Equal prior probabilities: Can one do any better?

This paper discusses recommendations concerning the use of prior probabilities that underlie recent, but in no way novel, proposals of presenting scientific evidence in terms of posterior probabilities, in the context sometimes referred to as the `full Bayes' approach'. A chief issue of this procedure is a proposal that - given the unavailability of case-specific circumstantial information - scientists should consider the prior probabilities of the propositions under which scientific evidence is evaluated as equal. The discussion presented here draws the reader's attention to the fact that the philosophical foundations of such a recommendation (in particular, attempted justifications through the Principle of Maximum Entropy (PME)) are far more controversial than what is actually admitted by the advocates for their use in the theory and practice of forensic science. Invoking only basic assumptions and the mathematical rules of probability calculus, the authors of this paper propose an argument that shows that there can be other more feasible and defensible strategies for eliciting reasonable prior probabilities. It is solely demanded that the reasoner is willing to make up his mind seriously on certain standard issues of fairly general criminal cases, such as evidential relevance or the probability of a suspect's guilt. However, because these issues intimately pertain to the responsibility of the trier of the fact, it is argued here that scientists' attempts to define appropriate prior probabilities should continue to be considered as untenable for the need. (c) 2006 Elsevier Ireland Ltd. All rights reserved.