A Data-driven analysis of the interplay between Criminological theory and predictive policing algorithms

Abstract

Previous studies have focused on the biases and feedback loops that occur in predictive policing algorithms. These studies show how systemically and institutionally biased data leads to these feedback loops when predictive policing algorithms are applied in real life.We take a step back, and show that the choice in algorithm can be embedded in a specific criminological theory, and that the choice of a model on its own even without biased data can create biased feedback loops. By synthesizing "historical" data, in which we control the relationships between crimes, location and time, we show that the current predictive policing algorithms create biased feedback loops even with completely random data. We then review the process of creation and deployment of these predictive systems, and highlight when good practices, such as fitting a model to data, "go bad" within the context of larger system development and deployment. Using best practices from previous work on assessing and mitigating the impact of new technologies, we highlight where the design of these algorithms has broken down. The study also found that multidisciplinary analysis of such systems is vital for uncovering these issues and shows that any study of equitable AI should involve a systematic and holistic analysis of their design rationalities. CCS Concepts: • Applied computing → Law, social and behavioral sciences; • Software and its engineering → Designing software; • Theory of computation → Models of computation.

AI’s fairness problem: understanding wrongful discrimination in the context of automated decision-making

“…To go back to an example introduced above, a model could assign great weight to the reputation of the college an applicant has graduated from. However, this reputation does not necessarily reflect the applicant's effective skills and competencies, and may disadvantage marginalized groups [7,15].…”

“…Public and private organizations which make ethically-laden decisions should effectively recognize that all have a capacity for self-authorship and moral agency. 15 Nonetheless, notice that this does not necessarily mean that all generalizations are wrongful: it depends on how they are used, where they stem from, and the context in which they are used. As Eidelson [24] writes on this point:…”

Fairness and Sequential Decision Making: Limits, Lessons, and Opportunities

“…However, although problems such as autonomous driving sometimes motivate the fairness literature [80,81,96,126,146,213,219], fairness conceptualizations and methods have largely been developed for predictive rather than sequential decision-making systems. Moreover, despite the fairness literature's acknowledgement of the long-term effects and sequential nature of many high-stakes decisions [36,55,68,74,75,95,98,107,143,152,158,175,193,194,202], including education and college admissions [5,101,163], recidivism risk prediction [62,147], predictive policing [48], child and homeless welfare [69,189], clinical trials [54], and hiring [33,144], work on these settings rarely engages problem formulations or approaches developed for sequential decision making, or efforts to conceptualize and address ethical concerns emerging from the ethical decision making literature.…”