It is easy to imagine Benthamite behavioral economics; it would embrace Bentham's form of utilitarianism (and draw on Bentham's understanding of human motivation). It is not as easy, but not terribly hard, to imagine Aristotelian behavioral economics; it would be rooted in a conception of human flourishing (and draw on Aristotle's understanding of human motivation). Kantian behavioral economics would emphasize the importance of treating people as ends rather than means. It may or may not have a great deal to say on the positive side – compared with Bentham and Aristotle, Kant did not focus much on people's actual motivations and behavior – but in terms of the moral foundations of behaviorally informed public policy, it would be highly prescriptive.

To be sure, general propositions do not decide concrete cases. Would Benthamite behavioral economics favor or disfavor a prohibition on the sale of cigarettes (Conly, Reference Conly2013)? Would Aristotelian behavioral economics favor or disfavor taxes on sugar-sweetened beverages (Allcott et al., Reference Allcott, Lockwood and Taubinsky2019)? Would Kantian behavioral economics favor or disfavor automatic enrollment in savings policies? Would Bentham, Aristotle, or Kant embrace behavioral welfare economics as understood by Bernheim, with its choice-theoretic foundations (Bernheim, Reference Bernheim2016)? You cannot obtain answers to these particular questions simply by reading and rereading the works of Bentham, Aristotle, and Kant. You cannot point to the categorical imperative (and thump the table). With respect to the views of dead thinkers, answers to many particular questions are inevitably interpretive in Ronald Dworkin's sense (Dworkin, Reference Dworkin1985): they must attempt (1) to fit the materials to be interpreted and (2) to justify them, that is, to put them in the best constructive light.

It is, of course, true that if an approach does not fit the materials, it cannot claim to be interpretive. But with respect to many specific questions of interpretation, more than one conclusion will do well, or well enough, along the dimension of fit, and the real question is one of justification. What looks like (1), or what purports to be (1), is often (2). That is, when a follower of Kant urges that ‘Kant would say x,’ or that ‘Kantianism entails y,’ the enterprise may be to make the best constructive sense of Kant and Kantianism, not to follow some edict of Kant himself. The interpreter might even be said to be authoring a new chapter in a kind of chain novel, whose initial chapters were written by Kant, but which Kant did not complete (Dworkin, Reference Dworkin1985).

Which understanding of Kant puts his views in the best constructive light? Because this is often the inevitable question, it should not be amazing to find that Jones’ preferred interpretation of Kant fits with Jones’ normative views, while Smith's interpretation fits with Smith's. The dispute between Jones and Smith might not be resolvable without speaking of justification – of what does, in fact, put Kant in the best constructive light. Inevitably, that judgment is an interpretation, in which we attempt to make the best constructive sense of what he actually said and thought.

My main goal in Hayekian Behavioral Economics (Sunstein, Reference Sunstein2021) was interpretive in Dworkin's sense. I sought to sketch an approach that (1) recognizes the importance and pervasiveness of individual errors, (2) emphasizes the epistemic limits of planners, (3) builds on individual choices rather than planner preferences, and (4) gives authority to choices made under epistemically favorable conditions, in which informational deficits and behavioral biases are least likely to be at work. The heart of the article, and the key step, of course, is (4). On that step, there is a large and growing theoretical and empirical literature, with important contributions from Bernheim and Rangel (Reference Bernheim and Rangel2007; Reference Bernheim and Rangel2009), Goldin (Reference Goldin2015), Allcott and Knittel (Reference Allcott and Knittel2019), Allcott and Taubinsky (Reference Allcott and Taubinsky2015), Bernheim (Reference Bernheim2016), Bernheim and Taubinsky (Reference Bernheim, Taubinsky, Bernheim, DellaVigna and Laibson2018), and others. One of my purposes was to unify that literature and to organize it under a general framework, which is insistently Hayekian in the sense that it is committed to (2) and (3), even as it is also committed to (1). To discipline (4), I pointed to five subsidiary questions:

1. (1) What do consistent choosers, unaffected by clearly irrelevant factors or ‘frames’, choose?

2. (2) What do informed choosers choose?

3. (3) What do active choosers choose? (If we focus on active choosers, we will protect against the possibility that outcomes are a product of inertia or procrastination.)

4. (4) In circumstances in which people are free of (say) present bias or unrealistic optimism, what do they choose?

5. (5) What do people choose when their viewscreen is broad, and they do not suffer from limited attention?

