Safer Places

Can we design and create a neighborhood that is safe and makes residents feel safe?

Our Safer Places project leverages large-scale street view images and cutting-edge AI (Vision Language model, Generative AI, and Deep Learning) methods to measure human subjective place perceptions and examine how these perceptions relate to crime and urban land use. Our assumption refers to the fact that people’s perception of street view images can serve as a proxy for how they experience the built environment. We collect human subjective safety ratings, then train AI algorithms to “see” streets and produce scalable measures of perceived safety that can inform urban planning and design. This allows us to identify which built-environment features, such as greenery, buildings, or roads, are associated with feeling safe or unsafe.

Beyond measuring perceived safety, we focus on Responsible and Human-centered use of GeoAI. In particular, we study perception bias by comparing the safety perceptions using GeoAI with: (1) real-world criminal activities (Paper), and (2) safety perceptions collected using traditional survey-based measures (Paper). By mapping these mismatches across neighborhoods, we show how perception bias may shape our understanding of urban safety.

This work is a close collaboration among the MIT Senseable City Lab, the City Planning Office in Stockholm, KTH Royal Institute of Technology, the Police Department in Columbia, South Carolina, and the University of South Carolina. Together, these partnerships demonstrate what Human-centered GeoAI can look like in practice: using AI to better understand human experience in place, and then integrating those insights back into real-world policies.

References

Kang, Y., Abraham, J., Ceccato, V., Duarte, F., Gao, S., Ljungqvist, L., Zhang, F., & Näsman, P. (2023). Assessing differences in safety perceptions using GeoAI and survey across neighbourhoods in Stockholm, Sweden. Landscape and Urban Planning, 231, 104768.

Zhang, F., Fan, Z., Kang, Y., Hu, Y., & Ratti, C. (2021). “Perception bias”: Deciphering a mismatch between urban crime and perception of safety. Landscape and Urban Planning, 207, 104003.

Abraham, J., Kang, Y., Ceccato, V., Näsman, P., Duarte, F., Gao, S., Ljungqvist, L., Zhang, F., & Ratti, C. (2025). Crime and visually perceived safety of the built environment: A deep learning approach. Annals of the American Association of Geographers, 115(7), 1613–1633.

Ceccato, V., Kang, Y., Abraham, J., Näsman, P., Duarte, F., Gao, S., Ljungqvist, L., Zhang, F., & Ratti, C. (2025). What makes a place safe? Assessing AI-generated safety perception scores using Stockholm’s street view images. British Journal of Criminology.

Kang, Y., Chen, J., Liu, L., Sharma, K., Mazzarello, M., Mora, S., Duarte, F., & Ratti, C. (2026). Decoding human safety perception with eye-tracking systems, street view images, and explainable AI. Computers, Environment and Urban Systems.