dax-med-relabeling-improved-hypothesis-testing

IN premise — summaries/2026/08/24/wiki-Confirmation_bias-chunk-2.md

Created 2026-08-25T02:58:45+00:00

Replacing 'fits/doesn't fit' labels with neutral 'DAX/MED' labels in Wason's task significantly improved participant hypothesis-testing performance, supporting the positive-test-strategy interpretation.

Summary

What looked like a deep logical flaw in how people test hypotheses was actually a framing trick: simply swapping out the word "fits" for a neutral code letter like "DAX" was enough to make participants reason much better. This matters because it means the system should treat classic "cognitive bias" findings as potentially sensitive to surface-level wording, rather than assuming they reflect stable reasoning errors.