Nudge: The Gentle Prison

Libertarian paternalism is only the latter.

Posted Mar 17, 2018

If you choose (god), you have been chosen. — Calvinist minister to an astonished Catholic at the Brown University Faculty Club.

Nudging is a popular concept: both among nudgers and nudgees. Nudgers get to indulge their paternalistic impulse (I have power over you!) while preserving a moral self-image (I am doing this for you!). Nudgees get to delight in an easy-to-navigate “choice architecture,” which purports to do the hard work of decision-making for them while allowing them to hold on to the soothing illusion of free will. Nudgers assert that nudgees are predictably irrational, but that benign authority can help them out, while at the same time letting them believe that they are helping themselves. In a brilliant and diplomatically devastating review of Thaler and Sunstein’s iconic book, Thomas Leonard (2008) sliced through the sanctimonious veil. His analysis culminates in the conclusion that

Behavioral economics, having attacked Homo Economicus as an empirically false description of human choice, now proposes, in the name of paternalism, to enshrine the very same fellow as the image of what people should want to be. Or, more precisely, what paternalists want people to be. For the consequence of dividing the self has been to undermine the very idea of true preferences. If true preferences don’t exist, the libertarian paternalist cannot help people get what they truly want. He can only make like an old fashioned paternalist, and give people what they should want” (p. 359).

If nudge theory is intellectually bankrupt and morally suspect, perhaps it can be defended on pragmatic grounds. Perhaps there is a case for the claim that people should vaccinate their children and save more for retirement. These actions are good for them individually, and they are good for society. So why not nudge them if we cannot force them? Ironically, if paternalisim in some shape or other is inevitable, as Thaler and Sunstein claim, might it not be more honest to make it coercive? The illusion of free choice, the perception of not being manipulated, chafes against a basic conception of human dignity.

Nudgers seem to think that they can have it both ways: manipulate (er, nudge) people, while granting them genuine options for dissent. This is another topic about which the irrepressible Mr. Sunstein has written a bestselling book. But if dissent is as essential to a well functioning society as he claims, then why not encourage dissent from the nudge offensive? Do the nudgees have the free choice that nudgers and the nudgees themselves think they have (Krueger, 2018)? A nudge is a subtle manipulation in the choice architecture making a desired choice more likely. A man is more likely to aim at the optimal spot in the urinal if a fly is painted on the porcelain. No harm done. Although the man has the freedom to aim at a different spot at the cost of greater splatter, there is no real sense of his dignity being infringed upon. Now consider the man who must choose between spending more now and saving more for later. Here we have a potential intra-psychic conflict between the desirous ego of the present and the well-being of the future ego. Intertemporal choice wipes away the idea of unitary preferences. The person of the present is weak, and the nudger comes to the rescue of the future self, a hero both by the lights of rationality and morality (Krueger, 2010). 

Suppose that without a nudge (an opt-out savings plan offered by the employer), 20 percent of people choose to save adequately, whereas with such a plan, the number rises to 80 percent. A tremendous victory for the nudgers, society, and the nudgee’s future self! The present-day nudgee is pleased as well because he thinks he could have opted out of the savings plan just as easily as inertly doing nothing. But could he? As an experimental intervention, or ‘treatment,’ a nudge is a causal force. The nudge causes 60 percent of people to do what they otherwise would not do (assuming the nudge does not cause anyone to shift in the opposite direction). What a cause does to an effect is deterministic even if its efficiency is not 100 percent. If we sample Sam from our group of nudgees, we know that he is now a saver with a probability of .8 as opposed to a probability of .2. In this sense, the causal force is probabilistic; it reveals its power at the group level and our ignorance as to which person is caused to save and which person is not. Once we know more about the individuals, we might learn why Sam was nudged and Samantha was not.

