research

For the complete list, see all publications.

AI and Alignment

Alignment as a game between policies; the aligned policy shifts toward higher reward within the base model’s coverage

Inference-time alignment

Aligning one model to diverse preferences, without retraining.

Nash equilibriummaximingeneral preferences

Foundations

Education

Bloom's taxonomy from Remember to Create, with AI cutting diagonally across the levels and metacognition spanning all of them

AI and CS education

How students learn with and about AI, and how the mistakes of AI tools can teach critical thinking.

AI literacygenerative AI in educationhigher-order thinkingcomputing education

Talks