The Gentle Push: Why a Pre-Ticked Box Achieves More Than Any Argument — and Whether It Is Permissible to Nudge People Toward Their Own Good

The Hook: Two Neighboring Countries, One Form

Germany and Austria share a language, a long common history, a similar healthcare system, and, if you ask their citizens, nearly identical attitudes toward organ donation. And yet, at the start of this century, a gap yawned between the two countries that was hard to believe: in Germany, roughly 12 percent of the population counted as organ donors. In Austria, it was 99.98 percent.

The difference lay not in morality, not in religion, and not in public education. It lay in a form. In Germany, you had to actively consent to become a donor; if you did nothing, you were not one. In Austria, everyone counted as a donor who did not expressly object; if you did nothing, you were one. The same inertia that made nearly everyone in Germany a non-donor made nearly everyone in Austria a donor. In 2003, the psychologists Eric Johnson and Daniel Goldstein documented this pattern across Europe in the journal Science: in countries with opt-in rules, the effective donor rate averaged around 15 percent; in countries with opt-out rules, around 98 percent. Denmark and Sweden, culturally as close as any two countries can be, stood at 4.25 percent and 85.9 percent.

This finding is the big bang of an idea that has occupied governments, corporations, app designers, and philosophers for nearly two decades: if something as trivial as the default setting of a form helps decide over life and death, then the design of decision situations is never neutral. Someone determines what happens when you do nothing. Someone determines which option is listed first, which one is pre-ticked, how many clicks an objection costs. That person is a choice architect, whether they want to be or not. And anyone who has grasped this faces a moral question that can no longer be shaken off: if I determine the architecture anyway, may I build it so that people are more likely to do what benefits them? Or is that paternalism through the back door, more dangerous precisely because it is so gentle?

This article takes you through the complete story of nudging: from the idea of libertarian paternalism through the most spectacular successes of defaults, the philosophical charge of manipulation, the defense via welfare, autonomy, and dignity, the sobering effectiveness debate of recent years, all the way to the dark twins of the nudge, the dark patterns, and the response of European law to them. At the end, as so often in ethics, there is no comfortable answer, but there is a checklist you can apply tomorrow to your next form.

Part 1: The Core Concept — Libertarian Paternalism

A Contradiction as a Program

In 2003, the economist Richard Thaler and the legal scholar Cass Sunstein published a short paper in the American Economic Review with a provocative title: Libertarian Paternalism. The term was deliberately an oxymoron. Libertarian means: no one is to be coerced; every option remains available; exiting the suggestion must be easy. Paternalistic means: the designer may and should arrange the situation so that people are more likely to choose what is good for them, by their own judgment. Five years later, the two turned the paper into the book Nudge: Improving Decisions About Health, Wealth, and Happiness (Yale University Press, 2008), which became one of the most influential non-fiction books of the century and helped carry Thaler to the Nobel Prize in Economics in 2017.

The definition the book gives is precise and worth knowing by heart. A nudge is any aspect of choice architecture that alters people's behavior in a predictable way without forbidding any options and without significantly changing their economic incentives. Three elements are decisive here. First: a nudge forbids nothing. A smoking ban is not a nudge; a warning on the pack is. Second: a nudge does not change prices. A sugar tax is not a nudge; fruit at eye level on the cafeteria shelf is. Third, and this is the core: a nudge must be cheap to avoid. Thaler and Sunstein offer the rule of thumb that the cost of avoiding it should be so low that a fully rational person would practically not notice it.

Econs and Humans

Why do nudges work at all? Because we are not Econs, as Thaler mockingly calls the ideal figure of classical economics: omniscient, strong-willed, consistent utility maximizers. We are Humans. We have limited attention, weak self-control, a preference for now over later, a tendency to feel losses about twice as strongly as gains, and the habit, when faced with complicated decisions, of simply taking whatever is already there. The entire body of psychological research that found its most famous expression in prospect theory (see Losses Loom Larger: How Prospect Theory Dethroned Rational Man – and Why Its Most Famous Principle Is Now on Trial Itself) is the foundation of nudging: if people deviate from rationality systematically and predictably, then those deviations can either be exploited or compensated for.

