EyeGPT: A Cognitive Load-Adaptive GenAI Assistant with Eye Tracking for Programming Education

Generative AI (GenAI) is increasingly used in education, yet it often fails to account for learners’ cognitive load, potentially leading to overload. To address this challenge, we propose EyeGPT, an eye-tracking-adaptive zero shot GenAI assistant that dynamically adjusts its responses based on users’ cognitive load. We initially tested its feasibility with five participants. To explore its impact, we propose a controlled lab study where participants without programming experience complete two programming tasks under different AI conditions: (1) an adaptive GenAI assistant using eye-tracking data and (2) a conventional GenAI assistant. By integrating biosignals into AI-driven learning, this study explores how real-time adaptations can enhance learning efficiency and mitigate cognitive strain, offering insights for the development of intelligent, user-centered educational AI systems.