The Role of Instructional Guidance in Generative AI-Assisted Learning: Empirical Evidence from Construction Engineering Education

A study demonstrates that structured instructional prompts significantly improve student learning outcomes when using generative AI for construction education, with prompted AI-assisted learning yielding 2-3 point improvements on reasoning tasks compared to unprompted AI use. The research introduces a five-step prompting framework based on learning theory, showing that AI effectiveness depends critically on how interaction is designed rather than AI capability alone.

This research addresses a fundamental challenge in educational technology: the gap between AI capability and actual learning outcomes. While generative AI systems possess substantial knowledge, students often interact with them inefficiently without proper guidance, limiting deeper cognitive engagement. The study's controlled design isolates the impact of instructional structure by comparing three conditions, revealing that unprompted AI interaction performs no better than traditional slide-based learning despite access to advanced tools. The framework grounded in Generative Learning Theory bridges this gap by systematizing how students prompt and interact with AI systems. The improvement concentrated on open-ended tasks requiring explanation and reasoning is particularly significant, as these competencies are central to professional engineering practice. This finding suggests that AI systems amplify learning when students engage in generative cognitive processes rather than passive information retrieval. For educational technology developers, the research indicates substantial market opportunity in packaging AI tools with pedagogically sound interaction designs rather than raw capability. For construction education specifically, AI-augmented learning could accelerate training efficiency and consistency across programs. The neutral performance on multiple-choice tasks implies that AI's primary value lies in supporting complex reasoning rather than factual recall, reshaping how institutions should integrate these tools into curricula. Future research should examine whether these principles transfer across disciplines and whether long-term retention benefits emerge from structured AI interaction.

• →Structured prompting frameworks improve AI-assisted learning performance on reasoning tasks by 11-17% compared to unprompted interaction
• →Unguided generative AI use provides no learning advantage over traditional instruction, highlighting design as critical to effectiveness
• →Instructional guidance effectiveness concentrates on complex cognitive tasks rather than factual recall
• →Learning science principles must be integrated into AI system design rather than relying on AI capability alone
• →Construction education serves as a testbed for scalable AI-augmented learning methodologies