Research Topic 9: Transforming Virtual Learning for Construction Safety Using Large Language Models

Research Description

Online, self-paced safety training has become increasingly common in the construction industry. Yet, as anyone who has taken an online driver’s education course knows, these modules—while incredibly convenient—often lack the interactive dynamics that makes in-person learning effective. Through this research project, we will investigate ways to develop and embed a virtual learning tutor powered by large language models (LLMs) within these digital training modules. The objective is to simulate the expertise and responsiveness of a seasoned instructor, thereby providing a dynamic, personalized learning experience that will ultimately contribute to enhanced safety on construction sites.

REU Research Plan: 

During the summer months, REU students will work with graduate student mentors and faculty advisors to develop and evaluate an AI-powered safety tutor that supports interactive learning within virtual environments. The students’ involvement will include the following:

• Review scholarly literature on adult learning principles and engagement strategies;
• Apply these research-based insights to instruction-tune and fine-tune LLMs for optimized learning engagement;
• Assess the tool’s efficacy against its intended objectives; and
• Contribute to the writing of a conference/journal paper to disseminate the project

Keywords: adult learning, safety training, human-computer interaction, artificial intelligence, language models, civil engineering, construction management

Required Skills: basic knowledge of statistics and fundamental programming concepts.

Undergraduate Degrees: Computer Science, Civil Engineering, Building Construction, Construction Science/Management.

Faculty Advisor: Dr. Zhenyu Zhang