This paper presents the curiosity erosion model (CEM), a conceptual framework that investigates the way large language models (LLMs) affect human curiosity—a vital catalyst for learning. Although LLMs provide immediate access to extensive information, their influence on curiosity has not been thoroughly examined. The CEM delineates four mechanisms through which LLMs may either promote or inhibit curiosity: information accessibility, depth of engagement, metacognitive awareness, and serendipitous discovery. Each mechanism encompasses testable hypotheses that analyze how interactions facilitated by AI modify exploratory motivation. Utilizing meta‐analytic data and empirical studies conducted in educational settings, the model uncovers bifurcated effects, demonstrating that while LLMs can enhance initial engagement, they may simultaneously pose a risk of diminishing independent inquiry, with results being significantly contingent upon the implementation methodology, learner proficiency, and educational context. By synthesizing perspectives from cognitive science, pedagogical theory, and human–AI interaction, the CEM offers a comprehensive framework for comprehending how LLMs transform learning experiences and presents actionable strategies for the development of AI systems that promote rather than hinder curiosity. These recommendations assist educators, designers, and developers in cultivating environments that bolster curiosity‐driven learning, nurturing adaptive and self‐directed learners both within traditional educational frameworks and beyond.
Ganuthula et al. (Thu,) studied this question.