I also aimed to put Hayekian behavioral economics in an appealing light, though it does not, in fact, reflect my own view. I would not give decisive authority to individual choices, even under epistemically favorable conditions; human welfare is the criterion, not respect for choices, though the two are of course related (Sunstein, Reference Sunstein2020a). If we respect choices, we will often and even generally promote welfare, but in important cases, respect for choices will undermine welfare, and welfare is authoritative (Conly, Reference Conly2013; Sunstein, Reference Sunstein2020a). The task of specifying ‘welfare’ is, of course, contentious and takes us into contested philosophical waters. Are we speaking in utilitarian terms? Bentham's version, or Mill's? Might we favor a nonutilitarian form of consequentialism? Might consequentialism import deontological considerations? The best answers to these questions will, I think, lead us to question Hayekian behavioral economics, as I have elaborated it, and to undertake the specification, or at least to notice the need for it (Sunstein, Reference Sunstein2020b).

Robert Sugden (Reference Sugden2021) suggests that my article is ‘merely re-naming a familiar approach to normative economics’, which, in his account, was ‘initiated in’ an article introduced by Richard Thaler and me in 2003. Nothing could be further from the truth. With the exception of wine (he likes it, I hate it), Thaler and I tend to agree, and I would not want to saddle Thaler with a view that I myself do not hold. I can report with some authority that the view we sketched in 2003 (which I still do hold, mostly), is not at all the same as the Hayekian view I sketched in 2021 (which, to repeat, is not my own). In 2003, Thaler and I argued for ‘libertarian paternalism’, that is, an approach that preserves freedom of choice while steering people in certain directions, but we did not elaborate (4) in any way. By emphasizing certain choice-attuned ways of deciding on the right defaults (Sunstein, Reference Sunstein2020b), I suppose that we made some gestures in broadly Hayekian directions. But we did not engage Hayek's work in any terms, and we did not have a time-travel machine. Certainly, we could not have anticipated the important research and findings of Goldin, Allcott, Bernheim, Daubinsky, and others.

Rejecting my account of Hayekian behavioral economics, Sugden urges that the concepts of ‘bias’ and ‘error’ make sense only if ‘individuals have context-independent latent preferences’, which, in his view, they lack. I have no idea what the term ‘latent preferences’ means. (I have never used it, and I can't recall anyone else who has.) When information is absent and when behavioral biases are at work, things usually aren't all that complicated. We ought not to get all tied up in nouns and abstractions.

Suppose that I buy an apple thinking that it is an orange. We do not have to get fancy, or to speak of ‘latent preferences’, in order to say that I have made an error (Bernheim, Reference Bernheim2016). Or suppose that someone buys a new car, believing that it has certain safety features; suppose that he would not buy the car if he did not hold that belief. Suppose that the car lacks those features. We do not have to get fancy, or to speak of ‘latent preferences’, in order to say that the purchaser has made an error. Or suppose that someone refuses to get vaccinated against COVID-19 because she is unrealistically optimistic (she says that ‘she never gets sick’) or because of availability bias (she read about someone who died as a result of getting vaccinated). We do not need to get fancy, or to speak of ‘latent preferences’, in order to conclude that a bias has led to an error (Allcott & Sunstein, Reference Allcott and Sunstein2015). If someone says that she will finish a project on February 1 and does not finish until April 15, overconfidence, and the planning fallacy, may well have led to the mistake (Kahneman et al., Reference Kahneman, Sibony and Sunstein2021). To reach this conclusion, we do not need to say a word about ‘latent preferences’.

Maybe you don't love those examples. Matthew Rabin has coined the term ‘explainawaytions’ for efforts to explain away behavioral findings by offering some not-certifiably-crazy rational explanation, and then declaring victory. I am acutely aware that if someone refuses to get vaccinated, drinks a ton, gets addicted to opioids, or eats enough to become morbidly obese, we might be able to come up with a rational explanation; a bias might not be at work. But as President Biden likes to say: ‘Come on, man!’ (These words apply as well, I am afraid, to Sugden's suggestion ‘that the concept of error-correction, as used in behavioural welfare economics, is not philosophically coherent’.)

Sugden likes the idea of ecological rationality, and he believes that many heuristics are fast and frugal: ‘simple rules by which human beings can make the best use of their limited cognitive powers in navigating complex environments’. We agree! I like the idea of ecological rationality too, and I believe that many heuristics are fast and frugal. This claim lies at the heart of the research program inaugurated by Amos Tversky and Daniel Kahneman. Even so, heuristics that generally work well can get us into a heap of trouble (Kahneman et al., Reference Kahneman, Sibony and Sunstein2021); in some cases, they can cost us our lives. A great deal can be done to help (Hu, Reference Hu2021), and much of the help can be introduced in a Hayekian spirit.

