CONCEPT
AI Shaming
The 2025 empirical finding that workers suppress AI use when it is visible to evaluators — accepting measurable performance costs to avoid the shame of being seen needing the machine.
AI shaming is the empirically documented pattern — first formalized in 2025 research — by which workers systematically reduce their reliance on AI recommendations when usage is visible to evaluators, even at measurable performance costs. The observed accuracy decline is approximately 3.4%, with one in four potential successful human-AI collaborations lost to visibility concerns. The workers are not making rational calculations about output quality; they are managing shame. Visible AI reliance is perceived to convey weakness in judgment, lack of confidence, or insufficiency. Workers would rather perform worse than be seen needing the machine. This is shame's signature pattern — choosing self-protection over effectiveness, appearance over reality, the performance of competence over the practice of competence.
In The You On AI Field Guide
The phenomenon matters because it reveals the gap between what AI could produce and what organizational conditions permit it to produce. A tool that improves decision quality when used privately is actively suppressed when use becomes observable, with measurable consequences for downstream