AI governance and model risk
Rules, standards, risk classification, validation, documentation, human oversight, liability, monitoring, and institutional accountability.
Research Program 02
AI, Data, and Public Policy
AI, data, and public governance
Examines how algorithms, data infrastructures, organizational practices, expertise, standards, and legal authority shape automated decisions and public accountability.
AI policy requires analysis of institutions, markets, and technical systems.
Artificial intelligence systems operate through data, computing infrastructure, organizational practices, labor, procurement, and legal authority. The program examines these arrangements and the consequences of deploying AI in commercial and public settings.
Research addresses safety and performance together with privacy, discrimination, transparency, competition, worker power, public participation, and the capacity of institutions to govern complex systems.
Areas of inquiry
Research priorities
Data privacy and governance
Collection, access, retention, consent, security, data markets, and organizational responsibility.
Public-sector AI
Procurement, administrative decisions, public services, benefits, law enforcement, defense, and democratic oversight.
Platforms and power
Market concentration, cloud and computing infrastructure, data advantages, labor, and political influence.
Automated decisions and evaluation
Validity, reliability, robustness, fairness, explanation, auditing, contestability, review, correction, and consequences for affected people.
AI infrastructure and applied analysis
Semiconductors, data centers, energy, supply chains, datasets, computational evaluation, technical labor, and the material foundations of AI.
Guiding questions
- What evidence should institutions require before deploying high-impact AI?
- How can regulation make automated decisions reviewable and contestable?
- How does concentrated technological power affect public authority?
- Which governance arrangements distribute AI's benefits and risks fairly?