When small-group workplace training uses the attempt–worked‑example cycle with an always‑on AI assistant, how does who controls the AI (e.g., rotating AI‑driver role vs. free control vs. assigning control to the least talkative member) interact with dominant-solver patterns to affect long‑term retention, transfer of skills, and social loafing, holding other structures like individual pre‑work and turn‑taking constant?

Answer

Control of the always-on AI mainly matters insofar as it amplifies or disrupts the dominant-solver pattern. With individual pre‑work and basic turn‑taking held constant:

• Rotating AI‑driver role (per task or per segment) is the safest default. It modestly improves retention and transfer for quieter members and reduces social loafing, because it guarantees everyone sometimes (a) decides when to call the AI, (b) formulates prompts, and (c) interprets AI output for the group. It prevents a single person from becoming the exclusive solver–AI gateway, so the dominant-solver pattern is weakened rather than reinforced.

• Free, unstructured AI control tends to let the dominant solver monopolize both problem solving and AI use. This reinforces the dominant-solver pattern, increasing social loafing and concentrating retention/transfer gains in the dominant member. Others still benefit somewhat from exposure and discussion, but with less retrieval and explanation, so their long‑term learning is smaller and more uneven.

• Assigning AI control to the least talkative member can reduce dominance and loafing if the group already has reasonable psychological safety and the facilitator supports that person (e.g., scaffolding prompts, backing their decisions). In that case, it pulls quieter members into higher‑level decisions and interpretation, improving their retention and transfer and slightly compressing gaps. But in low‑safety or highly hierarchical groups, this setup can backfire: the nominal driver defers to the dominant solver for every prompt, or is overruled in real time, so the pattern reverts to de facto dominant control plus extra anxiety for the quiet member.

Overall:

• Any scheme that concentrates AI control in the same person who already dominates reasoning (typical with free control) will worsen or maintain social loafing and sharpen learning disparities.
• Any scheme that systematically distributes AI control (rotating driver; occasionally assigning quieter members but with support) will modestly reduce social loafing and spread retention/transfer benefits more evenly.
• These effects are incremental, not transformative: with individual pre‑work and turn‑taking already in place, AI-control policies mainly fine‑tune who gets the deeper practice of deciding when and how to use the AI and making sense of its output.