TRACE: Testing the Reliability of AI in Consequential Evidence

AI is increasingly shaping the evidence behind decisions that affect us all. We test where it is reliable, where it is not, and how anyone can tell the difference.

TRACE is an interactive investigation into how AI changes the production, analysis and use of evidence. We explore this through a practical case study, using AI to analyse education debates across England, Scotland and the Republic of Ireland.

Follow the analysis from outputs to methods to governance: explore what AI found, see how data and methodological choices shaped the findings, and ask who decides whether AI-assisted evidence is good enough to use.

Why we built TRACE

TRACE does not offer a single score or final verdict. It shows why meaningful AI evaluation requires more than technical metrics. Those measures matter, but they must be combined with the knowledge of people who understand the system and the experience of those affected by the decisions the evidence informs.