
Reason from evidence.
We study how AI connects clinical evidence to its decisions, and how to measure when those decisions go wrong.
HARVARD MEDICAL SCHOOL · BIOMEDICAL INFORMATICS
Reasoning from evidence. Decisions over time.
Learning to assist.

We study how AI connects clinical evidence to its decisions, and how to measure when those decisions go wrong.

We study clinical agents and how to evaluate decisions across changing patients, resources, and care.

We study how AI learns from expert technique to support bounded forms of robotic assistance alongside clinicians.
SELECTED RESEARCH

Clinical reasoning
Testing clinical language models through simulated patient conversations.
Nature Medicine · 2025
Human–AI collaboration
Measuring why AI assistance helps some radiologists and hinders others.
Nature Medicine · 2024
Clinical monitoring & evaluation
Benchmarking clinical prediction across emergency-care records, vital signs, and waveforms.
NeurIPS Datasets and Benchmarks · 2023
Clinical reasoning failures
Testing how repeated user challenges change a model’s clinical judgment.
HeaLing at ACL · 2026
Learning from clinical language
Learning to recognize chest X-ray findings from reports without disease-specific labels.
Nature Biomedical Engineering · 2022

THE PEOPLE BEHIND THE WORK
Computer scientists and clinicians, working together at Harvard Medical School.
Meet the labJOIN THE LAB