Problemet med ursprungssannolikheter och naked statistical evidence i bayesiansk bevisvärderingsmetod

Sammanfattning: The thesis investigates the problem of naked statistical evidence and the problem of priors in the Bayesian theory of legal evidence. To fulfil this purpose, a transdisciplinary and legal philosophical method is used which enables an examination of the epistemological and normative considerations that characterize inquiries in legal evidence theory. The theoretical foundations of the Bayesian method are presented in the introductory part of the thesis to set the scene for the two problems subject to investigation. Thereafter, an account is given of the problem with naked statistical evidence and the problem of priors. Solutions to the problems are surveyed and considered. The problem with naked statistical evidence suggests that a probability-based view on legal proof is incompatible with basic intuitions regarding legal justice. Defenders of the Bayesian method, however, believe that the problem can be accommodated within the framework of the method. This thesis does so by defending a solution to the problem that has its basis in legal incentive structures. It is argued that this solution not only explains the problem of naked statistical evidence, but also provides a more extensive explanatory value in relation to related problems in legal evidence theory. The problem of priors is paramount for the validity of the Bayesian method, since a well-justified prior is necessary for the method to be epistemologically sound. The thesis defends two proposals as to how the problem can be addressed, the frequency principle and the normative prior. It is argued that the frequency principle can handle the problem of naked statistical evidence and other related problems concerning statistical evidence. Furthermore, it is emphasized that the frequency principle can be supplemented by the more practically oriented normative prior.