Access to an LLM assistant during practice did not significantly lower subsequent unaided test performance¶

Claim · All claims
Evidence · 1 study · causal r3 · q3
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? AI-only participants did not have significantly lower odds of answering a final test item correctly than control participants (OR = 0.80, p = .244). → Sebastian Maier 2026

Evidence¶

Sebastian Maier 2026¶

Sebastian Maier, Kai Schwabe, Manuel Schneider, and Stefan Feuerriegel. (2026). Designing Against Deskilling: Metacognitive Feedback Reduces Cognitive Offloading to LLM Assistants. arXiv. https://arxiv.org/abs/2609.20143 doi:10.48550/arxiv.2609.20143

q3 · i? · causal · r3

Confirmatory test of H1 comparing the AI-only condition (M = 0.62, SD = 0.34) with the no-AI control (M = 0.65, SD = 0.32) on the six unaided test items, using item-level mixed-effects logistic regression. The article reports no significant difference; equivalence was not tested.

"AI-only participants did not have significantly lower odds of answering a final test item correctly than control participants (𝑏=− 0.22,𝑆𝐸= 0.32,𝑂𝑅= 0.80, 95% CI[0.42, 1.51],𝑧=− 0.69,𝑝=. 244)"

Discussion¶

Related Claims¶

• How genAI is used, not who uses it, predicts whether independent performance declines or improves — the general claim this finding bears on, with the studies behind it
• Greater answer offloading during learning is associated with lower subsequent unaided performance (exploratory) — related
• In one preregistered experiment, metacognitive feedback on learners' own LLM use lowered the odds of answer offloading and raised the odds of a correct unaided answer on one-sided tests, with 95% intervals reaching no effect — related
• An effort-based reward did not significantly reduce answer offloading or improve unaided performance — related
• Unguarded answer-giving AI harmed unaided exam performance while a guarded version of the same model erased the harm (Bastani et al., 2025, as reported) — reports the opposite
• GenAI use weakens effort adjustment to incentives on subsequent unaided reasoning — related
• Prior genAI access shows only exploratory evidence of poorer subsequent unaided reasoning, within the no-incentive condition, and no preregistered main effect — related
• AI-assisted performance during skill-building does not transfer to subsequent unassisted performance — reports the opposite
• AI support can improve task performance while weakening users' metacognitive accuracy and leaving their own reasoning unimproved — related
• Participants deprived of relevant relational knowledge in Stage 1 take longer on the subsequent network-reasoning task — related