An information theory model for the evaluation of circumstantial evidence

The problem of assessing the weight of circumstantial evidence linking a suspect with the perpetrator of a crime is considered. The quantitative measure P_G of weight of circumstantial evidence that is used is the probability that culprit and suspect may be identified with certainty as one and the same based solely upon the circumstantial evidence. A unifying model for evaluating P_G taking into account expert testimony is established on the basis of the Kullback-Leibler discrimination information measure. This model is applied to deriving P_G based upon a jurist´s interpretation of the expert testimony. An important property of the model is the first-order agreement, as the testimony becomes compelling, of P_Gs based on different jurists´ viewpoints. The methodology developed is applied to the well-known Collins case.