To an Econ, the default setting of a form would be completely irrelevant. He would make the same choice in both cases, because the order, the wording, and the checkbox do not concern him. To a Human, the checkbox is half the decision. Precisely for this reason, so goes the central argument of Thaler and Sunstein, libertarian paternalism is not an intrusion into an otherwise free choice, but the deliberate design of an influence that takes place anyway.

There Is No Neutral Architecture

The strongest argument of the two is also the simplest. Imagine the manager of a school cafeteria who has to decide in what order the dishes are arranged. She can put the fruit up front or the fries. She can draw the order by lot. She can order them by the suppliers' profit margins. She can arrange them so that the children eat as healthily as possible. What she cannot do is: choose no order. Something is always up front.

The same applies to every form, every piece of software, every piece of legislation, every company pension scheme. Whoever builds a sign-up form for retirement savings must decide what happens if the new employee does not fill it in. Whoever ships an operating system must decide whether telemetry is on or off by default. Whoever regulates organ donation must decide what silence means. The question is therefore never whether people are influenced, but only by whom, in which direction, and with what accountability. That is the ground on which the ethical debate must be conducted, and anyone who overlooks it is arguing against a phantom.

Part 2: The Choice Architect's Toolbox

Nudges come in many forms, and the differences matter for the ethical assessment. The following table organizes the most important techniques according to the scheme used by the meta-analysis of Mertens and colleagues (2022): interventions in the structure of the decision, in the information about the options, and in the assistance of one's own intention.

The table already shows why the ethical discussion cannot be conducted about the nudge. A reminder SMS that reminds me of an appointment I made myself is something completely different from a loss frame that exploits my fear. Both are nudges in the sense of the definition.

Part 3: The Great Successes — When the Default Wins

The Pension Plan That Enrolls Itself

The first and still most convincing example of the power of the default comes from retirement savings. In 2001, Brigitte Madrian and Dennis Shea, in the Quarterly Journal of Economics, examined a large U.S. company that switched its company pension plan, the so-called 401(k) plan, from an opt-in to an opt-out rule. Before, new employees had to actively enroll; afterwards, they were automatically enrolled and could opt out. The participation rate among new hires jumped from around 37 to around 86 percent. Not a single incentive had been changed, no employer match increased, no obligation introduced. The title of the paper said it all: The Power of Suggestion.

But Madrian and Shea also discovered the flip side. The automatically enrolled employees stayed in large numbers at the low default contribution rate of three percent and in the conservative money market fund that served as the default, even when, under the old rule, they would have chosen higher contributions and higher-yielding investments. The default worked in both directions: it brought people into saving, and it held them fast in an unwise setting. A default is never just a push; it is an anchor, and a badly placed anchor harms just as reliably as a well-placed one helps.

Save More Tomorrow: Committing the Future

Thaler and the behavioral economist Shlomo Benartzi built on this insight a program that may be the most elegant nudge of all. Save More Tomorrow (SMarT), documented in 2004 in the Journal of Political Economy, resolves a dilemma: people want to save more, but they do not want less money in their pocket right now, and they feel every cut in their net pay as a loss. The solution: employees commit today to raising their savings rate only at their next pay raise, and by a portion of the raise. Net pay never falls. Present bias is not fought, but outwitted: the painful decision lies in the future, where we are all more reasonable, and inertia afterwards ensures that people stick with it.

In the first field study at a mid-sized manufacturing firm, the participants' average savings rate rose over four pay raises from 3.5 to 13.6 percent of income. The program was adopted millions of times over in the United States and was incorporated into the Pension Protection Act of 2006, which encourages employers to use automatic enrollment with automatic escalation. Anyone looking today for a nudge that perfectly satisfies the definition will find it here: no one is coerced, everyone can exit at any time, no incentive is changed, and the effect is enormous.

