Police Use of Machine Learning Algorithms Needs New Regulatory Framework

21 Sep 2018

There is an urgent need for regulatory and oversight mechanisms as machine learning tools are used for policing purposes, argues a new report published by the Royal United Services Institute and jointly authored with the University of Winchester.

Key points:

  • Machine learning algorithms to support police decision-making are in their infancy
  • UK Police forces are using or trialing machine learning technologies to support a variety of decision-making processes
  • Currently there are no clear guidelines on how police forces should trial experimental technologies in an operational environment. Experimental innovation is vital in this field, but can only proceed within the bounds of a clear policy framework
  • A coordinated and interdisciplinary approach will be needed to assess the impact of the introduction of machine learning algorithms into policing and to generate consistent principles for use
  • To guard against potential bias, inaccuracies and mishandling, machine learning algorithms will require constant ‘attention and vigilance’.

The RUSI report, published in partnership with the Centre for Information Rights, University of Winchester, describes how the police’s use of machine learning algorithms is in its infancy, and ‘the potential outcomes of these tools are still poorly understood’.

While machine learning algorithms are currently being used for limited policing purposes, such as supporting custody decisions, there is potential for the technology to do much more.  The lack of a regulatory and governance framework for its use is concerning.

The report also warns that the incorporation of machine learning into the criminal justice system may have ‘unintended or indirect consequences that are difficult to anticipate’.  The potential benefits of these tools are likewise yet to be fully established.

The report states that ‘it is clear that new technologies must be trialled in a controlled way in order to assess their effectiveness, before being rolled out in an operational environment where they are likely to have a significant impact on citizens’ lives. However, there is currently no clear framework in place for how the police should conduct such trials. What is needed going forward are clear codes of practice to enable police forces to trial new algorithmic tools in a controlled way in order to establish whether or not a certain tool is likely to improve effectiveness of a certain policing function.’

With concerns expressed about the application and accountability of artificial intelligence in wider society, the report argues that ‘it is essential that such experimental innovation is conducted within the bounds of a clear policy framework, and that there are sufficient regulatory and oversight mechanisms in place to ensure fair and legal use of technologies within a live policing environment.’

Recommendations

The report proposes eleven recommendations, including:

  • The Home Office should develop codes of practice outlining clear and appropriate constraints governing how police forces should trial predictive policing tools, including instructions concerning ‘experimentation’ in live operational environments focused upon fairness and proportionality. Such limited trials must then be comprehensively and independently evaluated before moving ahead with large-scale deployment.
  • The College of Policing should develop guidance within the Authorised Professional Practice with respect to the deployment of a machine learning algorithm within a decision-making process. This should include guidance on how police forces should present algorithmic predictions to those about whom the prediction is made. A clear process for resolving disagreements when professional judgement and the algorithm come to different conclusions should also be established within this guidance.
  • The inspection role of Her Majesty’s Inspectorate of Constabulary and Fire and Rescue Services (HMICFRS) should be expanded to include assessment of forces’ compliance with the above-mentioned new guidance. This will provide an accountability mechanism to ensure police forces are developing new tools in accordance with relevant legislation and ethical principles.
  • Officers may need to be equipped with a new, different skillset to effectively understand, deploy and interpret algorithmic tools (including the potential for bias), in combination with their professional expertise, and to make assessments of risk using an algorithmically generated forecast. The College of Policing should consider developing a course to equip officers with such a skillset.
  • Further research is needed to determine how the introduction of algorithmic tools influences police officer behaviour and the decision-making process as a whole.
  • A future regulatory framework should establish minimum standards of technical transparency for algorithms used to support police decision-making relating to individuals, which can be adapted for particular contexts and decision-making environments. As a minimum, machine learning (ML) algorithms should only be permitted for criminal justice purposes if it is possible to retroactively deconstruct the algorithm in order to assess which factors influenced the model’s predictions and how the prediction has been generated. This requirement should be included in all relevant public procurement agreements, along with a requirement for the supplier to be able to provide an expert witness who can give evidence concerning the algorithm’s operation if needed.

NOTES TO EDITORS

  1. This RUSI Whitehall Report is entitled: Machine Learning Algorithms and Police Decision-Making: Legal, Ethical and Regulatory Challenges
  2. Authors: Alexander Babuta is a Research Fellow in the National Security and Resilience studies group at RUSI, Marion Oswald is a Senior Fellow in Law, Director of the Centre for Information Rights at the University of Winchester and Christine Rinik is a Senior Lecturer in Law at the University of Winchester.
  3. The paper can be accessed at: https://RUSI.org/MachineLearningPolicing2018
  4. The Centre for Information Rights at the University of Winchester, based in the Department of Law, examines the overlap between information and privacy law and new technologies and methods of data analysis.
  5. The Royal United Services Institute for Defence and Security Studies (RUSI), is the world’s oldest independent defence and security think tank. Its mission is to inform, influence and enhance public debate on a safer and more stable world. RUSI is a research-led institute, producing independent, practical and innovative analysis to address today’s complex challenges.
  6. For further information, and interviews with the author, please contact Saqeb Mueen smueen@rusi.org /+44(0)7917 373 069
Back to media centre