The Impact of AI on Risk Decision-Making

• In Event: Technology and control in community sanctions

Abstract

Actuarial risk assessments are a core practice in criminal justice. Critical thinking and levels of uncertainty are crucial to professionals’ reasoning in this context, yet concerns remain about practitioners a) engagement with the uncertainty that the risk assessment and management of complex cases requires and engenders b) exercise of appropriate reflexivity and discretion, and c) use of critical thinking when assessing risk (Heron 2020; 2023a&b). The exponential growth of interest in Artificial Intelligence presents opportunities, challenges and dilemmas to practitioners across a range of professions. There remains, however little evidence about the value and impact of Machine Learning, Natural Language Processing AI in justice contexts. This paper reports on the findings from an experimental study with 20 justice social workers in Scotland, exploring the extent to which the use of ChatGPT enhances or detracts from uncertainty when completing complex risk assessments; and, whether it supports or limits practitioners critical thinking when examining risk of serious harm, and to what effect.

Authors

• Beth Weaver, University of Strathclyde

• Fern Gillon, University of Strathclyde

• Gavin Heron, University of Strathclyde