Jump to content

Cass Sunstein

From AI Law Wiki

AI-generated text. This page was generated using artificial intelligence.

Scholarship · Algorithmic discrimination · Administrative law · Consumer protection

Cass Sunstein (Cass R. Sunstein) is the Robert Walmsley University Professor at Harvard and a legal scholar whose AI-related work examines algorithmic discrimination, administrative decision-making and consumer protection. His publications consider both the potential for algorithms to improve human decisions and the safeguards needed to address their harms.[1][2][3]

Roles and affiliations

Sunstein is the founder and director of Harvard Law School's Program on Behavioral Economics and Public Policy. From 2009 to 2012, he served as Administrator of the White House Office of Information and Regulatory Affairs. His institutional biography also records advisory work on law and public policy for the United Nations, European Commission and World Bank. These roles provide regulatory-policy context; his AI-specific arguments are set out in the scholarship below.[1]

Contributions and positions

Discrimination and accountability

In Discrimination in the Age of Algorithms, coauthored with Jon Kleinberg, Jens Ludwig and Sendhil Mullainathan, Sunstein examines how automated decisions change the detection of unlawful discrimination. The authors argue that human decision-making can obscure discriminatory treatment, while appropriately governed algorithms can make parts of a decision process easier to examine. They distinguish mathematical opacity from the practical ability to investigate discrimination. Their claim is conditional: transparency and equity benefits require safeguards and do not follow automatically from using an algorithm. They also argue that specifying an algorithm can expose tradeoffs between competing values.[4]

Administrative decision-making

In Governing by Algorithm? (2022), Sunstein distinguishes bias from “noise,” meaning unwanted variation in judgments. He argues that algorithms offer administrative agencies a way to reduce inconsistent decisions and some cognitive biases. At the same time, he acknowledges that discriminatory inputs or prediction targets can reproduce discrimination. The article presents an argument for carefully constructed systems in public administration, including adjudication and prosecution; it does not establish that automated decisions are invariably fair or legally permissible.[2]

Consumer protection

Algorithmic Harm (2025), written with Oren Bar-Gill, considers how algorithms can assist consumers but also exploit limited information and behavioral biases. The public publisher abstract identifies short-term thinking and unrealistic optimism as potential sources of vulnerability and argues for protective regulation. The book also extends these concerns to workers, investors and political participants. These are the authors' policy arguments, rather than a description of a single enacted AI regulatory regime.[3]

Limits of prediction

Discussing Imperfect Oracle: What AI Can and Cannot Do at Harvard Law School on November 12, 2025, Sunstein argued that AI can help where useful data exist and human biases impede judgment. He also emphasized limits to forecasting outcomes shaped by unpredictable social interactions. Harvard's account describes examples involving judicial decisions, medicine and political change; it reports his presentation rather than independently validating each empirical claim.[5]

Selected works

News and coverage

References

Profile sources last reviewed September 12, 2026. Dated positions and developments are identified in the text.