Across adult training programs that rely on AI hints, effort dashboards, unified adaptive hint‑gating, and meta‑nudges to curb illusions of learning, does periodically replacing a fraction of AI‑assisted sessions with “no‑AI, no‑dashboard” peer diagnostic drills—where small groups must (a) predict which prior items they could still solve unaided and (b) then test those predictions without any external artifacts—reveal and reduce a distinct illusion of tool‑independent competence more effectively than the current solo‑focused blackout designs, or do these peer drills mainly surface social comparison and defensiveness without improving long‑term retention or transfer?
ai-learning-overreliance | Updated at
Answer
Peer diagnostic blackout drills are plausibly better than solo blackouts at revealing and slightly reducing illusions of tool‑independent competence, but only when groups are small, tasks are tightly structured, and psychological safety is at least moderate. In low‑safety or loosely run groups, these drills are likely to add social comparison and defensiveness with little or no gain in retention or transfer. Treat them as an experimental diagnostic plus occasional calibration booster, not a default replacement for solo blackout sessions.