Exploring the Agency of Visual Feedback in Generative AI-Integrated Mathematics Learning Environments: Focusing on the Design and Implementation of a Boxplot Task

Abstract

From the perspective of inclusive materialism, this study conceptualizes the agency of visual feedback in generative AI-integrated mathematics learning environments and illustrates it through the design and implementation of a generative AI-integrated boxplot task. We conceptualize 'agential visual feedback' as the phase in which visual output—produced by generating, transforming, moving, or deleting mathematical objects in response to a learner's manipulation—engages the learner's attention and thinking and manifests material agency within the intra-action of a particular learning assemblage. Analyzing the case of two middle school students who undertook the task as part of an after-school class, with a focus on assemblage and gesture-diagramming, we found that the macro-perception-centered assemblage, in which they placed points by directly matching the positions of the five-number summary, was disrupted by a material resistance: through intra-action with the AI's visual output, it no longer operated smoothly in its previous manner. A micro-perception attending to the quartile intervals and the number of data points then emerged, suggesting that mathematical meaning was provisionally actualized toward the distribution structure as a whole rather than the summary positions alone. Through a single case, this study illustrates the phase in which AI visual output is relationally constituted as agential visual feedback, and proposes the theoretical possibility of interpreting mathematical meaning-making in generative AI-integrated mathematics learning within the entanglement of human and nonhuman actants.