Skip to content

Past event

Use of AI in Sentencing & Management of Offenders

27 February 2023

This half day seminar was jointly hosted by the Alan Turing Institute, Northumbria University’s Centre for Evidence and Criminal Justice Studies and the Sentencing Academy.
 
 
 

About this event:

 
This seminar explored the current and proposed future uses of AI in sentencing and more broadly the management of offenders. The event was intended to present different perspectives on the use of AI in this context and raise awareness of, and provoke engagement with, the ongoing debates in this area.
 
The seminar brought together academics, policymakers, practitioners and computer scientists. The first session discussed the use of AI in sentencing from a technical perspective, including speakers from the Alan Turing Institute, Northumbria University and the Centre for Data Ethics and Innovation. The second session explored the issues for practice arising from the use of AI in sentencing and included speakers from legal practice and the Sentencing Academy.

Executive Summary

  • The implementation of AI into sentencing is complex due to non-universal definitions, imperfect or non-uniform data, an overload of information of varying value, and the risk of mistakes reducing the legitimacy of sentencing on which the judicial system relies.
  • AI can provide some benefits, such as mitigating against possible judicial bias or improving the speed of access to justice in the case of nations with sentencing guidelines. However, these benefits are often offset by sacrificing the right to speak in one’s own defence at sentencing hearings, and risks such as unexplainable decision-making with inadequate reasoning given.
  • AI may overlook the fact that the exact same offence committed against two different victims will have different impacts, and this is something that currently requires a human’s lived experience to analyse the harm caused by an offence.
  • AI is built, trained, and validated on existing data which can risk creating a false sense of objectivity, and subsequently over-trust in outputs.
  • AI may have a place in a research or advisory capacity when it comes to sentencing and help to detect and/or reduce bias or discrimination in sentencing.
  • AI may be particularly useful to the judiciary in the tracking and analysis of their sentencing habits, helping them to self-assess their own potential hidden biases and correct them with training.
  • AI has a role to play within sentencing, but that role must be chosen carefully and only where it can make a measurable and safe impact. It should not be used for its own sake, or for the appearance of modernisation.