
Eugen Dimant (University of Pennsylvania & CESifo, Munich), and Shaul Shalvi (University of Amsterdam)
We are constantly surrounded by temptations that are not in our best interest. That is, they are often neither in the best interest for us individually (e.g., succumbing to a sugar intake that is beyond healthy) nor collectively (e.g., throwing trash on the ground instead of a bin). How can we make better decisions? Nudging, a flourishing line of literature initially kickstarted by Thaler and Sunstein in their 2008 book ‘Nudge’, is concerned with exactly this challenge.
A plethora of interdisciplinary literature has established that nudging people into better behavior works across various contexts. At the same time, the effectiveness of nudging has been shown to produce small effect sizes that are highly sensitive to the exact context (Hummel & Maedche, 2019; Beshears & Kosowsky, 2020; Mertens et al., 2021). However, the theory of nudging has little to say about the longevity of the positive effects that result from nudging. What happens once the nudge is removed? Will individuals have ‘learned’ what’s better for them and others, or will they revert to their original behavior? That is, not only nudging but also changing such behavior for good, however, is the real challenge. This is particularly true when choice architects attempt to change behavior that is collectively harmful but individually beneficial. Evidence on long-lasting positive spillover effects of nudging once the nudge is removed is scarce but a recent paper by Brandon et al. (2022) at the very least casts reasonable doubt on the ‘one-and-done’ idea. Thus, when we observe that the longevity of nudges is limited then one can plausibly assume that the underlying preferences may not have actually changed, at least not sufficiently so.

Before diving into potential solutions to this, it is worth noting that only few serious scholars would argue in favor of this ‘one-and-done’ nudging approach. Instead, they would reasonably argue that leaving the nudge in place (say, constantly nudging employees toward a successful retirement) is the most effective way to achieve behavior change. The reality is, however, that there are many instances where constant nudging – or nudging at all – is not feasible.

Take the famous study by Goldstein et al. (2008) showing that simple social comparison messaging encourages hotel guests to reuse their towels. This worked well in the U.S. where such behavior was not yet the existing norm and thus allowed hotel guests to learn what’s appropriate behavior from their peers. But what about instances where such a norm is already in place? As Bohner & Schlüter (2014) show, the same interventions were ineffective in Germany where towel reuse already exists at high rates. Or consider instances where social comparisons with peers are used to promote household energy conservation. While results by, e.g., Schultz et al. (2007) indicate that the positive effects on the initial under-performers are substantial, the existence of boomerang effects is just as real: initial top performers – those who already consume at low rates – get discouraged and adjust their behavior downwards. But this does not mean that additional savings, even for top performers, are not possible. Thus, should we simply leave those top performers alone? But what if there is still ample room even for those top performers to improve? Neglecting these growth opportunities prevents nudging to play to its full potential.
Thus, finding ways to supercharge nudging interventions can benefit us all. Here, we will argue that the ‘meta-nudging’ approach represents such a promising approach, which we formally introduce in our recent publication in Current Opinion in Psychology (Dimant & Shalvi, 2022). In short, meta-nudging implies that instead of nudging end-users directly, one would nudge them indirectly via “social influencers” who are in the position of enforcing social norms. This approach is in the spirit of the recent call by scholars urging a reconsideration of the classical nudging approach by putting more emphasis on the environment in which these individuals operate. In turn, this could help behavioral science to transition from amending choice architecture to creating choice infrastructure (Hallsworth & Kirkman, 2020; Chater & Loewenstein, 2022).
Our underlying argument relies on the argument that observing behavior change does not necessarily tell the choice architect much about the underlying mechanism. Does the nudgee simply comply with the nudge for one of the many reasons why they work so well (e.g., inertia)? Or did the nudgee actually learn and internalize what the better behavior is? It becomes obvious that these two instances produce very different expectations of what would happen if the nudge would be removed: arguably, the nudgee would revert to the original behavior more frequently in the former than in the latter case.
This is where the power of social norms come into play. While nudges are largely about individual change, social norms are about collective change (Bicchieri, 2006). Such a distinction is important if we want to move from choice architecture to choice infrastructure and develop interventions that have bite and are in the best scenario also long-lasting. Research by Dimant & Gesche (2021) suggests that ‘norm-nudging’ can be a potent application of the meta-nudging approach. Norm-nudging, which is a special case of behavioral nudging, aims at eliciting and changing existing social norms through systematic variation of social expectations. This approach has been theoretically conceptualized by Bicchieri & Dimant (2019) in that norm-nudge interventions aim at changing either the beliefs about what others in one’s reference network do (descriptive element of the norm, first-order belief) or what others in one’s reference network approve others to do (injunctive element of the norm, second-order belief). Consequently, the effectiveness of norm-nudging results from targeting (at least) one of three aspects: (i) pointing out bad norms that are currently in place, (ii) defining good norms more clearly, and (iii) facilitating the enforcement of good norms.

