Pre-registered hypotheses. Blind scoring. Whatever the results.

Each experiment is designed before running, scored without knowing the outcome class, and reported in full: accuracy figures, effect sizes, limitations, and replication notes.