Showing papers by "Cass R. Sunstein published in 2019"

Discrimination in the Age of Algorithms

Abstract: The law forbids discrimination. But the ambiguity of human decision-making often makes it extraordinarily hard for the legal system to know whether anyone has actually discriminated. To understand how algorithms affect discrimination, we must therefore also understand how they affect the problem of detecting discrimination. By one measure, algorithms are fundamentally opaque, not just cognitively but even mathematically. Yet for the task of proving discrimination, processes involving algorithms can provide crucial forms of transparency that are otherwise unavailable. These benefits do not happen automatically. But with appropriate requirements in place, the use of algorithms will make it possible to more easily examine and interrogate the entire decision process, thereby making it far easier to know whether discrimination has occurred. By forcing a new level of specificity, the use of algorithms also highlights, and makes transparent, central tradeoffs among competing values. Algorithms are not only a threat to be regulated; with the right safeguards in place, they have the potential to be a positive force for equity.

Trusting nudges? Lessons from an international survey

Abstract: In the past decade, policymakers have increasingly used behaviourally informed policies, including ‘nudges,’ to produce desirable social outcomes. But do people actually endorse those policies? Thi...

Discrimination in the Age of Algorithms

Abstract: The law forbids discrimination. But the ambiguity of human decision-making often makes it extraordinarily hard for the legal system to know whether anyone has actually discriminated. To understand how algorithms affect discrimination, we must therefore also understand how they affect the problem of detecting discrimination. By one measure, algorithms are fundamentally opaque, not just cognitively but even mathematically. Yet for the task of proving discrimination, processes involving algorithms can provide crucial forms of transparency that are otherwise unavailable. These benefits do not happen automatically. But with appropriate requirements in place, the use of algorithms will make it possible to more easily examine and interrogate the entire decision process, thereby making it far easier to know whether discrimination has occurred. By forcing a new level of specificity, the use of algorithms also highlights, and makes transparent, central tradeoffs among competing values. Algorithms are not only a threat to be regulated; with the right safeguards in place, they have the potential to be a positive force for equity.

Ruining popcorn? The welfare effects of information

Abstract: Some information is beneficial; it makes people’s lives go better. Some information is harmful; it makes people’s lives go worse. Some information has no welfare effects at all; people neither gain nor lose from it. Under prevailing executive orders, federal agencies must investigate the welfare effects of information by reference to cost-benefit analysis. Federal agencies have (1) claimed that quantification of benefits is essentially impossible; (2) engaged in “breakeven analysis”; (3) projected various endpoints, such as health benefits or purely economic savings; and (4) relied on private willingness to pay for the relevant information. All of these approaches run into serious objections. With respect to (4), people may lack the information that would permit them to make good decisions about how much to pay for (more) information; they may not know the welfare effects of information. Their tastes and values may shift over time, in part as a result of information. These points suggest the need to take the willingness-to-pay criterion with many grains of salt, and to learn more about the actual effects of information, and of the behavioral changes produced by information, on people’s experienced well-being.

How Much Is Data Privacy Worth? A Preliminary Investigation

Abstract: Do consumers value data privacy? How much? In a survey of 2,416 Americans, we find that the median consumer is willing to pay just $5 per month to maintain data privacy (along specified dimensions), but would demand $80 to allow access to personal data. This is a “superendowment effect,” much higher than the 1:2 ratio often found between willingness to pay and willingness to accept. In addition, people demand significantly more money to allow access to personal data when primed that such data includes health-related data than when primed that such data includes demographic data. We analyze reasons for these disparities and offer some notations on their implications for theory and practice. A general theme is that because of a lack of information and behavioural biases, both willingness to pay and willingness to accept measures are highly unreliable guides to the welfare effects of retaining or giving up data privacy. Gertrude Stein’s comment about Oakland, California may hold for consumer valuations of data privacy: “There is no there there.” For guidance, policymakers should give little or no attention to either of those conventional measures of economic value, at least when steps are not taken to overcome deficits in information and behavioural biases.

