What is it in virtue of which a fact is evidence for a hypothesis? Our philosophical accounts fall into two categories. First are accounts based on some features of the probabilistic relations between the evidence and the hypothesis – for example, increase in probability or various functions of likelihoods (see Mayo, 1996 Chapter 3 for an overview of such positions). These are not useful for evidence-based policy. What we need is a concept of evidence that we can use to judge whether some fact should be taken into consideration – whether it should be ‘on the table’ for consideration. Then we would expect to look at all the evidence on the table to decide on the probability of the proposed policy claim. Concepts of evidence based on facts about probabilities put the cart before the horse. We need a concept that can give guidance about what is relevant to consider in deciding on the probability of the hypothesis not one that requires that we already know significant facts about the probability of the hypothesis on various pieces of evidence.
Second are those accounts that are based on facts about explanation – for example, versions of inference to the best explanation (Lipton, 2004) or explanatory connectedness (Achinstein, 2001). The problem here is the concept of explanation. A good many accounts end up explaining explanation by reference to probability relations between the ‘explanans’ [the means of making plain] and the ‘explanandum’ [that which is being made plain]. This simply recreates the previous problem.