Care & Artificial Intelligence
The Care & Artificial Intelligence Laboratory studies how artificial intelligence can help us better understand healthcare and support care that fits the needs, circumstances, and lives of patients.
We combine clinical data with information from real patient–clinician encounters to study aspects of care that are difficult to capture through traditional data alone.
Our goal is not simply to automate healthcare, but to develop and evaluate AI approaches that make important features of care more visible, measurable, and useful to patients, clinicians, and researchers.
Clinical Encounter Science
Much of healthcare happens through conversation.
We study real-world patient–clinician encounters using video, audio, clinical information, and patient-reported measures to understand how care unfolds in practice.
This work examines features such as communication, shared decision-making, relational care, treatment burden, pace of interaction, and how patients and clinicians develop care plans together.
Multimodal Artificial Intelligence
Clinical care generates many forms of information at once: language, voice, behavior, imaging, laboratory results, clinical records, and patient experience.
We develop and evaluate multimodal AI methods that bring these sources together to identify patterns and understand aspects of care that may be difficult to capture using any single source of data.
AI-Assisted Measurement
Many important aspects of care currently require trained researchers to review encounters or clinical information manually.
We are studying how AI can support the measurement of shared decision-making, communication, treatment burden, and other features of patient–clinician interaction while maintaining appropriate human oversight.
AI for Better Thyroid Care
We study how artificial intelligence can help clinicians make sense of complex information across the thyroid care pathway.
This work brings together imaging, fine-needle aspiration biopsy (FNAB), cytology, molecular testing, clinical history, and other diagnostic data to support more accurate and individualized care.
Our research examines how AI can help identify when testing is appropriate, improve risk assessment, reduce unnecessary diagnostic cascades, and support better decisions about what should happen next.
This includes work on inappropriate thyroid ultrasound use, thyroid imaging, FNAB, molecular testing, and downstream diagnostic and treatment decisions.
Understanding Patients Through New Data
Our work also explores whether information in language, voice, clinical records, and longitudinal data can reveal aspects of patients’ experiences that are not readily visible in routine care.
This includes research examining whether these data can help identify when healthcare demands are becoming difficult to sustain or when care may not be fitting well into a patient’s life.
Our Approach
Our work brings together health services research, clinical expertise, artificial intelligence, data science, and patient-centered care.
Across these projects, we approach AI as a tool to augment human understanding and judgment—not as a replacement for the relationships, expertise, and conversations at the center of good care.
Our aim is to use AI to better understand care and help make it fit the person receiving it.