Towards Social Fairness in Smart Policing: Leveraging Territorial, Racial, and Workload Fairness in the Police Districting Problem

Recent events (e.g., George Floyd protests) have shown the impact that inequality in policing can have on society. Thus, police operations should be planned and designed taking into account the interests of three main groups of directly affected stakeholders (i.e., general population, minorities, and police agents) to pursue fairness. Most models presented so far in the literature failed at this, optimizing cost efficiency or operational effectiveness instead while disregarding other social goals. In this paper, a Smart Policing model that produces operational patrolling districts and includes territorial, racial, and workload fairness criteria is proposed. The patrolling configurations are designed according to the territorial distribution of crime risk and population subgroups, while equalizing the total risk exposure across the districts, according to the preferences of a decision-maker. The model is formulated as a multi-objective mixed-integer program. Computational experiments on randomly generated data are used to empirically draw insights into the relationship between the fairness criteria considered. Finally, the model is tested and validated on a real-world dataset about the Central District of Madrid (Spain). Experiments show that the model identifies solutions that dominate the current patrolling configuration used.