Answers are grounded in this paper and cite their sources.
Objective This study examined how GPT-assisted formative feedback could be integrated into classroom inquiry on linear functions. Methods A gamified AlgeoMath module combined rule-based graph checks, staged language-model hints, and rubric-based reports. The pilot involved 49 students in two convenience-sampled intact middle school classes. Analyses integrated submission logs, module scores, immediate pre-post surveys, expert review, and human-AI agreement. Results Task completion was near complete. Among initially incorrect task instances followed by a second submission, 57.5% were corrected on that submission. Module profiles highlighted covariation and mathematical justification as areas needing further support. Six of eight survey items increased after Holm correction in 48 matched pairs. Expert review primarily identified mismatches between feedback content and its intended scaffold level. Automated rubric categories agreed with each of two educators on 69% of 100 sampled responses. Conclusions The study connects classroom implementation with evidence about students' response patterns and feedback quality. It identifies scaffold calibration and the interpretation of partial reasoning as priorities for teacher-supervised AI feedback in linear function inquiry and provides a basis for evaluating learning outcomes in subsequent controlled studies.
No takes yet. Share an insight, caveat, or question.
Oh et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: