What is interaction design?
The Harvard Chan 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.
Interaction design concerns what a person can do with a system, how the system responds and how the person understands the result. It includes the sequence of actions, the feedback after an action and the ways to correct mistakes. It is broader than choosing the color or position of controls.
Imagine an AI assistant preparing an email. The person needs to distinguish asking for a draft, approving its wording and authorizing it to be sent. If the interface makes these actions look interchangeable, the system may behave in a way the person did not intend. Good interaction design makes the consequences legible.
Feedback is the information a system provides about what is happening. “Preparing draft,” “waiting for approval” and “sent” describe different states. State means the current condition of the software and its task. A progress message should correspond to the actual process, rather than simply being reassuring text generated by a model.
Interaction design also includes recovery. Can a person stop the operation, correct a recipient or return to an earlier version? What happens if the connection drops? A smooth ordinary path is not enough when important work depends on understanding exceptions and unfinished actions.
The user interface, or UI, is the set of controls and displays a person interacts with. Interaction design shapes how those elements behave over time. A screenshot may show the UI but reveal little about the experience of waiting, approving or recovering from an error.
Testing can begin with a simple question: what does the person expect will happen after pressing this control? Compare the answer with what the system actually does. A mismatch exposes a design problem even when the underlying software works as written.
In a clinic, “prepare a rescheduling message,” “approve this message” and “send it” should not feel like one indistinguishable conversational step. The user needs to see the proposed recipient, appointment details and current status, with a clear opportunity to correct them. If sending fails, the interface should preserve the draft and show that the patient has not been contacted. Interaction design connects the person’s expectation with what the software actually does, especially at the boundary between suggestion and action.
If you would like to find out more about AI and improve your leadership skills, consider 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