What do I not know? How does AI affect critical thinking?

Effective and ethical risk assessment in criminal justice depends on practitioners’ critical thinking and professional curiosity, yet both are constrained by structural, relational and emotional pressures. The rapid proliferation of Large Language Models (LLMs) introduces new opportunities and challenges in probation practice. Drawing on data from 16 practitioners who participated in exploratory, interactive workshops, this study examined how practitioners engaged with ChatGPT when completing an actuarial risk assessment, and how these interactions influenced their critical thinking. Participants identified potential benefits, including reduced cognitive load through idea generation, support in articulating justifications, and assistance in structuring complex information. However, they also highlighted risks, particularly the potential for cognitive offloading, over‑reliance on AI outputs, and diminished critical scrutiny in contexts characterised by time pressure and uncertainty. The article discusses these preliminary findings in relation to wider debates on human–AI interaction and the conditions under which LLMs may augment, or undermine, professional judgement.