Prompt an image model to distinguish visible evidence, uncertainty, and next evidence needed.
Damage claim photo: the box is visible, the product is not. Jordan needs an AI-assisted read of the photo before approving the carrier claim. Which prompt sets up the safest vision review? Approve or reject this damage claim from the photo. This asks for judgment before evidence. It violates the visible-evidence principle because the model may infer product damage that the photo does not show. List only what is visibly damaged, what cannot be determined from this image, and the one additional photo needed before a claim decision. This is strongest because it separates observation, uncertainty, and next evidence. It…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in