As we argue in our new paper, meta-nudging is particularly promising in supercharging classical nudging approaches because it taps into existing social constructs. That is, by targeting those who enforce behavior – rather than those whose behavior one wants to alter – behavioral interventions would aim at nudging individuals in positions of power who have the ability to enforce the transgressors’ adherence to social norms. One advantage of meta-nudging is that behavioral interventions that rely on delegated policing (“hired guns”) might both be perceived less intrusive and more successful in that they would capitalize on existing peer mechanisms. Arguably, this would increase the acceptability of enforcement, which has been shown to be a crucial ingredient of successful norm enforcement (Bicchieri et al., 2021). While the meta-nudge still needs to be potent enough to overcome the influencer’s inertia and other related individual costs such as the fear of potential retaliation from the subordinate nudgee, there are now also counteracting forces that facilitate the success of the meta-nudge. For example, any utility that the influencer derives from impacting others’ behavior or from enforcing norms, which motivates the influencer to positively react to the nudge. Indeed, supporting those assumptions, influencers were found to be ‘social trendsetters’ who are ready to bear a cost to initiate change because they are usually less sensitive to risk (Bicchieri, 2006). This can be fruitful in the context of social movements, such as #meToo, where social change can be facilitated by social influencers. Here, using meta-nudging to ignite the ‘change cascade’ can be particularly promising (Sunstein, 2019).
To provide some conceptual backing, the last part of our paper discusses these ideas in the context of nudging honesty. Changing behavior in the context of curbing dishonesty is challenging because of diverging incentives: dishonesty is often individually beneficial but collectively harmful. Thus, any behavioral intervention aimed at changing behavior directly needs to convince the individual to forego an individual benefit in favor of the collective good. Recent work further demonstrated the large impact one’s (dis)honesty has on others (dis)honesty. Specifically, when considering settings in which one finds justification for one’s own dishonesty in the dishonesty of peers (Weisel & Shalvi, 2015; Leib et al., 2022), people lie a lot. Here, social reinforcement via observing and being observed by one’s peers may be interpreted as a signal of the dominant social norm, which can accelerate the contagion of dishonesty (Dimant, 2019; Biccheri et al., 2022). From this, we can construct forms of meta- nudging that have the potential to reduce dishonesty. For example, since the level of dishonesty has been found to be sensitive to financial incentives, nudging influencers to enforce the deviance of others via costly punishment – as successfully tested by Dimant & Gesche (2021) – presents a promising avenue.
In conclusion, we see this approach as complementary to the classical nudging approach allowing the choice architect to select from a wider array of tools that policy-maker have at their disposal. The ‘best’ tool will be context dependent, will require testing and re-testing, and a careful roll-out when attempting to achieve success at scale. By complementing the arsenal of behavioral change techniques that target individual decision-making with the ‘meta-nudging’ approach proposed here, policy-makers can build momentum at the collective level. The long-term success of such an approach remains an empirical question. We encourage the behavioral science community to meet this challenge head on!