What should an AI helm computer say when it isn’t sure?

I’m building d3kOS, an open-source AI helm computer for the Raspberry Pi. It reads your engine data over NMEA 2000 and explains a reading, such as a rising coolant temperature, using your own manuals and sensor history.

The hardest design question has been what it should do when it isn’t sure. Guessing confidently on a boat is worse than saying nothing. Right now d3kOS cites your manual and says plainly when it can’t find an answer, and the safety features like SafeHelm are advisory only.

For boaters: when your instruments alarm, would you want a short answer, a step-by-step checklist, or a “call a mechanic” flag?

For developers: how would you handle uncertainty in an AI that people may rely on at the helm?

I’m in the thread all day.