Nudge Units: The State as Choice Architect

The success of the book led to an institutional innovation. In 2010, the British government under David Cameron founded the Behavioural Insights Team, immediately dubbed the Nudge Unit by the press. It tested in randomized field experiments whether, for instance, a single sentence in tax reminder letters ("Most people in your area pay their taxes on time") increased payment rates. It did. Sunstein himself headed the Office of Information and Regulatory Affairs under President Obama from 2009 to 2012 and brought the book's ideas into the regulatory practice of U.S. federal agencies. Today, more than 200 such units exist worldwide, from the World Bank and the OECD to many national governments. What they have found in their thousands of experiments, however, as we will see in Part 5, is less glamorous than the founding stories.

And Organ Donation?

Back to the hook. Johnson and Goldstein's finding was spectacular, but it concerned the registration rate, not the actual number of transplanted organs. And here it gets instructive. The transplant physician Adnan Sharif and colleagues compared, in 2019 in Kidney International, the donation and transplantation rates of 35 OECD countries, 17 with opt-out and 18 with opt-in systems. The result: in the number of deceased organ donations per million inhabitants, there was no statistically significant difference between the two systems. The opt-in countries even had significantly more living donations. Spain, world champion of organ donation for decades, owes its success, by unanimous assessment, not to its opt-out law (formally in place since 1979), but to its organizational model: transplant coordinators in every hospital, trained conversations with relatives, a high-performing infrastructure.

The reason for the gap between registration and donation is as human as it is sobering: in almost all countries, the relatives are asked, and physicians do not remove organs against their will, regardless of what the register says. A default can change the answer on a form; it does not change what happens at the bedside. The German Bundestag conducted this debate in January 2020 and rejected the opt-out solution by 379 votes to 292; instead, it adopted the so-called decision solution with an online register that launched in March 2024. England introduced opt-out in May 2020, Wales already in 2015; a decisive rise in actual donations has so far not been demonstrated in either country. The most famous nudge in the world is thus also the best example of the fact that a nudge carries only as far as the causal chain that begins behind the form allows.

Part 4: The Ethics — Manipulation or Care?

The Charge: Hausman and Welch

The philosophical critique was not long in coming. In 2010, Daniel Hausman and Brynn Welch published in the Journal of Political Philosophy the essay To Nudge or Not to Nudge, which still sets the standard for the critique today. Its core is a distinction. There are two fundamentally different ways to move someone toward a decision: you can rationally persuade them by giving them reasons, information, and arguments they can examine for themselves. Or you can steer them by exploiting the psychological weaknesses of their decision-making apparatus: their inertia, their fear of loss, their tendency toward conformity. The first way treats them as a rational being; the second treats them as a mechanism that needs to be set correctly.

Many nudges, according to Hausman and Welch, belong to the second kind. A default that relies on inertia persuades no one; it bypasses judgment. A loss frame that generates fear gives no reason; it generates an impulse. Such nudges, they write, exploit flaws in human decision-making to move people toward a particular choice, and in doing so they do exactly what libertarian paternalism claims to avoid: they violate autonomy, understood as the capacity to direct one's own life according to one's own judgment. That freedom of choice is formally preserved does not make it better, but subtler. Coercion is visible and can be fought; a nudge works in the dark.

Two aggravations are added. First: when the state nudges, it does so with a power and reach that no cafeteria manager has, and in the process it learns that steering is easier than persuading. A government that gets used to pushing citizens instead of persuading them changes the relationship between state and citizen. Second: the nudgers are themselves human. Who guarantees that the choice architects are not subject to the same biases they want to correct in others, or that they do not simply pass off their own preferences as the "welfare" of those affected? The question of who nudges the nudgers is not a rhetorical one.