Heuristics and Public Policy: Decision-making Under Bounded Rationality:

Abstract: How do human beings make decisions when, as the evidence indicates, the assumptions of the Bayesian rationality approach in economics do not hold? Do human beings optimize, or can they? Sev...

Algorithms, Correcting Biases

Abstract: A great deal of theoretical work explores the possibility that algorithms may be biased in one or another respect. But for purposes of law and policy, some of the most important empirical research finds exactly the opposite. In the context of bail decisions, an algorithm designed to predict flight risk does much better than human judges, in large part because the latter place an excessive emphasis on the current offense. Current Offense Bias, as we might call it, is best seen as a cousin of “availability bias,” a well-known source of mistaken probability judgments. The broader lesson is that well-designed algorithms should be able to avoid cognitive biases of many kinds. Existing research on bail decisions also casts a new light on how to think about the risk that algorithms will discriminate on the basis of race (or other factors). Algorithms can easily be designed so as to avoid taking account of race (or other factors). They can also be constrained so as to produce whatever kind of racial balance is sought, and thus to reveal tradeoffs among various social values.

Nudging: A Very Short Guide

Abstract: This brief essay offers a general introduction to the idea of nudging, along with a list of ten of the most important “nudges.” It also provides a short discussion of the question whether to create a separate “behavioral insights unit” or instead to rely on existing institutions.

Sludge and Ordeals

Abstract: In 2015, the United States government imposed 9.78 billion hours of paperwork burdens on the American people. Many of these hours are best categorized as “sludge,” reducing access to important licenses, programs, and benefits. Because of the sheer costs of sludge, rational people are effectively denied life-changing goods and services; the problem is compounded by the existence of behavioral biases, including inertia, present bias, and unrealistic optimism. In principle, a serious deregulatory effort should be undertaken to reduce sludge, through automatic enrollment, greatly simplified forms, and reminders. At the same time, sludge can promote legitimate goals. First, it can protect program integrity, which means that policymakers might have to make difficult tradeoffs between (1) granting benefits to people who are not entitled to them and (2) denying benefits to people who are entitled to them. Second, it can overcome impulsivity, recklessness, and self-control problems. Third, it can prevent intrusions on privacy. Fourth, it can serve as a rationing device, ensuring that benefits go to people who most need them. In most cases, these defenses of sludge turn out to be more attractive in principle than in practice. For sludge, a form of cost-benefit analysis is essential, and it will often argue in favor of a neglected form of deregulation: sludge reduction. For both public and private institutions,“Sludge Audits” should become routine. Various suggestions are offered for new action by the Office of Information and Regulatory Affairs, which oversees the Paperwork Reduction Act; for courts; and for Congress.

Which nudges do people like? A national survey

Abstract: In surveys, majorities of Americans disapprove of twelve hypothetical nudges (seven involving default rules, five involving education campaigns or disclosure requirements). These results provide an illuminating contrast with the majority support for twenty-two nudges that were also tested, and that are more realistic examples of the kinds of nudges that have been adopted or seriously considered in democratic nations. In general (and with some interesting exceptions), there is a strikingly broad consensus, across partisan lines, about which nudges do and do not deserve support. The best understanding of the data is that people dislike those nudges that (a) promote what people see as illicit ends or (b) are perceived as inconsistent with either the interests or values of most choosers. A ranking of the thirty-four nudges, in terms of their popularity, is provided, along with reports of differences (when they exist) among Democrats, Republicans, and Independents.

