What will it mean for justice when AI-based evidence becomes mainstream in our criminal courtrooms? Will juries understand it or mistrust it? Will lawyers know how to test and challenge it? Will it result in fair decisions or miscarriages of justice?
Mock trials are used by researchers to simulate courtroom proceedings, and we have incorporated AI evidence into a fictional albeit realistic criminal case to examine how counsel, the judge, the expert witness, and the jurors make sense of it. Though imagined as a futuristic scenario, it may not be so far away!
Getting the materials and process right took many months of preparation. Beginning in March 2025, our first step was to scope how previous mock trial studies had been designed, which involved a literature review and consultations with other researchers to understand how they had approached the task. Alongside this early work, the team carried out a substantial exercise to map the AI tools being deployed, piloted, or trialled across the criminal justice system in England and Wales. From that exercise, we identified a shortlist of tools suitable for our fictional scenario, and the evidence now going into the trial features outputs from facial recognition technology and AI transcription tools.
To guide this process, we convened a working group with members from the Crown Office and Procurator Fiscal Service (COPFS), the Crown Prosecution Service (CPS), and the charity JUSTICE. The project also held a World Café focus group, bringing together experts from policing, computer science, law, psychology, and linguistics to help refine the case scenario. We also worked with Scottish advocates who provided specialist advice on Scotland’s criminal justice process. As we head into the trials, we now have both the case materials and a data collection framework in place to explore issues such as the explainability of AI evidence, human understanding, and the legal impact of probabilistic AI outputs.
In Glasgow, 15 mock jurors will participate in our first trial in person before taking part in group deliberation. A real judge, prosecutor, defence advocate, police officer, and expert witness will assume their roles in the simulation, while actors will play the witnesses and the accused. The live mock trial will be video recorded and adapted as needed for future research.
These findings will feed into the project’s central output—a co-produced framework for responsible probabilistic AI, aimed at law enforcement bodies and policymakers, as well as the CPS, COPFS, and wider legal community. This research forms part of a Responsible AI UK-funded keystone project, led by Northumbria University alongside the universities of Glasgow, Northampton, Leicester, Newcastle, and Cambridge.
To follow our research findings, visit our website and LinkedIn page.
By Dr Angela Paul, Dr Elizabeth McClory-Tiarks, Dr Kyriakos Kotsoglou, Professor Marion Oswald, and Dr Mackenzie Jorgensen
This blog was compiled and authored by members of the PROBabLE Futures project team. Claude (an AI assistant developed by Anthropic) was used to support drafting the blog, primarily to correct grammar and reduce repetition, working from the first author’s initial draft. The final text has been reviewed, edited and approved by the project team, who are responsible for its content and accuracy.