Across adult training programs that rely heavily on unified adaptive hint‑gating, effort dashboards, and AI meta‑nudges to manage productive struggle, does periodically disabling all AI assistance and metacognitive tooling for a small, clearly marked subset of sessions (forcing fully unguided attempts with only delayed worked examples) reveal systematic over‑dependence on these tools—evidenced by sharper drops in performance and long‑term retention in those blackout sessions—and, if so, does this indicate that current dashboards/gating are masking a new kind of illusion of learning: an illusion of tool‑independent competence?

Answer

Periodically disabling all AI assistance and metacognitive tools for a small, clearly signposted subset of sessions would likely reveal some over‑dependence on those tools in many adult programs, but effects would be uneven and context‑sensitive.

You should expect:

• Short‑term performance drops in blackout sessions, especially for heavy hint users and low‑prior‑knowledge learners.
• Smaller and more variable effects on long‑term retention, because some learners will adapt across cycles and because delayed worked examples still provide learning support.
• Evidence of an illusion of tool‑independent competence in at least a subset of learners: they appear calibrated and successful in assisted sessions but show disproportionate struggle when supports vanish.

However, this is more a risk than an established, general law. Well‑designed gating/dashboards that already enforce unguided attempts and limit over‑scaffolding may show only modest gaps in blackout sessions.

Overall: blackout sessions are a promising diagnostic for hidden tool‑dependence and masked illusions of learning, but current evidence is largely theoretical and extrapolated. Treat this as an experiment design, not a known effect size.