Senior Researcher, Microsoft Research Cambridge

Dr Advait Sarkar is an affiliated lecturer at the University of Cambridge, honorary lecturer at University College London, and a senior researcher at Microsoft Research Cambridge. His work focuses on how artificial intelligence reshapes human thinking, judgment, and expertise, and on how digital systems can be designed to support critical thinking rather than replace it. He studies AI not primarily as a tool for automation or content generation, but as an intervention in how people reason, learn, and make decisions.

His research spans human‑AI interaction, the future of knowledge work, and the design of tools for thought, with applications ranging from spreadsheets and data analysis to generative AI systems used by professionals. He is known for empirical work showing how different ways of eliciting human judgment, such as ranking rather than scoring, can produce more reliable decisions, as well as for critical analyses of common AI narratives, including the limits of “human‑AI collaboration” as a policy metaphor.

Advait’s work regularly informs debates on AI governance, productivity, and digital skills, and he has spoken internationally on the societal implications of AI, including at TEDAI Vienna. He works closely with policymakers, industry, and academia on questions of how AI systems should be designed and regulated to preserve human agency, accountability, and long‑term cognitive capacity.

Some key papers:

• The Impact of Generative AI on Critical Thinking — Empirical evidence from a large survey of knowledge workers showing that higher reliance on generative AI is associated with reduced cognitive effort and inflated confidence, with direct implications for decision quality and accountability in policy and government contexts.
• AI Should Challenge, Not Obey — Proposes that AI systems should act as provocateurs that surface risks, alternatives, and critiques, rather than compliant assistants, reframing how AI should be used in high‑stakes judgment and policy analysis.
• The Metacognitive Demands and Opportunities of Generative AI — Introduces metacognition as a lens for understanding why generative AI is hard to use well, and why institutional capability depends on supporting users’ ability to monitor and regulate their own thinking.
• Enough With “Human‑AI Collaboration” — Critiques the dominant “AI collaborator” metaphor in policy and industry discourse, showing how it obscures responsibility, labour, and power, with consequences for governance and procurement.
• Exploring Perspectives on the Impact of Artificial Intelligence on the Creativity of Knowledge Work — Argues that generative AI shifts creativity from material production to critical integration, reframing policy debates about originality, authorship, credit, and regulation by showing creativity is a property of socio‑technical processes rather than AI outputs alone. Synthesises early empirical studies of AI‑assisted work showing how tasks shift from production to evaluation and oversight, a pattern now mirrored across many forms of professional work.