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

“(Shrout and Fleiss 1979), indicating excellent agreement between raters (ICC = 0.91).”

“Generative Learning Theory conceptualizes learning as a constructive process in which learners organize information and relate it to prior knowledge to build coherent mental representations (Wittrock 1989).”

“Generative artificial intelligence (AI) is increasingly being incorporated into educational settings as a tool for learning support and instructional assistance (AlAli and Wardat 2024).”

“…studies in computer-based environments found that metacognitive prompts significantly improved both self-regulated learning activities (g = 0.50) and learning outcomes (g = 0.40), providing quantitative support for structured prompting as an effective instructional intervention (Guo 2022).”

“…in hypermedia and related digital settings report substantial variability in learners’ planning, monitoring, and evaluative behaviors, and many students demonstrate limited spontaneous use of self-regulatory strategies when instructional support is minimal (Azevedo and Cromley 2004; Bannert 2006).”

“…conversational agents has shown that AI-generated explanations can assist learners in articulating conceptual understanding, revisiting previously learned material, and reflecting on problem-solving strategies during independent study (Aluko et al. 2025; El Fathi et al. 2025; Mueller et al. 2025).”

“Research in technology-enhanced learning has extensively examined how learners regulate cognitive and metacognitive processes during self-directed study in computer-based environments (Persico and Steffens 2017; Sui, Yen and Chang 2024).”

“Core generative processes include selecting relevant information, organizing relationships among ideas, and integrating new material with existing schemas (Fiorella and Mayer 2015) (Lee, Lim and Grabowski 2008).”

“…evidence shows that while AI-assisted explanations can enhance the clarity and depth of student responses, unstructured use often results in superficial engagement, reinforcing the need for instructional scaffolding that emphasizes critical evaluation over answer acceptance (Uddin et al. 2024).”

“…have further reported that generative AI–supported learning environments can facilitate formative feedback and sustain learner engagement when students interact with conceptually complex or information-dense instructional content outside scheduled class time (Wu and Chiu 2025; Zhan et al. 2025).”

“Sequential analyses of inquiry-based learning tasks show that structured prompts influence learners’ regulatory patterns, with more successful learners demonstrating systematic monitoring and evaluation during prompted activities (Hsu, Wang and Zhang 2017).”

“Research in technology-enhanced learning has extensively examined how learners regulate cognitive and metacognitive processes during self-directed study in computer-based environments (Persico and Steffens 2017; Sui, Yen and Chang 2024).”

“Interactive graphic organizers are associated with deeper integrative processing and improved comprehension compared with static text or completed representations (Wang et al. 2021).”

“…conversational agents has shown that AI-generated explanations can assist learners in articulating conceptual understanding, revisiting previously learned material, and reflecting on problem-solving strategies during independent study (Aluko et al. 2025; El Fathi et al. 2025; Mueller et al. 2025).”

“Prior studies have explored applications such as on-demand explanation generation, interactive question answering, and adaptive support that respond to learners’ evolving needs during review activities (Bernasconi, Redavid and Ferilli 2025; Leon 2024; Yogi, Chowdary and Santhoshi 2024).”

“Prior studies have explored applications such as on-demand explanation generation, interactive question answering, and adaptive support that respond to learners’ evolving needs during review activities (Bernasconi, Redavid and Ferilli 2025; Leon 2024; Yogi, Chowdary and Santhoshi 2024).”

“…cognitive and metacognitive supports, commonly implemented through prompts that activate reflection and regulation, have been systematically investigated as instructional design mechanisms to promote deeper processing and improve learning outcomes (Hsu, Wang and Zhang 2017; Li et al. 2025).”

“Empirical studies indicate that LLMs are most effective in facilitating conceptual understanding and interpretive reasoning, while their performance remains limited in advanced analytical and calculation-intensive problem-solving tasks (Gesnot 2025).”

“Research on interaction design demonstrates that structured prompting strategies can significantly improve LLM performance on previously incorrect engineering exam questions, underscoring the role of students’ inquiry practices in mediating educational benefit (Genç 2025).”

“Building on prior work in which a course-aligned generative AI platform was developed for construction education contexts (Hou et al. 2026), the present study shifts the focus from system feasibility to the pedagogical design of student–AI interaction during self-directed learning.”