Designing for "I don't know"
What a check-in form should do when a patient can't answer, and where an AI companion might fit.
"I don't know" is an answer, not an error.
Some check-in questions have answers patients simply don't know. "Does anyone in your immediate family have a history of diabetes?" is a classic. Not knowing an answer was one of the pain points I set out to research, and most forms handle it badly: a required field you can't get past.
When Claude and I wireframed the Check In App, this moment got its own screen, built around one idea: "I don't know" is an answer, not an error. We explored two directions:
- Help in the flow. "I don't know" is a normal option. Picking it opens a small help panel with three next steps: ask in the room, check your records, or skip for now. Reassuring, not punishing.
- An AI companion. The question dims and an assistant steps in, "here to help with tricky questions," offering the same three paths plus a chat box to ask directly.
The AI version is more helpful on paper, but it also asks for more trust. The PRD scoped the AI companion as V2, so for now it's a design concept, not a promise. Before it could ship, I'd need answers to some hard questions: what can it say without giving medical advice, and how does it show when it's unsure?
Both directions share the part I'm sure about: never trap someone on a question they can't answer.
What worked
What didn't
"I don't know" is an answer, not an error.
