On Your Best Behavior: How to Nudge Ethically

By Alain Samson

This article discusses ethical considerations surrounding behavioral nudges by focusing on three core areas: Goals (of nudgers), Autonomy (of the nudged), and Effects (of nudges). The framework prompts practitioners to evaluate the alignment of nudgers’ intentions with the best interests of the nudged, the preservation of individual autonomy, and the evaluation of costs and benefits associated with behavioral interventions. It offers a set of practical questions for professionals seeking to build an ethically sound nudge practice, emphasizing the importance of a nuanced understanding of the backgrounds and needs of nudge targets, tailored interventions, and continuous monitoring.

Introduction

If you’ve ever come across online subscription options that highlight the most popular choice or a fruit basket placed next to the cash register in a cafeteria, you’ve encountered what behavioral scientists call a “nudge.”

Nudging is a practice that uses tweaks in the way choices are presented to people to steer behavior, such as changing decision defaults, asking people to precommit to an action, or making choices more salient. The use of nudges has arguably become the most prominent technique in applied behavioral science. Nudge projects have produced many success stories, but the practice has also raised important concerns about imposing specific ideas about what’s best for individual decision makers, influencing choices on a non-conscious level, and potentially producing unforeseen and unwanted outcomes.

These critical debates, reinforced by a lack of regulation and recent cases of academic misconduct, make it clear that nudge practice is in need of ethical guidelines. This is not just crucial for behavioral scientists. Whether you design a new banking app, implement Covid-19 policy, or sell men’s fashion, your work probably influences choices. Some of this may occur in the traditional way through economic incentives (e.g. providing discounts), information provision (e.g. communicating benefits), or mandates (e.g. prohibiting an action), but it may also involve nudges. This article briefly discusses the three main problem areas associated with nudging—Goals, Autonomy, and Effects—and offers practical questions that can help you build ethical nudge application.

Goals (of the Nudger)

To build ethical nudge practice we first need understand who does the nudging and, most importantly, why it occurs—the nudger’s intentions and goals.

Nudgers’ motives can be selfish (nudges serve the interest of the nudger), paternalistic (the nudger acts in the interest of the nudged), or pro-social (individuals are nudged to help others). In addition, nudges can benefit the nudger, the nudged, others (another individual or group of people), or the collective (i.e. society as a whole, including the nudged).

How you ethically evaluate these nudges depends on the perspective you adopt. Selfish nudges, for example, may ignore the nudged’s preferences. But does a selfish motive matter if the nudge doesn’t cause harm or is even mutually beneficial? In addition, some public opinion research suggests that nudges classified as benefiting society are significantly less accepted than those which benefit the nudged as an individual. Does this make pro-social nudges that don’t benefit the nudged less ethically legitimate?

When nudgers wish to act with the interests of the nudged in mind (most probably do), an important problem arises: How can they know what is in the nudged’s best interest? This question is at the very heart of ethical critiques of nudging, especially with respect to paternalistic nudges. Nudgers may project or impose their own values and goals onto others. At best, they assume that the targets of a nudge all have the same or context-independent preferences. Consider interventions that nudge people to take the stairs instead of the elevator. Some people may prefer to get their exercise elsewhere, are physically handicapped, or simply care more about convenience than health.

Addressing Concerns

In defense of nudging, it has been argued that having to determine what’s right for targets of a policy isn’t exclusive to nudging and that individuals most susceptible to nudges tend to be those with high uncertainty in the first place. Many practitioners have also adopted an utilitarian view of costs and benefits that focuses on bringing about the greatest good for the greatest number. (Covid-19 behaviors are a dramatic example.)

Richard Thaler, the father of nudge, maintains that practitioners should “nudge for good”. There should also be “good reason to believe that the behavior being encouraged will improve the welfare of those being nudged.” It’s good practice to take a deep dive into data you may already have about your nudge targets or conduct new research to better understand their preferences. This is not always an easy task, though, since preferences can be uncertain, context-dependent, and variable over time.

When preferences are diverse, choice architects (especially those in the digital and/or the commercial world) can attempt to target different consumers better with programs and interventions that fit their preferences. To better target individuals, a bit of data science can help. If you are unable to work with fancy algorithms (like Netflix or Amazon’s recommendation engines), traditional segmentation research should do the trick. Nudges to increase savings rates, for example, may not fit the preferences of some low-income segments with liquidity constraints, who may be better served with other programs, such as helping them avoid high-interest loans.

Autonomy (of the Nudged)

Ethical discussions often point out that nudges may exploit people’s irrationality, are excessively convenient, or even manipulate folks on an unconscious level.

In order to differentiate degrees of conscious thought (see examples in table below), nudges have been classified as either Type 2 (affecting reflective choice or triggering an active response) or Type 1 (the targeted outcome is more automatic or the nudged’s reaction more passive). The descriptive norm in the message “80% of our guests reuse their towels” is a Type 2 nudge, whereas room service not changing the towels unless requested by the guest is a Type 1 nudge.

It has also been argued that nudges sometimes lack transparency. In a transparent nudge, the intention behind it and the way it operates is more apparent to the person being nudged. While we can expect most people to hear and notice ambient music, its influence is more transparent if it’s played at an airport to lighten travelers’ moods than at a wine store to make shoppers not just feel good, but buy more premium wines.