The Defense: Welfare, Autonomy, Dignity

In the years that followed, Sunstein wrote an entire body of defense, the core of which is contained in the essay The Ethics of Nudging (Yale Journal on Regulation, 2015). He accepts that nudges must be measured against three standards: do they promote or undermine the welfare, the autonomy, and the dignity of those affected? And he claims that most nudges that deserve support promote all three and violate none.

The argument about welfare is the simplest: if people actually want, by their own judgment, to save more, eat more healthily, or pay their taxes on time, then a nudge helps them do what they want anyway. It is not an intrusion into their preferences, but a crutch for their implementation.

The argument about autonomy is more sophisticated. Sunstein denies that autonomy requires making every decision oneself and actively. A life in which one must think through all the options oneself for every form, every insurance policy, every device is not an autonomous life but an exhausted one. Attention is a scarce resource, and good defaults protect it so that people can spend it on what really matters to them. Autonomy, according to Sunstein, does not mean choosing everything, but determining what one wants to think about. A nudge that keeps the exit option open respects exactly that.

The argument about dignity is at the same time the concession. Sunstein admits that a nudge can treat people like children who are led to their happiness by tricks, and that precisely therein lies a violation of dignity. The line he draws is the line to manipulation: an influence is manipulative if it does not sufficiently engage the affected person's capacity for reflective and deliberative choice, but bypasses it. Information is not manipulation. A reminder is not manipulation. A subliminal stimulus, an exploited affect, a misleading frame are.

The Publicity Principle

The most important practical touchstone Sunstein proposes comes from the political philosophy of John Rawls (see Behind the Veil of Ignorance: How Rawls Reinvented Justice – and Why His Thought Experiment Still Echoes in the Ethics of Artificial Intelligence): the publicity principle. A nudge is legitimate only if the choice architect is willing and able to defend it publicly, and if the nudge still works when those affected know about it. Automatic enrollment in retirement savings passes this test: you can put it on the front page of the newspaper, and it still works, because inertia does not disappear just because one knows about it. A subliminal advertising insert does not pass it: it works only as long as no one notices it. The test is elegant because it defines manipulation through its dependence on concealment. What shuns the light is suspect.

Thaler himself has condensed the debate into three rules that he names as conditions for a legitimate nudge: the nudge must be transparent and never misleading; opting out must be as easy as possible, ideally with one click; and there must be good reason to believe that the behavior being encouraged improves the welfare of those affected. Nudges that violate one of these criteria he calls not nudges but sludge, more on that shortly.

The Third Position: Boosts Instead of Nudges

Between the charge and the defense, a third position has formed, which the psychologists Ralph Hertwig and Till Grüne-Yanoff elaborated in 2017 in Perspectives on Psychological Science: boosting. If the problem with the nudge is that it bypasses human competence, why not strengthen the competence itself? A boost teaches people to understand risks in natural frequencies instead of percentages, gives them simple rules of thumb for financial decisions, or trains them to recognize fake news by certain features. The difference from the nudge is one of principle: a nudge works only as long as the architecture stands; a boost remains when the person leaves the room. The nudge treats the bias as given and works around it; the boost treats it as changeable and works on it.

Boosts are more laborious, slower, and require the participation of those affected. Precisely therein lies their ethical advantage: they cannot manipulate at all, because they do not work without understanding. For decisions that are rare and consequential, such as choosing health insurance or a retirement plan, they are therefore often superior to the nudge. For the thousand small everyday decisions for which no one has time, the nudge remains the more realistic tool.

A Checklist

From this debate, a grid can be derived with which concrete nudges can be evaluated. It is not a formula, but it forces the right questions.

Anyone who puts these six questions to a nudge will find that the distinction between legitimate choice architecture and manipulation is neither trivial nor impossible. It does not run between nudge and non-nudge, but straight through the toolbox.

Part 5: The Effectiveness Debate — How Big Is the Push Really?

