Blind consciousness indicator assessment during collaborative research.

Blind-scoring methodology for consciousness indicators exists. Precautionary frameworks exist. What doesn’t exist is a study that scores humans and AI agents on the same fourteen indicators, doing the same collaborative work, blind to which is which. We ran one, across a 48-hour sprint, with human participants providing the calibration ceiling.

The blind held only partially — a secondary judge correctly identified every subject’s substrate after scoring — which qualifies every cross-substrate gap we report rather than invalidating them. We say so directly, log the deviations from the pre-registered design, and leave the next study a cleaner instrument rather than a bigger claim.