Signal 1: Prevention of Adverse Interaction between Police Officers and Citizens using Machine Learning

Adverse interaction between police officers and the public has become a sensitive issue in recent times. Recurring events like fatal shootings of unarmed citizens and excessive use of force against cooperative citizens have led to multiple large-scale protests against the police force.

Most of these adverse events are confined to a small group of police officers which has led many police departments to adopt Early Intervention Systems (EISs). These EISs generally evaluate officers by setting threshold levels for certain factors like total adverse events and total work hours over a specific duration of time. This kind of approach is not sufficient to predict adverse event risk which depends on complex human and situational factors like personality traits of officers, past experiences of officers and neighborhood conditions relative to time.

Machine learning can be used to solve this problem more efficiently by using past police data for risk prediction of real time dispatcher feeds. An increase in predictive accuracy of adverse event risk will lead to fewer police shootings, increase confidence in the police and save legal costs.

Next steps include fetching publicly available police data and using that to train a machine learning model for risk prediction.

Source: https://dssg.uchicago.edu/wp-content/uploads/2016/04/identifying-police-officers-3.pdf