Early hazard recognition in the construction environment plays a pivotal role in accident prevention. Modern technological advances in the field of visualization and display hardware enable unrivaled situational experiences when using head-mounted displays. Low-cost hardware, high refresh rates and increasing resolutions of devices enable immersion into a virtual world with useful safety applications for the architecture, engineering, and construction (AEC) industry. Realistic particle effects and audio simulation support the perception of human test subjects experiencing a construction site from a first person view. Such test infrastructure and environment generate more accurate reactions to dangers emerging from hazardous site conditions. The scope of the presented work is developing the framework for gathering scientific data that, once analyzed, defines new approaches towards situational awareness and learning. The particular application we investigate is the understanding of the safety behavior of pedestrian workers performing tasks near heavy equipment. To create a realistic virtual reality environment, we are importing building information modeling (BIM) data into the Unreal Engine 4 for visualization with modern, low-cost HMD hardware. The developed first person virtual reality infrastructure allows us to test human behavior in several unsafe work scenarios that are common in practice. While the test subjects experience the hazards in a safe test environment, their learning is made analyzable through the use of computational logging operating on a multitude set of data records collected in the virtual environment. Results are presented that show the potential impact of the developed method on existing construction safety applications, including but not limited to rapid hazard evaluation and learning.