A new resource is available for forces who wish to use machine learning (ML) in their work.

The Machine Learning Guide for Policing is an online resource that provides guidance to those considering embarking on using ML models to address problems within policing.  ML is a form of artificial intelligence (AI) that can analyse and identify patterns in large amounts of data. Many of these patterns might take too long to identify and are often impossible for humans to identify at all. ML could free up the manpower used for pattern-identification, making for more efficient use of staff and officer time.

In 2023, the Police Science and Technology Analysis and Research (STAR) Fund provided crucial funding for this project in the West Midlands Police (WMP), partnered with a group of academics. Titled: “The prioritisation of potentially harmful subjects and the estimation of future risk”, it sought in part to develop a framework to assist policing in implementing ML in their work.

In support of the strategy to end violence against women and girls, WMP are developing a tool to predict which stalking and harassment offenders are more likely to go on to commit high harm offences, in order to assist officers to prioritise their investigative workload.  At the same time, a group of academics (Middlesex University, Bournemouth University and the University of Suffolk) developed the Rational Unification Development and Implementation (RUDI) framework for policing which encapsulates best practice when developing ML algorithms to improve: transparency, justifiability, lawfulness and accountability.

This guide is an example of recent efforts within policing to develop responsible and transparent AI. Another example is the Covenant for Using Artificial Intelligence (AI) in Policing, a set of principles which all forces have agreed will define how AI is used in its business.