Abstract. According to the value-free ideal, the internal workings of science, including the evaluation of evidence, should be kept free from the influence of non-epistemic values as much as possible. We identify an underappreciated limit on the extent to which the value-free ideal can be achieved in practice. Our argument – which differs from inductive risk and other recent arguments – is grounded in the fact that, in some fields, scientists use complex scientific models as a replacement for background information in Bayesian and other methodologies that assign probabilities to hypotheses in light of evidence. We explain how this practice gives rise to conclusions that are influenced by non-epistemic values, and we consider the nature of the bias that it introduces. [This is joint work with Eric Winsberg.]

Wendy Parker is a Reader in Philosophy at Durham University.