Rear Visibility and Some Unresolved Problems for Economic Analysis (With Notes on Experience Goods)

Abstract: In 2014, the National Highway Traffic Safety Administration finalized its rear visibility regulation, which requires cameras in all new vehicles, with the goal of allowing drivers to see what is behind them and thus reducing backover accidents. In 2018, the Trump administration embraced the regulation. The rear visibility rule raises numerous puzzles. First: Congress’ grant of authority was essentially standardless – perhaps the most open-ended in all of federal regulatory law. Second: It is not easy to identify a market failure to justify the regulation. Third: The monetized costs of the regulation greatly exceeded the monetized benefits, and yet on welfare grounds, the regulation can plausibly be counted as a significant success. Rearview cameras produce a set of benefits that are hard to quantify, including increased ease of driving, and those benefits might have been made a part of “breakeven analysis,” accompanying standard cost-benefit analysis. In addition, rearview cameras significantly improve the experience of driving, and it is plausible to think that in deciding whether to demand them, many vehicle purchasers did not sufficiently anticipate that improvement. This is a problem of limited foresight; rearview cameras are “experience goods.” A survey conducted in 2019 strongly supports this proposition, finding that about 56 percent of consumers would demand at least $300 to buy a car without a rearview camera, and that fewer than 6 percent would demand $50 or less. Almost all of that 6 percent consists of people who do not own a car with a rearview camera. (The per-person cost is usually under $50.) These conclusions have general implications for other domains in which regulation has the potential to improve social welfare, even if it fails standard cost-benefit analysis; the defining category involves situations in which people lack experience with a good whose provision might have highly beneficial welfare effects.

Rear Visibility and Some Unresolved Problems for Economic Analysis (With Notes on Experience Goods)

Abstract: In 2014, the National Highway Traffic Safety Administration finalized its rear visibility regulation, which requires cameras in all new vehicles, with the goal of allowing drivers to see what is behind them and thus reducing backover accidents. In 2018, the Trump administration embraced the regulation. The rear visibility rule raises numerous puzzles. First, Congress’ grant of authority was essentially standardless – perhaps the most open-ended in all of federal regulatory law. Second, it is not easy to identify a market failure to justify the regulation. Third, the monetized costs of the regulation greatly exceeded the monetized benefits, and yet on welfare grounds, the regulation can plausibly be counted as a significant success. Rearview cameras produce a set of benefits that are hard to quantify, including increased ease of driving, and those benefits might have been made a part of “breakeven analysis,” accompanying standard cost-benefit analysis. In addition, rearview cameras significantly improve the experience of driving, and it is plausible to think that in deciding whether to demand them, many vehicle purchasers did not sufficiently anticipate that improvement. This is a problem of limited foresight; rearview cameras are “experience goods.” A survey conducted in 2019 strongly supports this proposition, finding that about 56 % of consumers would demand at least $50.) These conclusions have general implications for other domains in which regulation has the potential to improve social welfare, even if it fails standard cost-benefit analysis; the defining category involves situations in which people lack experience with a good whose provision might have highly beneficial welfare effects.

Drawing false inferences from mandated disclosures

Abstract: Disclosure mandates are pervasive. Though designed to inform consumers, such mandates may lead consumers to draw false inferences – for example, that a product is harmful when it is not. When deciding to require disclosure of an ingredient in or characteristic of a product, regulators may be motivated by evidence that the ingredient or characteristic is harmful to consumers. But they may also be motivated by a belief that consumers have a right to know what they are buying or by interest-group pressure. Consumers who misperceive the regulator's true motive, or mix of motives, will draw false inferences from the mandated disclosure. If consumers think that the disclosure is motivated by evidence of harm, when in fact it is motivated by a belief in a right to know or by interest-group pressure, then they will be inefficiently deterred from purchasing the product. We analyze this general concern about disclosure mandates. We also offer survey evidence demonstrating that the risk of false inferences is serious and real.

