Resisting Nudges

By Gabriel De Marco

Consider the following case:

Classic Food Placement (FP): In order to encourage healthy eating, cafeteria staff place healthy food options at eye-level, whereas unhealthy options are placed lower down. Diners are more likely to pick healthy foods and less likely to pick unhealthy foods than they would have been otherwise.

This intervention is a paradigmatic case of what are often called nudges. Though many will think that it is OK to implement this sort of intervention for these sorts of purposes, there is a large debate about when exactly this is OK.

One common theme is that whether such an influence is easy to resist is going to be relevant to when the intervention is OK. If the intervention is not easy to resist, then, at the very least, this counts as a strike against implementing it. However, though there is often reference to the resistibility of a nudge, there is rarely explicit discussion of what it is for a nudge to be easy to resist, or for it to be easily resistible.

To begin giving an account of what it is for a nudge to be (easily) resistible, we need to figure out what it is an ability to do. So, what is it to resist a nudge?

Though authors tend not to explicitly answer this question, one natural, and perhaps common, answer is that to resist a nudge is to behave contrary to it. What sort of behavior is contrary to the nudge? An initially attractive thought is that, since nudges are intended to get agents to do (or not do) something, behaving contrary to a nudge just involves behaving contrary to what the nudge is intended to get the nudgee to do. Call this the intention conception of resistance. In Classic FP, the nudge is intended to get agents to pick the healthy food – suppose it is salad. The intention conception would tell us that resisting would involve not picking the salad. Though people tend not to be explicit about how they are understanding the claim that someone resists a nudge, the intention conception seems to capture the underlying theory.

There are, I will argue, issues with the intention conception. First, this way of conceiving of resistance is not helpful in cases where there is no intention behind the influence. Consider:

Random FP: The cafeteria manager is not aware of the placement effect on customer behavior. But she still needs to decide where to place the salad and the pudding; so, she flips a coin. It lands heads, the salad is placed at eye-level, and the cafeteria layout is identical to that found in Classic FP.

Here we have, at the very least, a nudge-like influence. The relevant feature of the environment is the same as in Classic FP, and if it has an effect in Classic FP, it presumably has the same effect in Random FP. And it is plausible that resisting the influence involves the same thing in both cases. The intention conception, however, would not tell us what resisting the influence involves in Random FP, insofar as there is no intention behind the nudge-like influence.

Second, although it is typically assumed that the effect of the nudge and the intention behind it are aligned, this need not always be the case. Consider:

Confused FP: The cafeteria manager receives a memo concerning the effect that food placement can have on customers. However, he is confused about the effect – perhaps there was a typo, or he misread it – such that he thinks that people are less likely than they would have been otherwise to pick the food that is at eye-level. Since he intends to get customers to purchase more chocolate pudding, he places the salad at eye-level, and the cafeteria layout is identical to that found in Classic FP.

As with Random FP, the relevant feature of the environment is the same as in Classic FP, and presumably has the same effect on customers (assuming it has one). On at least one way of understanding this, one might think that resisting the nudge, or the influence, involves the same thing across all three cases. But the intention behind the food placement in Confused FP is to get people to pick the pudding. On the intention conception, resisting this nudge involves not picking the pudding, which is something quite different than resisting the nudge in Classic FP. If resisting the nudge, or the influence, in all three FP cases involves the same behavior – if resisting the influence in all three cases involves not picking the salad – then the intention conception gets it wrong.

However, it seems at least plausible that there is some sense in which resisting the cafeteria manager’s attempt at influencing customers does involve different behavior in Classic FP and Confused FP. These managers tried to influence customers into doing different things, and whereas the former used a method that may help him achieve his goal, the latter did not.

So, I suggest that when we talk about resisting a nudge, we might be talking about two different things at once. One thing we could be talking about is resisting the influence itself. Doing this would involve the same behavior in all three variations of Classic FP. Another thing we might be talking about is something like resisting the would-be influencer’s attempt to influence, and this might involve different behavior in Classic FP and Confused FP; and, since there is no attempt in Random FP, there is no sense in which one can resist the attempt in this case. This difference is obscured when we make the common assumption that the effects of the nudge and the intention behind it are working in unison, but Confused FP pries these apart.

For the purposes of this post, we can just say that resisting the attempt at an influence involves acting contrary to the intention behind it; the intention conception is correct when it comes to resisting the attempt. But what does it take to resist the influence itself? Finding the answer to this is complicated, and I won’t resolve this in this post.

