Handling Manipulated Evidence.pdf

Gianluca Baio, Fabio Corradi U n i v e r s i t à d e g l i S t u d i d i F i r e n z e Handl ing Manipulated Ev idence W O R K I N G P A P E R 2 0 0 4 / 1 3 D ip a rt im e n to d i S ta ti s ti c a “ G . P a re n ti ” – V ia le M o rg a g n i 5 9 – 5 0 1 3 4 F ir e n z e - w w w .d s .u n if i. it S t a t i s t i c s HANDLING MANIPULATED EVIDENCE GIANLUCA BAIO AND FABIO CORRADI University of Florence Department of Statistics “G. Parenti” 59, Viale Morgagni, Florence 50134 - Italy {baio,corradi}@ds.unifi.it Abstract. Bayesian Networks have been advocated as a useful tool to de- scribe the relations of dependence/independence among random variables and relevant hypotheses in a crime case. Moreover, they have been applied to help the investigator structure the problem and weight the observed evidence, typ- ically with respect to the hypothesis of guilt of a suspect. In this paper we describe a model to handle the possibility that one or more pieces of evidence have been manipulated in order to mislead the investigations. This method is based on causal inference models, although it is developed in a different, specific framework. Keywords: Bayesian Networks, Influence Diagrams, Forensic identification, Ma- nipulated Evidence, Causal Inference. 1. Introduction Bayesian Networks (BNs) have recently been advocated in forensic science as a powerful tool to establish the overall dependence between relevant hypotheses and observable random variables, and to evaluate the probabilistic effects of the latter on the former during an investigation, or in a trial (Dawid Evett 1997, Garbolino Taroni 2002). A further interesting feature of a BN is that it can be easily increased including relations with previously not considered variables, whenever required and as usually happens in real practice. A possible subtle drawback in the use of BN-assisted investigations