Addressing Concerns

Thaler and Sunstein have acknowledged that some nudges may operate outside of individuals’ awareness and suggested that it should be easy to opt out of a nudge. Preserving this freedom is a central part of ethical nudge practice. For example, if you offer a membership that renews automatically, you need to make sure customers can switch off renewal with the click of a mouse button rather than having to find it buried deep in the structure of your website.

It’s important to bear in mind that nudges are just one of many tools in the toolbox of those seeking to achieve corporate or policy goals. In order to address autonomy concerns, it’s important for practitioners to consider the full range of interventions (traditional vs non-traditional, Type 1 vs 2, etc.) at their disposal. If you’d like to decrease food waste at your buffet, for example, make sure you also test pricing options (e.g. prices for different types of eaters) and rational persuasion (e.g. “Help us reduce food waste…”) alongside your relatively non-transparent nudge (e.g. introducing smaller plates). This will not only increase your chances of finding the most effective approach, but also give different approaches a fair shake.

Should you go as far as disclosing your nudge to increase transparency? Some argue that disclosing nudges is impractical and transparency ‘in principle’ might suffice. While it’s commonly assumed that nudges work best “in the dark”, some studies on defaults have found that disclosing the default (e.g. “Please consider that the preselected default value might have an influence on your decision”) does not significantly reduce the effectiveness of the nudge.

Research on the effect of disclosures across different nudges is still limited and you may want to conduct your own tests. Disclosure would not only have beneficial effects on ethics, it may also elicit positive responses from the nudged.

Effects (of the Nudges)

There is a growing number of case studies and data about the effectiveness of nudge interventions (see this 2019 meta-analysis, this 2020 research and this study from 2022, for example). Critics point to a lack of evidence to-date showing a link between nudges and genuine behavior change with long-term effects. If a nudge succeeds at making me floss my teeth for a week, has it also successfully changed my preferences with regards to oral health?

There are various reasons why nudges may lead to unanticipated consequences as well, such as reactance (responding negatively to influence attempts) and other backfire or boomerang effects. For example, research on the effect of descriptive norms on energy use found increased consumption for those households whose past energy consumption was below the norm. Side effects may also occur. A nudge may reduce water consumption, for example, but inadvertently also increase electricity consumption.

Addressing Concerns

While backfire effects on the nudged population overall may become apparent with testing, differences across subgroups and side effects on non-targeted behaviors are less easy to spot. It’s good practice for nudgers to anticipate potential issues early by gaining an understanding of nudge targets’ backgrounds (e.g. in terms of their demographics, such as income, race, or gender), unique circumstances (e.g. life stage or employment status), and associated needs. Low-income individuals, mentioned previously, are a one of many examples of a group that may respond differently to nudges.

You can make an ethical assessment of your program by carefully weighing the benefits and costs of behavioral interventions, as the ones illustrated by the following cases:

In order to weigh costs against benefits, they need to be quantified. This is relatively easy for behaviors like energy use. Intangible costs, however, such as bad feelings about a diet nudge, are more difficult to assess. In addition, effects may carry costs for different stakeholders. Reminder emails, for example, can have both costs associated with the loss of subscribers to the organization and ‘annoyance costs’ for targeted individuals.

If you are unfamiliar with formal economic approaches, you may want to measure costs and benefits with some help from an economist, which would also allow you to account for so-called “welfare effects”. Economists typically measure these effects of nudges using willingness-to-pay. Analyses may identify costs that are easily overlooked, such as having to give up a pleasant activity. Direct effects of the nudge may also lead to costs, such as the “emotional tax” some experience as a result of calorie-labeling. If the average individual values a nudge’s benefits at $10 and its drawbacks at $11, you better rethink your approach.

Once an intervention is live, practitioners should continuously monitor processes and outcomes, which in turn allows them to spot unwanted effects. If your nudge is designed to increase sales of a product, for example, make sure you also monitor sales of other products, as well as customer feedback received through regular CRM channels.

Conclusion

The application of nudges has become a popular practice to solve problems in both the public and the private sector. Many professionals and organizations are looking to take their work to the next level by building an ethically-literate and future-oriented nudge practice.

I recommend for practitioners to start any project that involves nudges by first defining ethical guidelines. The 3×3 questions provided in this article can serve as a nudge ethics audit or checklist. In the beginning of the project, have each team member reflect on the questions, which can then be discussed as a team. The three problem areas are also relevant at different stages of a typical nudge project:

1. Defining the problem and possible solutions (GOALS of the nudger),
2. developing possible interventions or approaches (AUTONOMY of the nudged), and finally
3. testing and implementing the interventions (EFFECTS of the nudges).

Each group of questions should be revisited as the project progresses through corresponding stages and new knowledge emerges. Once the project has been wrapped up, you can use the recorded responses to questions as a guide for future projects.

Just as nudges may not provide one-size-fits-all solutions, their ethical use may need to be refined across projects and over time. In this article, I hope to have provided a starting point for practitioners to think about and implement ethical nudges.