One might think that something like the intention conception will still serve us well enough. The issue with the intention conception, perhaps, was not so much the focus on intention, but rather the focus on the actual intention. Perhaps we can rescue a version of the intention conception if we focus on what intention the would-be influencer would have if they were informed. Call this the informed-intention conception. This may have more promise with respect to Random and Confused FP; the fact that there is no intention behind the food placement in Random FP does not preclude there from being an intention that the manager would have, were he informed of the effects. Nor does the fact that the manager in Confused FP is mistaken about the effect, and for similar reasons. What matters, on this conception, is what intention the cafeteria managers would have, were they informed of the effect of the food placement.

This apparent benefit, however, may be short-lived, and this becomes apparent once we try to work out what it would actually say about these, or similar, cases. What intention would the manager in Confused FP have, were he informed of the food placement effect? One might worry that in this hypothetical case, he would have the same intention – to get people to buy the chocolate pudding – it is the intervention that would change: he would have placed the pudding at eye-level instead. But this doesn’t seem to answer what it would take to resist the influence in Confused FP, which presumably is the same as in Classic FP.

Perhaps, instead, we could focus on the intention the manager would have, were he to a) be informed about the effect and b) implement the same influence. But this version faces some issues as well, insofar as there may not be a clear answer to what the manager would intend, were he informed. This can be made clearest, I think, by focusing on cases in which we stipulate the actual intention, or lack thereof. Consider, for example:

Apathetic FP: Everything is as in Random FP, but the cafeteria manager knows about the food placement effect, yet does not care at all what his customers pick. He flips the coin to decide how to place the items.

This case, again, features the same environmental feature, and so the same influence itself. Yet the manager is informed about the effect, and has no intention to modify behavior. Given this, it is not clear that there is any relevant intention that the manager would have, were he to be informed of the effect. If there is no such intention, this account fails to give an answer for what resisting the influence itself would involve in this case.

We could further modify the view by focusing on what intention the manager would have were he to, a) be informed about the effect, b) implement the same influence, and c) intend to influence customers. This could help to get around Apathetic FP insofar as we would now only be concerned with hypothetical cases in which he is not apathetic, and does have an intention.

But even this formulation faces a further issue. Nudgers who are informed of the effect of a nudge-like influence, and who intend for the intervention to have an effect on individuals, may still differ on the content of that intention. Consider another nudge that is often mentioned in the literature:

Calorie Count (CC): The food menu in a restaurant displays the calorie-count of individual food items.

One effect this has is that many customers form beliefs about the number of calories in a particular option. Another effect it might have is that individuals, when deliberating about what to eat, take calories into account, whereas they may not have otherwise. A further effect this might have is that people, in general, pick items with lower calorie-counts. Given these different effects, one might implement this nudge for different reasons, and with different intentions, even if one is informed of all of the above. Thus, consider the following two cases:

CC-Autonomy: As in CC, but the restaurant owner intends for people to make a more informed decision about what to eat.

CC-Paternalistic: As in CC, but the restaurant owner intends for people to pick healthier, lower-calorie, foods.

In both cases, the restaurant owners are aware of the same facts. However, the intentions are different; one intends for customers to make a more informed decisions – regardless of what they decide to do – and the other intends for customers to pick lower-calorie meals. Yet the feature of the environment – the display of the calorie counts – is the same, and we can suppose that they would have the same effect on individuals. The last version of the informed-intention conception would therefore tell us that resisting the influence itself involves different behavior in these two cases; yet, whatever it is that resisting the CC influence involves, it would plausibly be the same across these two cases.

There is much more to say here, but ultimately, I suspect that appealing to the intentions of a would-be nudger – be they actual or hypothetical – will not help to give us an account of what it is to resist the influence itself.

 

Disclaimer:

Recently, there has been some pushback against the efficacy of nudges, sparked by PNAS’s publication of three letters responding to a recent meta-analysis of nudges (the authors reply here). These letters pointed out various issues with not only the meta-analysis, but the whole body of work surrounding interventions often called nudges, and possibly suggesting that, at the very least, we have no evidence for the effectiveness of nudges. However, others take a more moderate line, often pointing to another large study that seems to avoid at least some of these issues (and here is a brief interview with the authors of this large study). And even one of the critics of that study agrees that nudges sometimes work (see here for a response from the authors of that study).