Modeling Relevance

During the past decade, particularly during the years immediately following the California Supreme Court’s decision in People v. Collins, a number of articles have appeared suggesting ways in which jurors might use certain mathematical techniques of decision theory as aids in the rational evaluation of circumstantial evidence. Professor Tribe, in an important response to the post-Collins articles, argues against introducing these techniques into the factfinding process. Problems that Tribe foresees include the necessary imprecision of the probabilistic estimates that these techniques require, the dwarfing of soft variables by those that are more readily quantified, and the potential dehumanization of the trial in the name of rational factfinding.

In this article I try to show the utility of two simple models, Bayes’ Theorem and regret matrices, for thinking about the meaning of relevance and for analyzing those evidentiary rules, which I call the “relevance rules,” generally associated with this topic. The discussion assumes that the factfinder is a jury and, unless otherwise noted, that the issue to be resolved is a defendant’s guilt. However, the analysis may be readily generalized to the situation where the factfinder is a judge and/ or a question other than guilt is at issue. The first section of this article applies the two models to a simplified situation where the factfinder must evaluate only one item of indisputably accurate testimony. The second section explores complexities that can arise when a case involves two or more items of possibly unreliable evidence.