Digital Science
Privacy-preserving synthetic health records and computational methods that let researchers work with clinical data safely.
Trustworthy AI
for healthcare.
The Computational Thinking Lab at Indiana University studies how AI can make care safer, more private, and more efficient.
What we work on
Research across healthcare AI.
Three research directions, from synthetic clinical data to human-centered and efficient models for healthcare.
Human-Centered AI
Health coaching grounded in behavior science, designed around the people it serves.
Green AI
Small, efficient language models that lower the compute and energy cost of healthcare machine learning.
Published work
Selected publications.
Peer-reviewed papers in medical AI, model evaluation, and trustworthy machine learning.
[Placeholder] Rigorous Evaluation of Large Language Models in Clinical Decision Support
Conference Conference on Health, Inference, and Learning (CHIL), 2025
[Placeholder] Calibrated Uncertainty for Trustworthy Language Model Predictions
Conference Neural Information Processing Systems (NeurIPS), 2025
[Placeholder] A Benchmark for Compositional Reasoning in Foundation Models
Conference International Conference on Machine Learning (ICML), 2024
From the lab
Research & explorations.
Notes, announcements, and write-ups from ongoing work.
Trustworthy evaluation, in the open.
We build open evaluation frameworks for healthcare AI that measure a model on several axes at once, from statistical fidelity and privacy to clinical usefulness, and tie each result to a real clinical use. The datasets and tools are public so others can test their own models the same way.
See the papers ›How we work
Built for impact and trust.
A cross-disciplinary team that turns hard questions in healthcare AI into systems people can rely on.
Rigorous evaluation.
We judge models on the axes that matter for care, including fidelity, privacy, fairness, and clinical usefulness, not just top-line accuracy.
Real-world stakes.
We work on problems where reliability is not optional, from clinical decision support to patient data.
Open science.
We release datasets, benchmarks, and code so the community can build on and scrutinize our work.
Cross-disciplinary.
A team spanning machine learning, statistics, behavior science, and clinical expertise that takes ideas from question to result.
Work with the lab