Abstract:

This study aims to investigate how generative AI influences creative problem solving in Project-Based Learning (PBL) by examining the reconfiguration of the role structure of human cognitive activities during the problem-solving process. In recent years, the educational use of generative AI has expanded rapidly, and it has been suggested that AI can enhance learning efficiency and support creative activities by assisting with information exploration and idea generation. At the same time, concerns have been raised that generative AI may substitute for parts of the thinking process, potentially reducing learners’ cognitive engagement and leading to superficial problem solving. However, in complex problem-solving environments such as PBL, the causal mechanisms through which generative AI reshapes the distribution of cognitive roles between humans and AI—namely which cognitive activities are undertaken by AI and which remain with learners—have not yet been sufficiently clarified.

To address this issue, this study proposes a Cognitive Load Redistribution Model, which suggests that generative AI changes the distribution of cognitive load within the problem-solving process. Specifically, cognitive activities in problem solving are categorized into three types of cognitive load: exploration load, involved in searching for information and generating ideas; evaluation load, involved in assessing the relevance and validity of information or ideas; and integration load, involved in synthesizing multiple pieces of information or ideas to construct new solutions. Generative AI may reduce exploration load by supporting information generation and search processes. At the same time, because learners must evaluate the validity of AI-generated information and integrate multiple ideas into coherent solutions, evaluation load and integration load may increase. In this sense, generative AI does not simply reduce cognitive load; rather, it may reshape the distribution of cognitive load across different stages of problem solving.

Furthermore, this study focuses on the possibility that a decline in integration load may lead to shallow thinking. Shallow thinking is operationally defined through observable indicators such as insufficient integration of ideas, superficial evaluation of information, and weak self-explanation. Based on this definition, the study examines how different conditions of generative AI use influence the distribution of exploration, evaluation, and integration load, and how these changes affect the outcomes of creative problem solving, measured in terms of novelty, usefulness, and feasibility.

Preliminary observations in PBL-based courses indicate that the use of generative AI reduced the time required to complete project outcomes by approximately 60%. However, it remains unclear whether such efficiency gains correspond to improvements in the quality of creative problem solving or merely reflect the substitution of cognitive effort by AI. By clarifying how generative AI reconfigures the role structure of cognitive activities and the distribution of cognitive load in problem solving, this study aims to provide a theoretical foundation for designing PBL in the era of generative AI and to contribute to an extension of Cognitive Load Theory.

Keywords:

Generative AI, Cognitive Load Redistribution, Project-Based Learning (PBL).