What is AI in health care?
The AI Glossary
In our classes, we often pause to explain the ideas behind AI and the language used to describe it. These posts grew out of those conversations. They’re for anyone who would like to understand AI better, with no technical background assumed.
AI in health care is the use of computational systems that learn patterns, generate content or select actions to support health-related work. The field includes care delivery, administration, public health, biomedical research and support outside clinical facilities. These uses do not all require the same models or the same evidence.
A useful starting point is the job being done. Predicting which referrals may stall is different from diagnosing a disease. Drafting a note is different from signing it. Searching published evidence is different from establishing that a treatment benefits a particular patient. Broad claims about “medical AI” often conceal these distinctions.
The model is only one part of a working service. It needs appropriate information, a user interface, connections to other systems and rules governing actions. People need to know when to rely on it, when to question it and how to recover from a failure. A strong model cannot compensate for an undefined workflow.
Evidence should match the claim. Performance on an examination or historical dataset can justify further investigation. It does not by itself demonstrate improved care. A simulation can reveal errors, while a prospective evaluation observes what happens when the system is used in its intended setting. Patient outcomes, workload and access may need separate measurement.
Human involvement is important, but human–AI combinations are not automatically better than either alone. Studies have found that results depend on the task and interaction. A clinician can miss useful advice or be misled by a confident error. The collaboration itself therefore needs evaluation rather than being assumed safe because a person remains present.
Clinical and administrative categories also do not create a simple high-risk/low-risk divide. An appointment-routing error or coverage decision can harm someone even without making a diagnosis.
The practical question I would start with is: what should happen differently for the person receiving or delivering care? A referral completed sooner, a clear explanation or a reliably followed-up result provides something concrete to test. AI becomes useful through those changes in the service, not through the sophistication of the demonstration alone.
Sources: Panch, Mattie and Celi — The “Inconvenient Truth” About AI in Healthcare (2019); Vaccaro, Almaatouq and Malone — When Combinations of Humans and AI Are Useful (2024); Goh and colleagues — Large Language Model Influence on Diagnostic Reasoning: A Randomized Clinical Trial (2024)
To learn more, take a look at the AI in Health Care Certificate of Specialization from the Advanced Learning Academy at Harvard T.H. Chan School of Public Health. You can take an individual course or work toward the full certificate.
AI is a fast-moving field. This explanation reflects our class discussions and related material at the time of writing. Terminology, capabilities and tools may change.
For suggested additions or corrections, please email me at tpanch@hsph.harvard.edu.
Dr. Trishan Panch
Program Director, AI in Health Care
Harvard T.H. Chan School of Public Health — Advanced Learning Academy