Metacognitive feedback disrupted the tendency to repeat answer offloading across consecutive items without changing whether participants engaged with the assistant¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r3 · q2
Currency · newest study 2026 · checked against its sources 2026-10-20 (automated, 2 of 2 entries) · not reviewed by a person

Subclaims¶

q2 i? After offloading an item, participants offloaded the subsequent item in 69.7% of cases without feedback versus 55.9% with feedback; the interaction was significant (OR = 0.54). → Sebastian Maier 2026 q2 i? Feedback participants were similarly likely to use the assistant at least once (66.7% vs. 68.6%) and to request at least one complete answer (38.2% vs. 39.4%). → Sebastian Maier 2026 (2)

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

q2 · i? · causal · r3

Exploratory mixed-effects logistic regression on pairs of consecutive learning items, predicting offloading on item t+1 from offloading on item t, feedback, and their interaction. The article concludes feedback "disrupted the tendency to repeat answer offloading across consecutive items."

"After offloading an item, participants offloaded the subsequent item in69.7%of the cases without metacognitive feedback and55.9%with metacognitive feedback. The interaction was significant ( OR= 0.54, 95% CI[ 0.30, 0.97],𝑧=− 2.05,𝑝=. 040)"

Sebastian Maier 2026 (2)¶

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

q2 · i? · causal · r3

Descriptive comparison of the metacognitive feedback and AI-only conditions in the same experiment, reported in Section 5.3: feedback did not change whether participants engaged with the assistant at all, only how often they offloaded once engaged.

"Participants in the two conditions were similarly likely to use the assistant at least once (66.7% vs. 68.6%) and to request at least one complete answer (38.2% vs. 39.4%; Table 2)"

Discussion¶

Learner Variables¶

• Self-Regulation — outcome: instruction changes it

Related Claims¶

• Evidence suggests over-reliance on generative AI may negatively affect students' self-regulated learning behaviors, potentially harming retention and transfer to novel contexts — the general claim this finding bears on, with the studies behind it

• Feedback interventions show phase-specific effects: advice feedback reduces reminder bias, but ranking feedback alters beliefs without increasing offloading — 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
• The effort-based reward reduced how often participants used the assistant, whereas metacognitive feedback changed how they used it — 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 — a broader claim this one bears on