(Im)probable stories: combining Bayesian and explanation-based accounts of rational criminal proof

A key question in criminal trials is, ‘may we consider the facts of the case proven?’ Partially in response to miscarriages of justice, philosophers, psychologists and mathematicians have considered how we can answer this question rationally. The two most popular answers are the Bayesian and the explanation-based accounts. Bayesian models cast criminal evidence in terms of probabilities. Explanation-based approaches view the criminal justice process as a comparison between causal explanations of the evidence. Such explanations usually take the form of scenarios – stories about how a crime was committed. The two approaches are often seen as rivals. However, this thesis argues that both perspectives are necessary for a good theory of rational criminal proof. By comparing scenarios, we can, among other things, determine what the key evidence is, how the items of evidence interrelate, and what further evidence to collect. Bayesian probability theory helps us pinpoint when we can and cannot conclude that a scenario is likely to be true. This thesis considers several questions regarding criminal evidence from this combined perspective, such as: can a defendant sometimes be convicted on the basis of an implausible guilt scenario? When can we assume that we are not overlooking scenarios or evidence? Should judges always address implausible innocence scenarios of the accused? When is it necessary to look for new evidence? How do we judge whether an eyewitness is reliable? By combining the two theories, we arrive at new insights on how to rationally reason about these, and other questions surrounding criminal evidence.