Start with a precise claim
Ask which distance, angle, direction or contact time the claim predicts. “The world looks flat” is not yet a numerical prediction; a horizon angle at one specified camera height is. Agree the units and the reference direction before changing the inputs.
Allow the same test to challenge both models
Write both results before observing. Avoid granting one explanation adjustable parameters after the measurement while holding the other fixed. If both fit within uncertainty, choose another observation with greater separation. The three-shadow tool illustrates how an additional site can test the shape of a relationship.
Why a pile of facts may not settle a conversation
Beliefs can connect to trust, group belonging and the sources people consider credible. Experiments on perceived in-group beliefs have found an association with a person’s own conspiracy beliefs. That result describes a studied pattern; it does not establish why any particular person holds a view.
Make the next step practical
Keep a source file or raw reading both people can inspect. Ask what each would expect if their explanation were wrong. Let an unresolved observation remain open instead of turning it into a contest over intelligence or identity.