Towards Evaluating Student Agency in AI-Mediated Learning

AI-mediated learning systems increasingly shape how students act, choose, and engage within digital classrooms, yet designers lack concrete ways to evaluate how these systems support or constrain student agency. This work-in-progress paper proposes an initial, design-oriented framing of student agency for interactive learning environments and introduces a prototype methods suite for analysing it in practice. Synthesising insights from HCI, the learning sciences, and child-rights guidance, we conceptualise agency as an interactional and multidimensional phenomenon produced through system affordances, decision policies, feedback structures, and orchestration behaviours. We outline early methods (i.e., log-data probes, scenario-based interface walkthroughs, and mixed-methods classroom studies) that reveal how students interpret system cues, negotiate control, and enact agency socially and developmentally. Our aim is to offer designers an emerging evaluative lens that identifies when AI systems narrow or expand students’ room to act, and to support the creation of more intentional, agency-sensitive learning technologies.