Current Research

Construction sites face persistently high safety risks, yet training in practice often remains instructor-centered and one-directional, which limits its ability to reflect diverse learners and changing site contexts. Worker and engineer proficiency can vary substantially; when the same content is delivered in the same way, novices may be overwhelmed while experienced personnel receive oversimplified guidance, reducing overall training effectiveness. In addition, limited opportunities for timely questioning and clarification on site can prevent learning from translating into concrete behavioral change. 

The key contribution of this study is to treat the idea that proficiency cues are embedded in the question itself as a central design principle, and to propose a personalized safety Q&A framework that adapts the depth and explanation strategy of answers accordingly. This approach avoids heavy reliance on static user profiling (e.g., pre-registered roles or experience), which is often missing or unreliable in real deployments, while still improving educational relevance. Beyond personalization as a mere change in tone, the framework emphasizes document-grounded answering so that learners can verify, trace, and trust the basis of the guidance. Overall, this research integrates proficiency-aware interactive learning with evidence-based trustworthiness, suggesting a scalable direction for practical, on-site construction safety education.