<![CDATA[Designing mathematically worthwhile problems requires teachers to coordinate clear language, cognitive demand, openness, and contextual relevance. Generative artificial intelligence (GenAI) may provide rapid feedback, but little is known about which feedback dimensions preservice teachers actually enact in revised problems. This exploratory descriptive study examined 48 coded feedback units linked to the final problem designs of 10 sixth-semester preservice mathematics teachers. Each unit was classified by feedback aspect, i.e. clarity and language, cognitive depth, open-endedness, or contextuality, and by adaptation level from ignored (AL-0) to significant (AL-3). Descriptive analysis showed that 40 of 48 units (83.3%) reached AL-3; the overall mean adaptation score was 2.63 on a 0–3 scale. Clarity and language had the highest mean adaptation (M = 2.73; 90.9% AL-3), followed by open-endedness (M = 2.70; 90.0%), cognitive depth (M = 2.60; 80.0%), and contextuality (M = 2.50; 75.0%). The findings indicate a pattern of stronger enactment for concrete and locally actionable suggestions. These findings do not imply that GenAI caused improvement in the proposed problem. The study suggests that the used of GenAI feedback should be paired with explicit checks for mathematical validity, intended reasoning, authentic context, and human accountability.]]>
Copyrights © 2026
