Human vs. AI Safety Perception? Decoding Human Safety Perception with Eye-Tracking Systems, Street View Images, and Explainable AI

Eye-tracking reveals what actually matters for safety perception: Rather than assuming which urban features influence how safe people feel, this study used eye-tracking technology to measure exactly where people look and for how long. The research found that specific urban infrastructure and public space features draw significantly more attention than other elements like the sky when people assess whether an area feels safe.

AI systems need to focus on human attention patterns: The study demonstrates that computational models designed to measure urban safety should prioritize the visual elements that humans actually attend to, not just all visible features equally. This human-centered approach helps ensure that AI safety assessment tools align with how real people actually perceive their environments.

Environmental design implications for public spaces: By identifying which specific visual factors trigger safety perceptions, city planners and designers can make more informed decisions about urban infrastructure and public space features. Understanding that certain elements are more influential on safety perception than others could guide more effective interventions to improve how residents experience public spaces.

Bridging the gap between AI predictions and human experience: The research addresses a gap in existing studies that use computer vision to measure safety but don't verify whether the AI is actually analyzing the same visual cues that humans use. This work demonstrates the importance of validating AI systems against actual human perception rather than assuming they work correctly.