Fire safety in buildings could be improved by supporting occupants during emergency and evacuation phases, thanks to interactive systems that monitor the environmental risk levels and provide suggestions to people on how to properly behave. In this context, the direct, dynamic support to wayfinding process is essential, since it reduces wrong path selection, limits the possibility of interaction with fires and smoke, and thus contributes to the decrease in evacuation timings. Given the above, this study aim is to improve emergency evacuation effectiveness using this kind of solutions, by enhancing human-signage interaction through behavioural-based design, that is including human behaviours as main drivers for the definition of such technological systems. This study presents the design and Immersive Virtual Reality (IVR)-based evaluation of a novel dynamic Emergency Wayfinding System (EWS) aimed at enhancing evacuation outcomes in buildings. Leveraging a behavioural design approach, the system integrates environmental sensing and adaptive signaling to address user-specific cognitive and perceptual needs. A prototype was developed using modular components for integration into building infrastructure and tested within a Cave Automatic Virtual Environment (CAVE) setting. Two EWS configuration, passive and dynamic, were compared across sensory, cognitive, and functional affordances using a controlled virtual evacuation scenario involving 66 participants. Results indicate that while dynamic signage improves perceptual and cognitive consistency, it introduces interpretive uncertainty under specific conditions, potentially affecting exit choice accuracy. Participant feedback emphasized the importance of clearer dynamic cues, such as sequential lighting. Findings support the utility of IVR for early-stage evaluation of emergency systems and highlight pathways for optimizing dynamic EWS designs for real-world deployment.