Who governs the algorithm?

Such has been the pace of technological advances and application of AI tools, their use offers policing significant new opportunities to prevent and detect crime, to reduce harm and to safeguard communities.

The implementation of AI in policing is driven by changing criminal threats, the increase in big data- the sheer amount of information that the police must process, the exponential rise in demand for police services, coupled with pressures on resources and the current agenda for police reform that will increasingly focus on structures, governance and mission.

There are also considerable challenges in addressing low public trust and confidence.

In applying powerful new technologies, police leaders have a responsibility to apply proportionality and necessity within evidence-based practice that promotes accountability, ethics, and human rights.

Currently, there is no all-encompassing legislation covering the use of AI in the UK, and the government has indicated that each sector should develop its own governance and inspection frameworks. Nevertheless, AI in policing should expect increased public scrutiny and others including, the Home Office, HMICFRS, College of Policing and IOPC.

We wouldn’t expect the police to be given new legal powers without parliamentary scrutiny, so why expect AI to be any different? For police leaders, whilst AI tools, models and agents have the potential to deliver more effective, efficient, and economic policing services, there are substantial moral and ethical questions around its capability to deliver safer and more equitable policing services.

Therefore, achieving ‘AI safety’ in the application of policing, given the dangers of data bias, unintended consequences of use and risks for community cohesion are essential areas to navigate.

The recent introduction of Police.AI, and the investment of £140 million in AI technology over 3 years signals the ‘responsible development, piloting and scaling of artificial intelligence in policing.’

This includes the expansion of new Live Facial Recognition (LFR) systems. For police leaders at local and national levels, there are key considerations around governance. Firstly, the technology must work effectively and reliably- this is obvious- but it must also be seen to work and have a direct impact on tackling crime and disorder and making all communities safer.

LFR has proved controversial with problems of data bias, adversely affecting women and people with darker skin complexions. These are all areas where the technology has to improve.

Secondly there must be a human in the decision-making loop- with AI systems being readily interpretable and outputs being understood by operational officers- this means effective training, not just on the technology but on the highest professional and ethical standards required to apply them safely and fairly.

We cannot leave AI to make autonomous decisions in policing and must guard against humans retreating into ‘machine bias mode’- the over estimating or underestimating of AI capabilities- including ‘the computer says no’ perspective.

Thirdly, police leaders must apply a compatibility test to ensure proportionality and necessity in the use of AI tools, models and agents. This will be familiar language to officers and staff and aligns with the Human Rights Act.

Finally, there should be the democratisation of the use of AI for policing purposes, with the creation of independent advisory groups or ethics committees made up of lay citizens and experts to act as critical friends in how AI systems are developed and implemented. The West Midlands Police Data, Technology and Operational Ethics Panel provides a significant example of best practice in this area. ‘The Panel’s job is to advise the PCC and Chief Constable on data science projects being proposed by WMP’s Data Analytics Lab’ and provides the model for similar structures nationally.

There will be bumps along the way, including legal challenges and questions of police legitimacy in the use of AI.

However, the prize for getting this work right, means safer communities and that can only come through balancing the huge technological advances with the application of excellence in professional and ethical policing practice.

Ron Winch is a former Superintendent with West Midlands Police. He is currently Course Director for MSc Policing and Criminal Justice Leadership at Birmingham City University