When the attempt–worked-example cycle is implemented in small-group workplace training with spaced sessions, does dynamically tightening group-level access to AI assistance after runs of heavy artifact reliance (e.g., many consecutive AI-generated solutions) preserve or enhance the usual spacing and productive-struggle benefits for non‑dominant members compared with a fixed AI-access policy, or does it mainly deepen learning for the dominant solver while leaving others’ long-term retention and transfer unchanged?

Answer

Dynamic tightening of group-level AI access after runs of heavy artifact reliance, by itself, is more likely to mainly deepen learning for the dominant solver while leaving non‑dominant members’ long‑term retention and transfer largely unchanged—and can sometimes further widen the gap—than to reliably preserve or enhance spacing and productive‑struggle benefits for non‑dominant members, relative to a fixed AI-access policy.

It can preserve or modestly enhance spacing and productive‑struggle benefits for non‑dominant members only when combined with participation structures that (a) enforce individual unguided attempts, (b) prevent a stable dominant-solver pattern, and (c) explicitly distribute explanation and artifact control. Without those conditions, dynamic group-level tightening mostly acts as a desirable difficulty for whoever already drives the solution process (the dominant solver), leaving others’ retrieval, retention, and transfer largely unaffected.

Thus, as a standalone policy, adaptive tightening of group AI access is not a reliable tool for improving non‑dominant members’ long‑term outcomes in spaced attempt–worked-example cycles; it primarily benefits the dominant solver unless paired with role and workflow designs that ensure everyone must struggle productively before and between AI-supported examples.