Back to Mill? Behavioral Welfare Economics

Abstract: A growing body of normative work explores how deference to people’s choices might be reconciled with behavioral findings about human error. This work has strong implications for economic analysis of law, cost-benefit analysis, and regulatory policy. in light of behavioral findings, regulators should adopt a working presumption in favor of respect for people’s self-regarding choices, but only if those choices are adequately informed and sufficiently free from behavioral biases. The working presumption should itself be rebuttable on welfare grounds, with an understanding that the ends that people choose might make their lives go less well. For example, people might die prematurely or suffer from serious illness, and what they receive in return might not (on any plausible account of welfare) be nearly enough. The underlying reason might involve a lack of information or a behavioral bias, identifiable or not, in which case intervention can fit with the working presumption; but the real problem might involve philosophical questions about the proper understanding of welfare, and about what it means for people to have a good life.

Is cost–benefit analysis a foreign language?:

Abstract: Do people think better in a foreign language? In some ways, yes. There is considerable evidence to this effect, at least to the extent that they are less likely to rely on intuitions that can lead to serious errors. This finding reinforces, and makes more plausible, a central claim in regulatory policy, which involves the value of cost-benefit analysis. In a sense, cost-benefit analysis is a foreign language, and it reduces the risk that people will rely on intuitions that cause serious errors.

El ascenso de la economía del comportamiento: Misbehaving, de Richard Thaler

Abstract: espanolLa economia del comportamiento surgio en la decada 1980, debido sobre todo a la obra de Richard Thaler, quien exploro la relevancia del efecto dotacion, la contabilidad mental, la preocupacion por la justicia y otras “anomalias”, segun la teoria economica estandar. Su maravilloso libro, Misbehaving, ofrece una descripcion narrativa del surgimiento de estas ideas, y explora algunas de sus implicaciones para el futuro. Los retos continuos incluyen el de hacer predicciones cuando los resultados del comportamiento inciden en direcciones diferentes (como, p. ej., cuando el sesgo optimista entra en conflicto con el sesgo de disponibilidad); entender la linea entre empujar y manipular; y aplicar los resultados del comportamiento a desafios apremiantes de politica publica, como la pobreza, la educacion, el terrorismo y el cambio climatico. portuguesA economia comportamental emergiu na decada de 1980, principalmente, devido ao trabalho de Richard Thaler, que explorou a relevância do efeito de dotacao, da contabilidade mental, da preocupacao com a justica e de outras “anomalias”, de acordo com a teoria economica padrao. Seu maravilhoso livro, Misbehaving, oferece uma descricao narrativa do surgimento dessas ideias e explora algumas de suas implicacoes para o futuro. Os desafios continuos incluem fazer previsoes quando os resultados comportamentais atingirem direcoes diferentes (como, por exemplo, quando a tendencia otimista entra em conflito com a de disponibilidade); entender a linha entre empurrar e manipular; e aplicar resultados comportamentais nos desafiosemergentes das politicas publicas, como pobreza, educacao, terrorismo e mudanca climatica. EnglishBehavioral economics emerged in the 1980s, above all because of the creative work of Richard Thaler, exploring the relevance of the endowment effect, mental accounting, concern for fairness, and other “anomalies” from the standpoint of standard economic theory. His engaging book, Misbehaving, offers a narrative account of how these ideas came about, and also explores some of their implications for the future. Continuing challenges include making predictions when behavioral findings cut in different directions (as, for example, where optimistic bias conflicts with availability bias); understanding the line between nudging and manipulation; and applying behavioral findings to pressing public policy challenges, such as poverty, education, terrorism, and climate change.

