Prompt Engineering - How to Use Roles, Context, and Constraints to Get Precise Medical Output
There is a concept in AI interaction that experienced users learn quickly and that most beginners never discover: the output you get from an AI is shaped as much by the framing of the conversation as by the specific question you ask. Two clinicians can ask about the same clinical topic and get very different responses, not because the AI is random, but because they set up the conversation differently.
This course covers three of the most powerful levers in that setup: roles, context, and constraints. Used deliberately, these three elements can transform AI from a generic information generator into something that feels much more like a specialist colleague who understands your specific situation.
What roles do and why they work
When you tell an AI to take on a specific role, you are not playing a game or using a gimmick. You are giving the model information about the vocabulary, the knowledge domain, the level of technical sophistication, and the communication style that is appropriate for the task. A model told to respond as a maternal-fetal medicine specialist preparing a teaching case for third-year residents will use different language, make different assumptions about baseline knowledge, and organize information differently than a model given no role instruction at all.
Roles also implicitly set expectations about accuracy and rigor. A model playing the role of a specialist writing for a peer-reviewed audience will be more careful about evidence qualification than one responding to a casual question. This is not magic. It is the model responding to the contextual information you have provided.
Useful roles for clinical ObGyn tasks include: a maternal-fetal medicine specialist, a patient education specialist, a clinical evidence reviewer, a medical ethics consultant, a senior ObGyn attending reviewing a resident’s work, a plain-language medical writer, and a clinical guidelines analyst.
