Problems of Forensic Sciences 2001 Vol. 46 (XLVI) 173-179
BAYESIAN NETWORKS AND THE EVALUATION OF SCIENTIFIC EVIDENCE: A THEORETICAL APPROACH
Franco TARONI1,2, Paolo GARBOLINO3
1Institut de Police Scientifique et de Criminologie, University of Lausanne, Lausanne, Switzerland
2Institut de Médecine Légale, University of Lausanne, Lausanne, Switzerland
3Dipartimento di Filosofia, Scuola Normale Superiore, Pisa, Italy
Streszczenie
Recently, methods that deal with formal analysis of decision making have been developed. Bayesian networks – also known as belief networks and causal probabilistic networks – provide a method of representing relationships between characteristics even if the relationships involve uncertainty, unpredictability or imprecision. This quantitative method assists the scientist not only in describing a problem and communicating information about its structure but also in calculating the effect of the truth of one proposition or piece of evidence on the plausibility of others. Notably, Bayesian networks are a network-based framework for representing and analysing situations involving uncertainty (i.e. evidence evaluation, criminal investigation, etc.). Information is presented in a graph as a set of nodes (representing the variables) linked by directed arcs (or edges) and the direction of the arc represents an influential relationship. The absence of an arc between two nodes implies that the two variables associated with these nodes are independent of each other, conditional on knowledge of the values of the other variables. The aim of this paper is to show how such a methodology could facilitate the representation and the evaluation of the scientific evidence. A simple scenario involving a transfer evidence is developed to show the role of different variables.
Słowa kluczowe
Evidence evaluation; Bayes’ theorem; Likelihood ratio; Bayesian networks.