Making Accurate Judgements in Child Welfare: Comparing ChatGPT With Qualified Social Workers

ABSTRACT This study compares the judgemental accuracy of child and family social workers ( n = 581) with ChatGPT, a generative AI model. Using 12 anonymized referrals, participants were asked predictive questions to evaluate accuracy through Brier scores. ChatGPT outperformed the average social worker on 11 of the 12 referrals, though the difference was not statistically significant. These findings highlight the potential and the limitations for AI to support decision‐making in social work while emphasising the need to address ethical concerns and AI's inadequacies for understanding complex human needs and social contexts. The study contributes to ongoing discussions on integrating AI into social work, advocating for a balanced approach that enhances effectiveness while preserving the profession's essential human elements.