Paper Type
Complete
Abstract
Generative AI (GenAI) is increasingly used for learning, yet its fluent explanations may bias metacognitive monitoring and learning outcomes. Drawing on metacognitive monitoring and generative learning frameworks, we tested whether a brief Self-explanation Task (SeT) reduces the Illusion of Knowledge (IoK) and improves performance. In a between-subjects experiment, mechanical engineering and computer-science students learned wind-turbine principles via (1) an 18-page PDF, (2) GenAI dialogue, or (3) SeT followed by GenAI. IoK was defined as the discrepancy between post-learning judgments and test scores (memorization, transfer). IoK effects were weak (trend-level for memorization; null for transfer). The clearest result was an unexpected interaction: SeT before GenAI impaired learning, reducing transfer relative to PDF and GenAI-only, especially for low-familiarity learners. These findings show that GenAI effects are nuanced and that self-explanation may hinder learning when prior knowledge is limited.
Paper Number
1423
Recommended Citation
Lepennetier, Sophie; Fleury, Sylvain; Lepont, Virginie; Hamdani, Fatima; and MEYRUEIS, Vincent, "Generative AI-Assisted Learning: An Unexpected Detrimental Interaction with Self-Explanation" (2026). AMCIS 2026 Proceedings. 9.
https://aisel.aisnet.org/amcis2026/sig_ed/sig_ed/9
Generative AI-Assisted Learning: An Unexpected Detrimental Interaction with Self-Explanation
Generative AI (GenAI) is increasingly used for learning, yet its fluent explanations may bias metacognitive monitoring and learning outcomes. Drawing on metacognitive monitoring and generative learning frameworks, we tested whether a brief Self-explanation Task (SeT) reduces the Illusion of Knowledge (IoK) and improves performance. In a between-subjects experiment, mechanical engineering and computer-science students learned wind-turbine principles via (1) an 18-page PDF, (2) GenAI dialogue, or (3) SeT followed by GenAI. IoK was defined as the discrepancy between post-learning judgments and test scores (memorization, transfer). IoK effects were weak (trend-level for memorization; null for transfer). The clearest result was an unexpected interaction: SeT before GenAI impaired learning, reducing transfer relative to PDF and GenAI-only, especially for low-familiarity learners. These findings show that GenAI effects are nuanced and that self-explanation may hinder learning when prior knowledge is limited.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
SIG ED