Generative AI and the balance of effort in group tasks: Appropriate, over-, and under-scaffolding

Generative artificial intelligence (Gen-AI) is increasingly used in higher education, yet its role in collaborative learning remains unclear. This study examines how students use Gen-AI during group lesson-planning tasks in an undergraduate education course. Drawing on Vygotsky’s sociocultural theory, we introduce the concept of “an effort space” and apply the categories of appropriate, over-, and under-scaffolding to Gen-AI use. The analysis of 75 students’ reflections, chat logs, and lesson plans identified three patterns: appropriate support that preserved learner effort, over-scaffolding through task outsourcing or free-riding that eliminates learner effort, and under-scaffolding through AI avoidance. These patterns show how Gen-AI reshapes collaborative learning.