Drawing on Cognitive Load Theory and Self-regulation Theory, this study proposes a structural equation model that explains the factors and underlying mechanisms influencing performance in complex problem-solving tasks involving generative AI (GAI). Sp...
Drawing on Cognitive Load Theory and Self-regulation Theory, this study proposes a structural equation model that explains the factors and underlying mechanisms influencing performance in complex problem-solving tasks involving generative AI (GAI). Specifically, nine independent variables were classified into three cognitive load categories—Intrinsic, Extraneous, Germane. The variables contents creation, information restructuring, and information integration were hypothesized to reduce intrinsic cognitive load; the variables multimodality, context continuity, and human-likeness were expected to reduce extraneous cognitive load; and the variables knowledge conversion, critical thinking, and creative thinking were hypothesized to enhance germane cognitive load. In addition, low intrinsic cognitive load, low extraneous cognitive load, and high germane cognitive load were hypothesized to influence complex problem-solving performance, with these relationships moderated by variable metacognitive self-regulation.
A total of 331 valid responses were collected and analyzed using SmartPLS 4.0. The structural model evaluation showed that all independent variables had significant positive effects on their respective mediators. All hypothesized paths were supported, confirming the proposed cognitive pathways in the model. Furthermore, the paths from all three mediators (intrinsic cognitive load, extraneous cognitive load, and germane cognitive load) to performance were statistically significant, supporting the hypothesized cognitive mechanism that enhances user outcomes in complex problem-solving. The moderation analysis further revealed that metacognitive self-regulation significantly strengthened the effects of all three paths.
This study theoretically contributes by identifying the characteristics of GAI tools and demonstrating distinct cognitive pathways that influence performance in problem-solving situations. Practically, the findings offer valuable implications for designing task work flows utilizing GAI and developing GAI training programs managing cognitive load.