Chevron As Law

Abstract: Chevron v. NRDC, the foundation for much of contemporary administrative law, is now under siege. The central objection, connected with longstanding challenges to the legitimacy of the modern regulatory state, is that the decision amounts to an unwarranted transfer of interpretive authority from courts to the executive branch. Some people think that the transfer is a recipe for a form of authoritarianism – and inconsistent with the proposition that it is the province of the judiciary to say what the law is. To assess such objections, the starting point is simple: Whether courts should defer to agency interpretations of law depends largely on legislative instructions. Under the Constitution, Congress has broad power to require courts to defer to agency interpretations (in the face of ambiguity), or to forbid them from doing so. If congressional instructions are the touchstone, and if the Administrative Procedure Act is the guiding text, there is a plausible argument that Chevron was wrong when decided; but the issue is intriguingly cloudy if the APA’s text is taken in its context. In these circumstances, Chevron should not be overruled. Doing so would introduce a great deal of confusion and increase the role of political judgments within the courts of appeals. Nonetheless, Chevron’s critics have legitimate concerns. Those concerns should be taken into account (1) by insisting on a fully independent judicial role in deciding whether a statute is ambiguous at Step One; (2) by invalidating arbitrary or unreasonable agency interpretations at Step Two; and (3) by deploying canons of construction, including those that are designed to serve nondelegation functions and thus to cabin executive authority.

Fuel Economy 2.0

Abstract: Motor vehicle fuel-economy standards have long been a cornerstone of U.S. policy to reduce fuel consumption in the light-duty vehicle fleet. In 2011 and 2012 these standards were significantly expanded in an effort to achieve steep reductions in oil demand and greenhouse gas emissions through 2025. In 2018, following a review of the standards, the Environmental Protection Agency and National Highway Traffic Safety Administration proposed to instead freeze the standards at 2020 levels, citing high program costs (and potential safety issues). The current debate over the future of U.S. efficiency standards provides an opportunity to consider whether the existing approach could be improved to achieve environmental goals at a lower cost. The current policy prescribes standards that focus on efficiency alone, as opposed to lifetime consumption, and treats vehicle categories differentially, meaning that it imposes unnecessarily high costs and does not deliver guaranteed petroleum/greenhouse gas (GHG) savings. On the basis of a commitment to cost-benefit analysis, defining U.S. regulatory policy for more than 30 years, we propose a novel reform that has three main features: 1) the regulation of expected fuel consumption/GHG emissions directly without consideration of the type or size of the vehicle; 2) use of existing data to assign lifetime fuel consumption/GHG emissions to each model; and 3) creation of a robust cap-and-trade market for automakers to reduce compliance costs. We show that this approach would increase the certainty of reductions in fuel consumption/GHG emissions in transportation and do so at a far lower cost per gallon/ton of GHG avoided. Such an approach would be consistent with existing statutory authorities at the Environmental Protection Agency and the Department of Transportation.

Lapidation and Apology

Abstract: Groups of people, outraged by some real or imagined transgression, often respond in a way that is wildly disproportionate to the occasion, thus ruining the transgressor’s day, month, year, or life. To capture that phenomenon, we might repurpose an old word: lapidation. Technically, the word is a synonym for stoning, but it sounds much less violent. It is also obscure, which makes it easier to enlist for contemporary purposes. Lapidation plays a role in affirming, and helping to constitute, tribal identity. It typically occurs when a transgressor is taken to have violated a taboo, which helps account for the different people and events that trigger left-of-center and right-of-center lapidation. One of the problems with lapidation is that it often accomplishes little; it expresses outrage, and allows people to signal their identity, but does no more. Victims of lapidation might be tempted to apologize, but apologies can prove ineffective or even make things worse, depending on the nature of the lapidators.

Behaviorally Informed Policy: A Brisk Progress Report

Abstract: The goal of this essay is to offer a brisk progress report about behaviorally informed policy and law, while also providing some conceptual clarifications. It identifies a diverse range of initiatives, focusing largely on experience in the United States, that involve nudging and uses of behavioral science. It also explores prominent objections, coming both from those who believe that nudges are unduly aggressive, and should be avoided, and from those who believe that they are too weak and timid, and should be replaced or supplemented with mandates and bans.

