|
Abstract: This article responds to Donald Berry’s article on inferring identity from DNA profiles and presenting a method for direct calculation of the probability that the suspect is guilty and the probability that an alleged father of a child is the true father. The method is Bayesian. This article examines, from the perspective of a lawyer, two connected issues: the forensic importance of quantifying measurement and sampling error and the desirability of combining likelihoods and priors for jurors or judges. It attempts to place Berry’s treatment off these matters in the context of the emerging case law on DNA profiling, and speculates about the advisability of bringing Bayes to the bar. It concludes that Berry’s article presents a cogent and powerful indictment of the matching and binning reasoning now used in single-locus DNA profiling, and that Berry builds an impressive case for using likelihoods that (a) make better use of the information in the test results and the population data and that (b) handle more of the uncertainties now present in DNA evidence. However, Berry’s strong Bayesian demand that jurors should quantify their prior probabilities and combine them with likelihood ratios based on certain simplifying assumptions to return a verdict of guilt or innocence is less persuasive, although the author is not prepared to say that there is no room for some form of a Bayesian presentation in a criminal trial. Considering the difficulties that many courts, attorneys, and jurors face in assessing quantitative evidence, Berry’s efforts and the efforts of others to develop suitable Bayesian analyses for forensic applications are a most welcome development.
|