Speaking of that spirit: Regulators all over the world are giving careful consideration to fuel economy and energy efficiency standards (Sunstein, Reference Sunstein2021). Such standards promise to reduce externalities, including greenhouse gas emissions. At the same time, aggressive standards are quite expensive, and in cost–benefit terms, they might be challenging to justify solely by reference to externalities; it might be necessary to invoke consumer savings as well. But on standard economic grounds, it is not at all clear that those savings should count. After all, consumers can buy fuel-efficient vehicles if they wish, and if they don't, they must not wish. Why force them? The behavioral response is based an empirical hunch: Consumers might be making errors, perhaps because of present bias, perhaps because of limited attention. A great deal of empirical work investigates that hunch (Sunstein, Reference Sunstein2021). The jury is still out, but in my view, the hunch has significant support, in the sense that under epistemically favorable conditions, consumers would be making more fuel-efficient choices (Sunstein, Reference Sunstein2021).

Sugden dips his toe, very lightly, into the empirical waters, investigating one of the many studies. Gillingham et al. (Reference Gillingham, Houde and Benthem2019) study the effect of an abrupt change in the fuel economy ratings of certain vehicle models, brought about by a correction of erroneous ratings. The question was whether the change would produce different choices. The authors find that consumers seem to value $1 in discounted fuel savings at some fraction of that, somewhere between $0.15 and $0.38 – which suggests that some kind of bounded rationality, perhaps present bias, is accounting for their choices. Sugden responds:

This sounds like an explainawaytion. It's a clever one. Is it right? As an objection to the careful work by Gillingham et al., it's pretty seat of the pants. We might want to test it. (Tacit knowledge exists, and it is important, but beware of modern-day phlogiston.)

Rizzo and Whitman (Reference Rizzo and Whitman2020) want to have a fight. They urge that I intend ‘to persuade the reader that behavioral economic policy has reached the stage where it either can or plausibly could overcome the problems of inadequate knowledge that the two of us have claimed it faces’. They lament that my article ‘only mentions us once in a footnote’. They conclude that I aim ‘to persuade the reader that knowledge problems pose no great barrier to behavioral policymaking. However, he has not yet addressed the knowledge problems we have presented in the past.’

Carly Simon, a student of human behavior, famously sang, ‘You're so vain/You probably think this song is about you.’ (A lot of men thought that Simon's song was about them!) My article is not about Rizzo and Whitman. (For the record, I like their book.) With respect to fuel economy standards, the evidence is not clear, which is why I was cautious. If I asked ‘the reader to imagine that the knowledge problems ‘have somehow been overcome’, it was because I was sketching what a defense of fuel economy standards, rooted in Hayekian behavioral economics, might look like, not asserting that those problems have, in fact, been overcome. (If you want a full-throated behavioral defense of such standards, invoking the idea of internalities, you would have to look elsewhere.) To the charge that my essay does not address the knowledge problems that Rizzo and Whitman have presented in the past, I plead nolo contendere. This was not the essay's purpose. (Rizzo and Whitman do not address the arguments about Star Wars that I have presented in the past, Sunstein (Reference Sunstein2016).)

To be sure, one of my main goals was to show that a Hayekian approach would examine not what planners would do, but what informed, unbiased choosers would do. Rizzo and Whitman are skeptical about that whole enterprise. They insist that there is no way of knowing what informed, unbiased choosers would do, and that in light of heterogeneity and the existence of local knowledge, the effort to find out is a kind of fool's errand. But (a request, a hope, a plea) might we avoid that level of dogmatism? Or that level of antecedent, uncomfortably-close-to-theological commitment to conclusions that precede engagement with actual problems?

Some cases are really easy; human beings are uninformed or biased, and choosers do err (Bhargava et al., Reference Bhargava, Loewenstein and Benartzi2017). GPS devices can help; they respond to a lack of information. Reminders can help; they respond to limited attention. Automatic enrollment can help; it can overcome inertia (Thaler and Sunstein, Reference Thaler and Sunstein2021). A lot of people, none of them fools, have been examining the underlying questions (Bhargava et al., Reference Bhargava, Loewenstein and Benartzi2017; Goldin, Reference Goldin2017; Allcott et al., Reference Allcott, Lockwood and Taubinsky2019; Bernheim & Gastell, Reference Bernheim and Gastell2021). It is uncharitable, and not a good idea, to dismiss their work with adjectives and nouns. Again: Tacit knowledge is definitely important, and Rizzo and Whitman are right to draw attention to it. But beware of explainawaytions, and let's post a sign: No Phlogiston Here. (Come on, man!)

Sugden urges that in speaking of Hayekian behavioral economics, I have a secret aim, which is ‘to head off the criticism that behavioural welfare economics is unacceptably paternalistic’. Actually, I was motivated by something mundane: curiosity. Hayek is a hero of mine (though I have substantial disagreements with him). It seems to me interesting, and perhaps more than that, to think about how Hayek, or a Hayekian, might respond to behavioral findings about individual error. In my view, the resulting approach – Hayekian behavioral economics – deserves respect. It is worthy of serious consideration, even if some of us, including the present author, would not entirely embrace it (Sunstein, Reference Sunstein2020b).