Understanding and Rewiring Cities

Bruno Lepri, Simone Centellegher, Marco De Nadai

Abstract: Nowadays, the ever increasing digitization of our societies is producing an unprecedented amount of data about human behavior. At the same time, advances in machine learning and complex systems enable us to build explanatory and/or predictive computational models of human behavior. Interestingly, these data and models can also be used to better understand the factors associated with specific neighborhoods’ outcomes such as vitality, safety perception, crime levels, innovation, segregation, traffic congestion, etc., and to design more efficient policymakers’ interventions. In particular, leveraging census data, mobile phone traces, information from OpenStreetMap, and street view images, we describe a set of studies where we (i) infer how vital and livable a city is; (ii) find urban appearance conditions that magnify and influence urban life; (iii) study the relationship of urban conditions with societal outcomes such as urban crime levels; and (iv) model the impact of pandemic shocks such as COVID-19 and related non-pharmaceutical interventions on human behavior.