The Big Meta-Analysis

An ethical debate about nudges makes sense only if nudges work. Exactly that seemed conclusively established in 2022. Stephanie Mertens, Mario Herberz, Ulf Hahnel, and Tobias Brosch of the University of Geneva published in the Proceedings of the National Academy of Sciences the most comprehensive meta-analysis to date: more than 200 studies, more than 450 effect sizes, a good 2.1 million participants. The result: a small to medium overall effect of Cohen's d = 0.45, with a confidence interval from 0.39 to 0.52. Structure nudges such as defaults worked more strongly than information or assistance nudges; nudges on food choices worked up to two and a half times more strongly than in other domains. The authors dutifully noted a "moderate publication bias" but stood by their conclusion: nudging is an effective and widely applicable tool for behavior change.

The Counterstrike

Half a year later, a commentary appeared in the same journal with a title that left no room for interpretation: No evidence for nudging after adjusting for publication bias. Maximilian Maier, František Bartoš, T. D. Stanley, David Shanks, Adam Harris, and Eric-Jan Wagenmakers had recalculated the data of Mertens and colleagues, which the latter had exemplarily made public, with a new method: robust Bayesian meta-analysis (RoBMA), which applies several models of publication bias simultaneously and weights them by their fit to the data.

The result was sobering. After correcting for publication bias, the overall effect shrank from 0.43 to 0.04, with a credible interval that included zero. For information nudges, the adjusted effect was exactly 0.00, for finance nudges likewise; the Bayes factors in favor of the null hypothesis here were 34 and 41 respectively, which counts as strong evidence for the absence of an effect. Only for structure nudges, that is, above all defaults, did the evidence remain undecided. At the same time, the authors found strong evidence of publication bias in practically all subdomains. Their conclusion: the nudge literature is characterized by severe publication bias, and after correcting for it, no evidence remains that nudges work on average.

The pattern is familiar to readers of this vault. As with the Dunning-Kruger effect (see The Peak of Not Knowing: Why Psychology's Most Famous Effect May Be Just a Statistical Mirage), a finding that appears replicated in a hundred studies can to a considerable extent be an artifact of the publication system: studies with large effects get published, studies without an effect disappear into the file drawer, and the meta-analysis then averages over a biased selection. Two further commentaries in the same issue, by Barnabas Szaszi and colleagues and by Jonathan Bakdash and Laura Marusich, reached similar conclusions with other methods: there is no reason to expect large and consistent effects from nudges. Mertens and colleagues replied that the heterogeneity of effects was the real result, and in December 2022 they also had to publish a correction because their dataset contained studies from a retracted paper and coding errors.

The View from Practice

Perhaps the most important study on the question comes from an entirely different direction. Stefano DellaVigna and Elizabeth Linos published in 2022 in Econometrica an analysis that uses no selection of published studies, but all 126 randomized experiments of two large American nudge units, with a total of more than 23 million participants, regardless of whether the results were positive, negative, or null. The comparison with the nudge studies published in academic journals is drastic: in the academic literature, an average nudge raises the target rate by 8.7 percentage points, an increase of around 33 percent over the control group. In the nudge units, it is 1.4 percentage points, an increase of around 8 percent. The effect is thus real and highly statistically significant, but about six times smaller than in the literature.

DellaVigna and Linos examined five explanations and found that publication bias combined with the low statistical power of academic studies can explain the entire gap. One detail of their study is particularly revealing: they had academics and practitioners predict how large the effects in the nudge units would be. The academics overestimated them considerably; the nudge practitioners were almost perfectly right. Those who work with the results every day evidently know full well that the push is small.

What Follows

These findings can be misread in two ways. The first: "Nudges don't work." That is false, as DellaVigna and Linos show, and as Maier and colleagues also explicitly concede: the strong heterogeneity means that some nudges do indeed work, even if the average is indistinguishable from zero. Defaults are very probably among them; the company pension plan and Save More Tomorrow are not artifacts. The second misreading: "The debate is academic." That is also false. If the typical nudge delivers 1.4 instead of 8.7 percentage points, then it no longer justifies crowding out other policy.

