De Marneffe (2006: 81, italics added) puts it, “personal autonomy involves more than not having one’s choices limited in ways that one does not want. It also involves having adequate control over one’s life as a whole and acting on the basis of independent reflection on what is true, right, and good. These aspects of autonomy provide reasons for government policies that limit choices and not only for policies that protect choices”. It’s often good to make one’s own mistakes, but in a large and potentially growing realm of goods and services, it may be prudent to also benefit from the experiences others have made. To do so, citizens may wish to engage in some controlled collective “self-nudging” (Lades 2014). To be sure, nudges may also be applied in order to achieve a variety of non-paternalistic policy goals, such as environmental protection.

When assessing the pros and cons of LP, we should engage in the comparative assessment of real-world institutional arrangements: Compared to hard paternalism – which may often be the only politically feasible alternative –, the “libertarian” variants have at least the advantage that they preserve freedom of choice – the normative price to pay for that (namely, the exploitation of cognitive biases) can be mitigated by thorough and open constitutional control.