Bayesian Decision Aids for Juries

If Bayesian models provide a standard for appropriate judgment, and if actual human judgment deviates from those models, then perhaps jurors need some assistance in drawing proper conclusions from evidence. This line of thinking has led to a number of proposals for the use of Bayesian decision aids, generally to help jurors to draw conclusions from forensic identification evidence, such as evidence of matching blood types or matching DNA profiles. Most proposals entail the use of likelihood ratios to characterize the value of the identification evidence, along with instructions to the jurors on how to use the likelihood ratio to update their prior subjective assessments of the probability of a relevant proposition. An early proposal of this nature (Finkelstein and Fairley, 1970) drew a famously eloquent response from constitutional scholar Laurence Tribe (1971), who acknowledged the appeal of a Bayesian perspective as a way for scholars to think about evidence, but lambasted the idea of providing Bayesian decision aids to jurors on doctrinal, constitutional and practical grounds. Tribe's analysis was in turn challenged as psychologically naive and anti-empirical (Saks and Kidd, 1980). Other psychologists (Faigman and Baglioni, 1988), statisticians (Friedman, 1997), philosophers (Cohen, 1977), and legal scholars (Kaye, 1979; Koehler and Shaviro, 1990) have joined the debate, which gained currency with the growing use of scientific evidence in the courtroom. Proponents of decision aids have made some progress. In paternity cases, for example, some courts provide jurors charts showing how a prior probability of paternity should be revised in light of the paternity index (likelihood ratio) associated with a genetic test (Kaye, 1988b). The National Research Council has suggested that experts might compute posterior probabilities to show jurors the power of DNA evidence for establishing identity (National Research Council, 1996, pp. 201-202), although this proposal remains controversial (see Thompson, 1997).

Was this article helpful?

0 0

Post a comment