Structured AI prompts improve student reasoning by 2-3 points

A study by Xiaoyu Hou, Bo Xiao, Hexu Liu, and Shane Mueller (arXiv:2606.05509) investigates how instructional guidance shapes the effectiveness of generative AI in self-directed learning within construction engineering education. The authors developed a five-step prompting framework based on Generative Learning Theory (GLT) to structure student interactions with AI during review activities. A controlled experiment compared three conditions: slide-based learning, unprompted AI-supported learning, and prompted AI-supported learning. Learning performance was assessed via multiple-choice and open-ended tasks, and user experience was measured using the User Experience Questionnaire (UEQ).

The results show that the prompted AI condition significantly outperformed the other two on open-ended tasks requiring explanation and reasoning, with an improvement of approximately 2–3 points on an 18-point scale (p<0.01). However, no significant differences were observed in multiple-choice performance, and the unprompted AI condition remained comparable to traditional slide-based learning. These findings highlight that the effectiveness of AI-supported learning hinges on how interaction is structured, rather than simply providing access to AI. The proposed framework offers a practical basis for integrating learning science principles into generative AI systems, particularly in technical and engineering education contexts.