Should Public Figures Apologize? Preliminary Evidence and Speculations

Abstract: In the modern era, the statements and actions of public figures are scrutinized with great care, and it often emerges that they have said or done things that many people consider objectionable, hurtful, offensive, or despicable. A persistent question is whether public figures should apologize for those statements or actions. Suppose that an apology has a purely strategic motivation: helping a politician to be elected or reelected, helping an executive to keep his job, helping a nominee to be confirmed by the U.S. Senate. Empirical work presented here suggests that an apology might well turn out to be futile or even counterproductive. One reason is Bayesian; an apology produces updating that can be unfavorable to the apologizer (by, for example, resolving doubts about whether the apologizer actually said or did the objectionable thing, and about whether what the apologizer did was actually objectionable). Another reason is behavioral; an apology triggers the public’s attention, makes the public figure’s wrongdoing more salient, and can help define him or her. But many open questions remain about the reasons why apologies by public figures fail, and about the circumstances in which they might turn out to be effective.

Cinquenta tons de manipulação

Abstract: RESUMO: Uma declaração ou ação pode ser considerada manipuladora se ela não fomenta ou estimula a capacidade de uma escolha reflexiva e deliberativa. Um dos problemas com a manipulação assim compreendida, é que ela falha em respeitar a autonomia das pessoas, e é afronta à dignidade delas. Outro problema é que se forem produtos de manipulação, as escolhas das pessoas podem falhar em promover seu próprio bem-estar, e podem ainda promover o bem-estar do manipulador. Nessa medida, a objeção central à manipulação está enraizada em uma versão do princípio do prejuízo de Mill: pessoas sabem o que é de seu melhor interesse e deveriam ter uma oportunidade (livre de manipulação) em fazer tal escolha. Com fundamentos em bem-estar, a norma contra manipulação pode ser vista como um tipo de heurística, um tipo que geralmente funciona bem, mas que também pode levar a sérios equívocos, pelo menos quando o manipulador é bem informado e de forma genuína interessado no bem-estar do optante. Para o sistema legal um quebra-cabeça complexo é o porquê de a manipulação ser raramente policiada. A resposta mais simples é que a manipulação tem tantos tons, e que em uma ordem social que valorize o livre mercado e é comprometida com a liberdade de expressão, é excepcionalmente difícil de regular a manipulação como tal. Mas à medida que os motivos do manipulador se tornam mais egoístas ou venais, e como os esforços para contornar as capacidades deliberativas das pessoas se tornam mais bem-sucedidos, as objeções éticas à manipulação tornam-se muito vigorosas, e o argumento para uma resposta legal é fortificado. A análise da manipulação se debruça em questões emergentes relacionadas a primeira emenda, questões essas levantadas, por exemplo, no caso de informações compulsórias, especialmente no contexto de advertências gráficas de saúde. De forma relevante também pode ajudar a orientar a regulação de produtos financeiros, onde a manipulação de escolhas do consumidor é uma preocupação evidente, mas raramente explícita.

Ismism, Or Has Liberalism Ruined Everything?

Abstract: There has been considerable recent discussion of the social effects of “liberalism,” which are said to include (among other things) a growth in out-of-wedlock childbirth, repudiation of traditions (religious and otherwise), a rise in populism, increased reliance on technocracy, inequality, environmental degradation, sexual promiscuity, deterioration of civic associations, a diminution of civic virtue, political correctness on university campuses, and a general sense of alienation. There is good reason for skepticism about these claims. Liberalism is not a person, and it is not an agent in history. Claims about the supposedly adverse social effects of liberalism are best taken not as causal claims at all, but as normative objections that should be defended on their merits. These propositions are elaborated with reference to three subordinate propositions: (1) liberalism, as such, does not lack the resources to defend traditions; (2) liberalism, as such, hardly rejects the idea of “constraint,” though the domains in which liberals accept constraints differ from those of antiliberals, and vary over time; (3) liberalism, as such, does not dishonor the idea of “honor.” There is a general point here about the difficulty of demonstrating, and the potential recklessness of claiming, that one or another “ism” is causally associated with concrete social developments.