Make AI defer to your data, not its guess
Lesson 5 of 7 · Compliance & safe AI use
1 · Learn the move · Verified-data override
A model will fill a gap with a plausible number - a typical incubation time, a 'commonly reported' efficiency, an n - and state it as fact. The skill is telling it up front to use only the values you paste, mark every gap UNKNOWN, tag each figure with its source, and let your measured value win any conflict. You catch the invented parameter before it reaches a reviewer, not after.
You are my analysis assistant. Use ONLY the values and outputs I paste. If you need a number I did not give you, write UNKNOWN - never estimate from what is typical. Tag each figure with its source so I can trace it. [my run data]
2 · Your turn. You write the prompt
You ask for the results paragraph and it confidently reports 'a mean recovery of 87%' and 'n = 3' that you never gave it. Write a prompt that forces it onto your measured values and flags anything it cannot source.
Write the prompt for this scenario as given, or swap in a real one from your own work. The score works the same either way.
Remember: the AI sees only your prompt, not this page. If the situation isn't in your prompt, it doesn't exist.
Optional. These shape the output when you run your prompt below, not your score.