Stop Guessing, Start Predicting
Guests: Pete Mansmann, Rich Epstein, and Oscar McKnight
In episode 41 of Just Verdicts, host Brendan Lupetin interviews the team behind Predict, a data-driven jury research tool that moves beyond traditional focus groups to provide scientific case forecasting. Pete Mansmann, Rich Epstein, and Oscar McKnight discuss how their unique Delphi groups and Human-Assisted AI (HA-AI) allow trial lawyers to measure anything – from case themes to a lawyer’s specific attire – to predict trial outcomes with remarkable accuracy.
The Just Verdicts podcast, hosted by attorney Brendan Lupetin, delves into compelling medical malpractice, medical negligence, and catastrophic personal injury cases. Each episode breaks down successful civil trials by featuring interviews with the trial lawyers, who discuss the innovative strategies, challenges, and key moments that led to significant verdicts for their clients.
Subscribe to Just Verdicts
The Science of the Delphi Group
Unlike a standard focus group, which often provides merely descriptive feedback from random participants, Predict utilizes Delphi groups to achieve predictive results.
- Normed Populations: Groups are “normed” to national or local demographics, ensuring a balance of ages (25 to 75), education levels (no diploma to PhD), and socioeconomic backgrounds.
- Volunteer Participants: Predict relies on unpaid volunteers rather than paid participants. Research shows that paying participants creates implied pressure to lean toward the side they believe is paying them, which ruins the data’s validity.
- Intact Groups: Delphi groups are “intact,” meaning researchers can go back to the same individuals days later to test how a new piece of evidence or a changed fact pattern alters their original decision.
- Segmented Fact Patterns: Instead of reading a full case summary at once, participants are “spoon-fed” one paragraph at a time. This allows researchers to see exactly where a juror’s mind shifts and how concepts build on one another as the “movie” of the case unfolds.
Vista vs. Fusion: Predict’s Current Offerings
Predict has boiled its services down to two primary products, both of which now include an AI component:
- Vista (Full Predict): This comprehensive study involves over 120 participants (10+ Delphi groups). It provides statistically valid demographic data to help identify your “best” and “worst” jurors for voir dire.
- Fusion (Hybrid): A quicker, less expensive option using 24 participants. While it lacks deep demographic statistical power, it is highly effective at identifying critical case concepts and award potentials early in litigation.
Human-Assisted AI (HA-AI) and “Fuzzy Logic”
A core evolution of the product is Human-Assisted AI, which learns from a proprietary 25-year database of jury research and the specific human comments from the current study.
- Pattern Recognition: HA-AI can identify patterns and predict outcomes even when human groups are deadlocked.
- Fuzzy Logic: This allows the AI to self-correct participant data. For example, if a participant identifies as a Republican but answers every question like a Democrat, the “fuzzy logic” formula self-corrects their demographic profile to improve prediction accuracy.
- Medical Treatises: The AI component can search global medical literature to find liability theories that even the retained experts might have missed.
The “62% Threshold”
Oscar McKnight explains that in probabilistic thinking, a 50/50 result is a coin flip. A lawyer should only feel truly confident when the Predict probability hits 62% or higher. This number represents a half-standard deviation above a normal curve, indicating the outcome is no longer happening by chance.
Need to refer a medical malpractice or personal injury case in Pennsylvania?
Lupetin and Unatin, Attorneys at Law, handle catastrophic injury and medical malpractice cases in Pennsylvania. If you are interested in co-counseling, local counseling, or referring a case, visit our attorney referral page.