Lab profile
About Healthy ML
The Healthy ML group focuses on creating and applying machine learning to understand and improve health in ways that are robust, private and fair. The group tackles novel technical opportunities for machine learning in health, working to ensure models do not learn biased rules or recommendations that harm minorities or minoritized populations. They work on developing machine learning models that can be safely deployed in healthcare settings while addressing challenges of robustness, fairness, privacy, and appropriate system design.
01 · Research focus
Research areas and highlights
Topics and recent work as stated on the lab’s own site.
- The Limits of Fair Medical Imaging AI in Real-World GeneralizationResearch on fairness and generalization challenges in medical imaging AI systems.
- BendVLM: Test-Time Debiasing of Vision-Language EmbeddingsWork on debiasing vision-language models at test time.
- When Personalization Harms: Reconsidering the Use of Group Attributes in PredictionInvestigation of when and how personalization using group attributes can be harmful in prediction tasks.
02 · People
Leadership and researchers
People listed on the lab’s team pages, by role.
03 · Research intelligence
Research Intelligence evidence
Publications, grants and clinical trials linked to the lab’s researchers in Health AI Central’s research data.
Publications & preprints
- MAGIC: an international network for evaluating generative artificial intelligence in global health.Nature medicine · 2026
- Avoiding common failures in AI for health and medicine.Cell · 2026
- Divergent impacts of explainable AI for dermatological diagnosis on clinicians versus lay people.Nature medicine · 2026
Clinical-trial records
- Stress and Recovery in Frontline COVID-19 WorkersNCT04713111 · Completed
04 · Updates
Updates and lab news
News and events from the lab’s own site.
- Escaping the Mode: Multi-Answer Reinforcement Learning in LMs published in ICML 2026Publication
- Position: Benchmarks Do Not Measure Deployment Readiness in Clinical AI published in ICML 2026Publication
- Rank-Constrained Adaptation for Reliable Real-World Performance published in NeurIPS 2026Publication
- Poison-then-Hide: Finetuning-Activated Backdoor Attack on Pretrained Vision Encoders published in NeurIPS 2026Publication
- Avoiding Common Failures in AI for Health and Medicine published in CellPublication
- Upcoming talks by Dr. GhassemiAnnouncement
05 · Official presence
Lab sources and channels
The lab’s own pages and accounts.
06 · Related
Related labs and centers
Labs at the same institution and labs working on similar topics.