This effort, a cross-disciplinary collaboration between machine-learning experts and those with expertise in social theory and ethnography, aims to investigate the predictability of human behavior in the context of criminal infractions transpiring in the city of Chicago. The project also aims to investigate the social and ethical issues that arise from enhanced methods of prediction. The project uses as its database the detailed spatio-temporal logs of reported criminal infractions publicly available from the City of Chicago Data Portal. Sophisticated machine-learning algorithms will be deployed to automatically infer predictive generative models, with the ultimate objective of gaining deeper understanding of the temporal evolution of complex social structures. Participants aim to assemble the vast number of local models they infer into a predictive causality network that resides live online, ingesting data as it updates and projecting predicted incidence risks across the city. Any predictability that the project successfully distills from the data raises epistemological questions about human behavior, culpability, and the societal share of responsibility for individual actions. Key issues pertain to the social origin of biases, particularly in a culturally, economically, and ethnically diverse urban environment, and even what we mean by a “systemic bias.” Related questions concern how to identify, quantify, understand, and mitigate such issues. For example, does aggressive selective surveillance, including that driven by predictive algorithms, simply lead to higher opportunity of getting caught, and hence to higher crime rates, thereby bringing to fruition self-fulfilling prophecies of specific communities being more crime-prone? Or does unbiased opportunity to commit crime, resulting from larger societal forces, drive such dynamics? How can we design systems that mitigate crime while at the same time limiting these feedback loops? More broadly, the project will attempt to interpret predictive models/algorithms of normative human behavior and investigate the extent to which interpretation is important for crafting policy informed by such predictive analytics.