Are the respectively high and low probabilities of saving in the nudge and no-nudge condition compatible with both determinism (i.e., the nudge’s causal force) and free choice? In other words, are paternalism and libertarianism as compatible as the nudgers claim? Suppose 100 individuals are nudged and 80 act as hoped (i.e., doing nothing if the nudge is a default), compared with 20 out of a 100 selecting the option favored by pater when there is no nudge. Now we review the 100 individuals in the nudge condition one by one. As soon as we have logged the ‘choices’ of 99 of them, the ‘choice’ of the last one is fixed. There are only n-1 degrees of freedom. Now we set this person aside, scramble the remaining 99, and review them again. Once we have logged the choices of 98, the 99th is fixed. And so on down to the last person; no one’s ‘choice is free. The argument extends to a single person making 100 ‘choices.’ Knowing, for example, that the person has a probability of .8 ‘to do the right thing’ (i.e., nothing), we infer that he will do nothing in 80 out of 100 rounds. Reviewing his choices one by one, we find that no a single choice is free.

One might think that this argument is sophistry, that choices are free and indeterminate when we look forward, and that they show the constraints we have seen with backward induction only when we review past behavior. A free-choice libertarian might agree that behavior is determined when we look back on it, but that it is free when we look forward. The question is still whether we can have the paternalistic cake of predictability (p = .8) and eat it as libertarians, claiming that individual choices are free. The answer is no. Let’s say each individual out of 100 does nothing (as hoped) with a probability of .8. This does not mean that the resulting proportion of doing nothing is exactly 80 percent. This value is just the expected (most likely) value. There is some variation around it, and more so if the sample is small. There is thus wiggle room in the sense that we can sample an individual and grant him the option of changing his mind. If he simply resamples his mind and does nothing, nothing much changes. His behavior, and that of the group is still characterized by necessity (the causal power of the nudge) and chance (the fact that the causal power is imperfect). If, however, we grant him libertarian freedom so that he may change his mind with any probability deemed necessary to be considered free, then still nothing much changes if this privilege is given only to this one person. If we give this privilege to a second person out of a large group, the effect is still small. At some point, however, the collective impact of free choice will pull the group proportion of do-nothing toward unpredictability, that is, toward .5, and this will be so in the no-nudge condition too (why should the folks in this condition not be free?). At that point the causal force of the nudge is eliminated. Libertarian free choice crowds out the paternalistic preference. You can’t have it both ways. When the desired paternalistic outcome is shown as a strong nudge effect, we can  infer that the nudgees are not not free. As Leonard (2008) suggested, libertarian paternalism is just paternalism; but it is more devious than old-fashioned coercive paternalism because it keeps people enslaved in the illusion of individual freedom. 

Statistical intuition. When arguments offend belief or intuition, judgment suffers. Here’s an example: Quoting from this essay, a commentator writes

“As soon as we have logged the ‘choices’ of 99 of them, the ‘choice’ of the last one is fixed." This is ridiculous. Statistics are not determinative. If I see an ad on TV that says two out of three doctors recommend nargles and I ask two doctors and they recommend nargles, does that mean the next doctor I ask is forced to say nargles are bad for you, no matter what she thinks? Of course not.

The omission of the preceding sentence makes all the difference. That sentence is “Now we review the 100 individuals in the nudge condition one by one.“ With that, we are assuming a fixed number of individuals, and the degrees-of-freedom argument holds. To make the ‘nargles’ story a valid analog, one would have to assume that there are only 3 doctors and 2 have rendered their opinion. Then the opinion of the third is fixed. What the commentator appears to be claiming is that if we have a known proportion (e.g., 2/3 as in his example), any individually reviewed case is perfectly free to vary without affecting the proportion on which the argument is premised. The point of this essay was to show the constraints that do apply.

And hoopla, I find myself making a libertarian argument (see here for my unexpected bedfellows). Who would have thunk it?

Krueger, J. I. (210). Discounting and the ethic of denial. Psychology Today Online,

Krueger, J. I. (2018). Five arguments for free will. None of them are compelling. Psychology today Online.

Leonard, T. C. (2008). Richard H. Thaler, Cass R. Sunstein, Nudge: Improving decisions about health, wealth, and happiness. Constitutional Political Economy, 19, 356-360. DOI 10.1007/s10602-008-9056-2 [Book Review]