Exactly this point was sharpened into a self-critique in 2023 by Nick Chater and George Loewenstein, both themselves pioneers of behavioral economics, in Behavioral and Brain Sciences. Behavioral economics, so their thesis, has shifted to the i-frame, the framing of problems as the result of individual bad decisions, and in doing so has neglected the s-frame, the framing as the result of systems, rules, and incentives. Obesity can be fought only to a limited extent with cafeteria nudges when the food industry optimizes its products for overconsumption; the climate crisis not with energy-saving tips when the carbon price is missing. Worse: the emphasis on the i-frame has been eagerly taken up by precisely those industries that fear structural regulation. Whoever locates responsibility with the individual relieves the system. Chater and Loewenstein do not demand that nudges be abolished, but that they be understood as a complement, not as a substitute for legislation. I am of the opinion that this self-correction is the most important development in the nudge debate since the book of 2008, because it poses the ethical question anew: not only "May one nudge?" but "What does one fail to do while nudging?"

Part 6: The Dark Twins — Sludge, Dark Patterns, and the Law

Sludge: Friction as a Weapon

Every tool that can steer behavior in one direction can also steer it in the other. In 2018, Thaler coined in Science the term sludge for the flip side of the nudge: friction deliberately built in to keep people from something that benefits them, or to hold them in something that harms them. The subscription you sign up for with one click and can cancel only by phone during business hours. The rebate for which you must print, fill in, and mail a form. The application for a social benefit so complicated that half of those entitled give up. Sunstein devoted an entire book to the phenomenon in 2021 and estimates that the administrative burden of government forms in the United States alone costs billions of hours per year.

Sludge is ethically clearer than the nudge, because it almost always violates the publicity principle: no company would publicly defend deliberately making cancellation difficult. And sludge strikes unequally: those with time, education, and patience fight their way through; those without them lose.

Dark Patterns: Nudging Against the User

In the digital world, a term of its own has become established for the manipulative relatives of the nudge: dark patterns, coined in 2010 by the British UX designer Harry Brignull. The list is long and familiar to every internet user. Confirmshaming: the unsubscribe link reads "No thanks, I don't want to save money." Roach motel: easy in, hard out. False urgency: the countdown that restarts with every page load. Hidden costs: the fees that appear only in the final step. Pre-selected add-ons: the travel insurance you have to deselect. Trick questions: the double negative in the cookie banner. Each of these techniques is, formally speaking, a nudge: no option is forbidden, no price changed. And each violates every criterion of the checklist from Part 4. They are not transparent, they are misleading, exit is expensive, and they serve the provider, not the user.

Dark patterns are the strongest argument of the nudge critics, because they show what happens when choice architecture falls into the hands of actors whose interest is opposed to the user's. They are at the same time the strongest argument of the nudge defenders, because they show that the distinction between legitimate and illegitimate nudges is very much possible in practice. No one has difficulty recognizing the difference between a reminder SMS and a roach motel.

The Response of the Law

The European Union has turned this insight into law. The Digital Services Act (Regulation (EU) 2022/2065), fully applicable since February 17, 2024, contains in Article 25 the first explicit prohibition of dark patterns in EU law: providers of online platforms must not design, organize, or operate their online interfaces in a way that deceives or manipulates users or otherwise materially distorts or impairs their ability to make free and informed decisions. Recital 67 expressly names examples: giving prominence to certain choices, repeatedly requesting a decision already made, making cancellation more difficult than signing up, and default settings that disadvantage the user. Violations can be punished with fines of up to 6 percent of global annual turnover. What the General Data Protection Regulation or the Unfair Commercial Practices Directive already cover, such as cookie banners and misleading advertising, remains reserved to their scope; Article 25 fills the gap in between.

