AI has become a fundamental component of the technological infrastructures that now automate all manner of decision-making. Its rapid ascent and adoption stand to improve the quality of these decisions, but also pose new ethical and policy challenges to fairness, non-discrimination, autonomy, due process, and privacy, among other values at the heart of liberal democracies. More generally, AI risks exacerbating information asymmetries and transforming power relations in ways that further disadvantage the most deprived in society—while also providing new ways to measure, model, and mitigate exactly these dynamics.
In practice, it can be challenging to bring together technical researchers who develop AI systems with scholars who focus on the social, legal, and ethical aspects of these systems. We believe that work in each of these areas can be greatly improved through cross-disciplinary feedback. In addition, some of the most important questions in AI policy and practice can only be answered through rigorous work that spans disciplines. One of our key goals is to illustrate—through a set of collaborative projects—the benefits of approaches that tightly integrate technical advances with social and normative considerations.
Our areas of research interest include:
- Fairness: AI has the potential to to improve the quality and fairness of high-stakes decision-making, but also poses risks that such decision-making may reproduce existing biases. Our work considers the role of AI as a supplement to, and analytic lens on, the policy decisions of human experts across a range of domains, including criminal justice, credit, and employment.
- Explainability: In some situations, the improved accuracy that AI can bring to decision-making may come at the cost of understanding the reasons behind those decisions. Further research is necessary to better understand this trade-off as a technical matter—and develop techniques and policies to manage it.
- Work: AI stands to greatly impact social and economic outcomes in the workplace, particularly for low-wage workers. Our research considers how AI is transforming the nature of work—not only through the risk of worker displacement, but by facilitating more granular control over workers who remain on the job.
- Data: AI depends on massive stocks of data, much of which is concentrated in the hands of a few commercial global actors. Our work will consider how the command of such data may shape the AI agenda, and how access to training data may be made available to a more diverse set of social actors and interests.
You can learn more about recent work by project participants on the People page.