Social workers are constantly required to make complex and consequential judgements – about risk, harm, parenting, and what is likely to happen next. These judgements are morally, legally, and professionally significant. While many factors contribute to good professional judgement, one key aspect is accuracy: when a social worker assesses that an event or outcome is likely or unlikely, how often are they right?

In this webinar, David Wilkins presents findings from an exploratory study comparing the judgemental accuracy of 581 child and family social workers with that of ChatGPT. Using 12 anonymised referrals and a structured set of forecasting questions, the study examined how accurately both groups predicted subsequent outcomes, using Brier scores as the measure of accuracy. Surprisingly, ChatGPT outperformed the average social worker on 11 of the 12 referrals – though the difference was not statistically significant.

This raises important questions. What does this tell us about the nature of human judgement in social work? Should accuracy be given greater weight in our thinking about professional decision-making? And what are the ethical implications of using AI tools to support or supplement human judgement? Join us to reflect on these questions – and on what the future might hold for child protection practice.