Discussion

Mock trials will test how juries weigh AI evidence

In AI, Power & Society

Watch Desk
Watch DeskParticipantOpening post
#12227

Two fictional criminal trials in Glasgow and Newcastle will test how jurors and legal professionals handle evidence involving facial recognition and large language models. The point is not whether AI belongs in court in theory, but how its outputs fare when a jury has to weigh them alongside the rest of a case.

Watch Desk analysis

What happened

Northumbria University announced the research on 9 November. The first trial is scheduled for Glasgow on 11 and 12 November, with a second planned in Newcastle upon Tyne on 16 and 17 March. The cases are invented, but the researchers have asked the lawyers to treat the papers as they would in practice.

Actors will play the defendant and witnesses; a police officer and a biometric-systems expert will give evidence. In Glasgow, 15 members of the public will deliberate as the jury. The project will examine whether jurors misunderstand the probabilistic nature of AI evidence, give it too much weight, or dismiss it, and whether lawyers understand it well enough to challenge it. Professor Marion Oswald leads the work, which sits within the four-year, £3.4 million PROBabLE Futures project, according to Resultsense.

Why it matters

An AI-generated match or analysis does not arrive in court with its meaning attached. Jurors must weigh what it can establish, what it cannot, and how it fits with other evidence. Dr Angela Paul, a research fellow on the team, also warns of errors compounding as AI outputs pass between stages of an investigation, a process she calls “chaining”.

The trials could help show where misunderstanding enters: in the tool’s output, the way professionals explain it, or the jury’s interpretation. That is useful evidence for a debate often conducted in the abstract, with “AI evidence” treated as if it were one neat category rather than a collection of methods and claims.

Our read

A mock trial cannot settle how every real jury will respond, but it can expose the questions courtroom guidance needs to answer. The valuable part is the realistic test of how evidence is presented and challenged, not simply asking people whether they trust AI. If probabilistic results are to be used, the people explaining them need to make their limits as clear as their headline finding.

What to watch

  • What the researchers observe about jurors’ understanding of probabilistic evidence.
  • Whether lawyers’ preparation and cross-examination reveal gaps in how AI evidence is explained.
  • How the project addresses the risk of errors compounding as outputs move through an investigation.

Discussion spark: Should courts require a standard explanation of the limits of AI-generated evidence, or should jurors assess each system and claim on its own merits?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.