Developed by an interdisciplinary team of Silicon Valley AI researchers, professional jury consultants, and lawyers, the project introduces an interpretable AI system that brings data-driven analysis to jury selection and trial preparation. The research, published on arXiv, will be presented in March 2026 at the ACM Symposium on Law and Computer Science at UC Berkeley.
Researchers built the system using responses from more than 400 mock jurors who reviewed the same case and rendered verdicts. The AI analyzes those responses to help predict how future jurors may decide and to show which themes or arguments tend to resonate with different groups of jurors.
In controlled evaluations, the AI system outperformed experienced trial consultants who assessed the same juror profiles under identical informational constraints. Unlike traditional jury consulting, the system is designed to be transparent, allowing attorneys to see which juror responses and questions carry the greatest predictive value. Ashwin Murthy, a Silicon Valley–based AI engineer who led the research, said the tool is intended to support—not replace—human judgment, particularly for litigants who lack access to costly trial consulting services.
Beyond jury selection, the system can also be used for data-driven trial preparation, including testing which legal theories resonate with different juror groups, examining the reasoning patterns behind juror decisions, and estimating potential damages in civil cases to help attorneys refine trial strategy.
Trial consulting often costs tens of thousands of dollars per case, putting it out of reach for most public defenders, said Lisa Pyle, a former Bronx prosecutor with more than 22 years of experience. AI-based decision-support tools, she said, could give public-interest litigators access to data-driven insights that have traditionally been available only to well-funded parties.
Beyond litigation, the tool is currently used in law school settings for hands-on jury selection exercises, giving students an opportunity to evaluate juror questionnaires in a data driven manner.
The researchers are now seeking collaborations with litigators to test the system in real-world settings, with an emphasis on public-interest cases. By making the tool available pro bono, the team aims to evaluate whether data-driven decision support can meaningfully expand access to trial-preparation resources that have traditionally been limited to well-funded parties.