Behavioral Safety Performance Assessment in Construction: a Data-driven Approach in Virtual Reality

The construction industry suffers from high incidence rates compared to other sectors.Lean and safety training is one effort to increase work productivity while preventing accidents, but the effectiveness of various existing training methods relies on intention-based evaluations.These can fail to capture behavioral lean and safety performance.This paper presents a novel Virtual Reality (VR)-based assessment method for quantifying (lean and) safety performance in construction, addressing the critical gap between intention and behavior.By simulating realworld hazards, a simplified VR learning environment allows for objective, data-driven assessment of workers' behavior using metrics that capture, for example, the interaction with poor construction site layout and hazards.This study created a virtual environment and tested the capabilities of objectively assessing aspects of lean and safety performance in an experiment with 60 subjects.Splitting them into groups of experts and novices and exposing them to different training methods allowed a qualitative discussion on the implications of such an assessment method.The results demonstrate its applicability for a data-driven metric to assess lean and safety behavior, given that the environment includes relevant hazards and represents lifelike tasks.Construct validity requires further considerations when using this concept for virtual performance assessment.A more nuanced metric could further strengthen the study design.This study provides an explorative framework for further research to investigate the effectiveness of training methods based on behavior and overcoming the intention-behavior gap.