AI and Transparency in Judicial Decision Making

Abstract Transparency is essential for maintaining the accountability of the judiciary and justice system as a whole. Yet, judicial reliance upon predictive artificial intelligence (AI) tools is not always compatible with this core value. Drawing upon semi-structured expert interviews with members of the judiciary and legal profession, case law and real-life examples, we consider four questions: why transparency matters for judicial decision making; what information judges must communicate to satisfy its demands; whether this information is accessible; and what we might do about any deficit. We argue that judicial reliance upon AI-generated behavioural predictions warrants a stringent transparency threshold, which demands access to the variables and formulae used to generate an output. At present, this threshold cannot easily be met in cases that involve trade secrecy and/or machine learning. Accordingly, we set out concrete steps to reconcile AI-informed judicial decision making with transparency, as a foundational aspect of justice and the rule of law.