The AI Act (Regulation (EU) 2024/1689) goes one step further (see The Pyramid of Risk: How the EU AI Act Tames Artificial Intelligence – and Why It Concerns the Whole World). Its Article 5, applicable since February 2, 2025, prohibits AI systems that deploy subliminal, purposefully manipulative, or deceptive techniques to materially distort people's behavior and thereby cause them significant harm, as well as systems that exploit vulnerabilities due to age, disability, or social situation. With that, the distinction of Hausman and Welch between persuasion and exploitation has become European law twenty years after its formulation: what bypasses judgment and causes harm is prohibited. Article 22 of the GDPR, the right not to be subject to a solely automated decision (see The Algorithm as Judge: Article 22 GDPR, the SCHUFA Ruling, and the Right to a Human Decision), belongs in the same line: the legislator increasingly protects not only what is decided, but the integrity of the decision process itself.

It is remarkable that the legislator has thereby codified, in essence, the same criteria the philosophical debate produced: deception, bypassing of judgment, asymmetry of entry and exit, exploitation of weakness. The ethics of nudging is no longer a purely academic question in Europe, but a compliance requirement for anyone who builds a user interface.

The Central Takeaway

If you want to take away a single lesson from this story, let it be this: you are already a choice architect, and your defaults are your ethics. Every form, every configuration file, every onboarding dialog, every permission prompt, every default setting in a piece of software you design decides what happens when the person at the other end does nothing. And the research says with great reliability: most will do nothing. The default is the decision.

Three concrete practices follow. First: set defaults for the user, not against them. Ask yourself, for every default, whether you could defend it publicly and whether it is what a well-informed user with time and leisure would choose themselves. Telemetry on by default, data sharing pre-ticked, the expensive subscription pre-selected: these are not neutral decisions but dark patterns, and since 2024 they are also legally assailable in the EU. Second: make exiting as easy as entering. Symmetry is the simplest test for sludge. If signing up costs one click and unsubscribing costs five, you know which interest your design serves. Third: trust no effect you have not measured yourself. The effectiveness debate teaches that published effect sizes are typically too high by a multiple. If you build in a nudge because a study recommends it, plan with one sixth of the reported effect, and test it with an adequate sample before you declare it a success.

And one more thing: where a decision is rare and consequential, consider whether you can build a boost instead of a nudge. A good explanation, an understandable comparison, a rule of thumb the user takes with them respects their competence in a way no default can.

Reflection Question

Thaler and Sunstein say there is no neutral choice architecture, and therefore the benevolent nudge is unavoidable and legitimate. Hausman and Welch say a nudge that exploits a person's inertia or fear treats them as a mechanism and not as a rational being, even if it benefits them. Suppose you are designing the retirement savings sign-up form for your colleagues and you know from the data that automatic enrollment more than doubles participation: do you set the default, even though you are thereby relying on their inertia rather than their insight? And if you do it because it benefits them: where exactly would you draw the line to the growth team that, using the same inertia, pre-ticks the same box for a subscription?

Cross-References in the Vault

• Losses Loom Larger: How Prospect Theory Dethroned Rational Man – and Why Its Most Famous Principle Is Now on Trial Itself – the psychological foundation of nudging: loss aversion, reference points, and the farewell to the rational utility maximizer.
• The Marshmallow Promise: How a Candy Test Redefined Willpower for Half a Century – and Was Then Put to the Test Itself – the present bias that Save More Tomorrow so elegantly outwits, and why self-control depends heavily on the environment.
• The Peak of Not Knowing: Why Psychology's Most Famous Effect May Be Just a Statistical Mirage – a case study in how a popular finding shrinks under publication bias and methodological artifacts, the pattern of the nudge debate.
• Behind the Veil of Ignorance: How Rawls Reinvented Justice – and Why His Thought Experiment Still Echoes in the Ethics of Artificial Intelligence – the origin of the publicity principle with which Sunstein separates legitimate nudges from manipulation.
• The Price of Justification: Cognitive Dissonance from Festinger to Neuroscience – why people retroactively justify decisions once made (or pre-set) and stick to them.
• The Pyramid of Risk: How the EU AI Act Tames Artificial Intelligence – and Why It Concerns the Whole World – the prohibition of manipulative AI systems in Article 5 as a codification of the manipulation boundary.
• The Algorithm as Judge: Article 22 GDPR, the SCHUFA Ruling, and the Right to a Human Decision – the protection of the decision process itself as a new line of European law.

