AI in the Courtroom: Developing a Juror Decision Making Model with Topic Modeling and Cross Validation

The present study mined 1037 free-response mock jury verdict justifications and computed a CTM, yielding 6 topics: context, DNA, weapon, testimony, relationship, and framing. These topics were entered into a 10-fold cross validated GLM to predict guilty verdicts, resulting in a 76% accuracy (13% higher than the baseline model).