Jump to content

Sandra Wachter

From AI Law Wiki

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

Scholarship · Privacy and discrimination

Sandra Wachter is Professor of Technology and Regulation at the Oxford Internet Institute, University of Oxford. Her research concerns the legal and ethical implications of AI, including algorithmic discrimination, explainability, profiling and data protection.[1]

Roles and affiliations

Wachter leads the Governance of Emerging Technologies research programme at the Oxford Internet Institute. Her work combines analysis of legal protections with proposals for technical methods that can help people understand or challenge automated decisions. Oxford identifies non-discrimination law, privacy, robotics and accountability among her research interests.[1]

Contributions and positions

Explanations that help people act

With Brent Mittelstadt and Chris Russell, Wachter developed an account of counterfactual explanations for automated decisions. Their paper asks what an affected person needs an explanation to accomplish: understand a result, contest it, or identify changes that could produce a different result in the future. A counterfactual describes a change that would alter a model's output without necessarily exposing the model's complete internal logic. The authors relate these purposes to the GDPR while recognizing barriers to providing conventional explanations of complex systems.[2]

This approach separates an explanation useful to an individual from a complete technical description of an algorithm. It is a proposed way to support understanding and challenge, rather than a claim that the GDPR mandates this particular method in every case.[2]

Inferences and data-protection rights

In A Right to Reasonable Inferences (2019), Wachter and Mittelstadt argue that data-protection law inadequately addresses some predictions and conclusions drawn about people. They propose greater justification for high-risk inferences, including the suitability of the underlying data, the purpose of the inference and the reliability of the methods. Their proposal also includes an opportunity to challenge unreasonable inferences. The article's proposed right should be distinguished from rights already established by legislation or a court.[3]

The focus is on what an organization concludes about an individual as well as what information it collects. In the authors' account, control over input data alone does not answer questions about potentially harmful predictions used in consequential decisions.[3]

Fairness and discrimination law

In Why Fairness Cannot Be Automated, Wachter, Mittelstadt and Russell examine the gap between statistical fairness measures and European non-discrimination law. They argue that legal assessment depends on context and interpretation that cannot simply be replaced by a metric. Their proposed conditional demographic disparity measure is intended to support consistent evidence and procedures, while preserving judicial interpretation. The manuscript therefore treats technical measurement as an aid to legal evaluation, not a substitute for it.[4]

Selected works

News and coverage

References

Profile sources last reviewed September 7, 2026. Dated positions and developments are identified in the text.