Sources

1. Thaler, R. H. & Sunstein, C. R. (2008). Nudge: Improving Decisions About Health, Wealth, and Happiness. Yale University Press. Precursor: Thaler, R. H. & Sunstein, C. R. (2003). Libertarian Paternalism. American Economic Review 93(2), 175–179. Critical contextualization of the concept: Maier et al. (2022), reference list, PNAS
2. Johnson, E. J. & Goldstein, D. (2003). Do Defaults Save Lives? Science 302(5649), 1338–1339. Columbia Business School (abstract and PDF); Science
3. Madrian, B. C. & Shea, D. F. (2001). The Power of Suggestion: Inertia in 401(k) Participation and Savings Behavior. Quarterly Journal of Economics 116(4), 1149–1187. Oxford Academic; NBER Working Paper 7682
4. Thaler, R. H. & Benartzi, S. (2004). Save More Tomorrow: Using Behavioral Economics to Increase Employee Saving. Journal of Political Economy 112(S1), 164–187. Overview by Shlomo Benartzi (UCLA Anderson)
5. Hausman, D. M. & Welch, B. (2010). Debate: To Nudge or Not to Nudge. Journal of Political Philosophy 18(1), 123–136. Wiley Online Library; PhilPapers
6. Sunstein, C. R. (2015). The Ethics of Nudging. Yale Journal on Regulation 32(2), 413–450. Yale Journal on Regulation; SSRN
7. Hertwig, R. & Grüne-Yanoff, T. (2017). Nudging and Boosting: Steering or Empowering Good Decisions. Perspectives on Psychological Science 12(6), 973–986.
8. Mertens, S., Herberz, M., Hahnel, U. J. J. & Brosch, T. (2022). The effectiveness of nudging: A meta-analysis of choice architecture interventions across behavioral domains. PNAS 119(1), e2107346118. PubMed; correction (December 2022): PNAS; authors' reply: PNAS
9. Maier, M., Bartoš, F., Stanley, T. D., Shanks, D. R., Harris, A. J. L. & Wagenmakers, E.-J. (2022). No evidence for nudging after adjusting for publication bias. PNAS 119(31), e2200300119. PNAS (full text with table of adjusted effects); Szaszi, B. et al. (2022). No reason to expect large and consistent effects of nudge interventions. PNAS 119(31). PNAS
10. DellaVigna, S. & Linos, E. (2022). RCTs to Scale: Comprehensive Evidence From Two Nudge Units. Econometrica 90(1), 81–116. Econometric Society; NBER Working Paper 27594 (abstract with effect sizes)
11. Arshad, A., Anderson, B. & Sharif, A. (2019). Comparison of organ donation and transplantation rates between opt-out and opt-in systems. Kidney International 95(6), 1453–1460. Kidney International; PubMed
12. Chater, N. & Loewenstein, G. (2023). The i-frame and the s-frame: How focusing on individual-level solutions has led behavioral public policy astray. Behavioral and Brain Sciences 46, e147.
13. Regulation (EU) 2022/2065 (Digital Services Act), Article 25 and Recital 67. EUR-Lex; annotated version: DSA Library, Art. 25; overview: European Parliament, EPRS: Regulating dark patterns in the EU (2025)

← All articles