Par auteur > Mudry Léna

Assuming that the absenteeism rate is higher for women than for men, an employer believes of a particular female candidate that she is more likely to be absent than her Male counterpart. Consequently he decides to hire the male applicant. Is there anything wrong with believing that someone is likely to substantiate a disvalued property on the basis of an accurate statistical generalisation about their demographic group? Purists argue that such beliefs are often not epistemically justified because believers fail to understand the statistical generalisation (Gardiner 2018, Munton 2019). Even if the statistic is accurate, believers often assume a mistaken link between correlated properties and therefore over project the statistic upon novel cases. For short they assume a false statistical model or Bayesian network. As we see it, the choice of the model itself, on which we rely to draw inferences about others, depends on moral considerations (pace purism). First we shall argue that there is no systematic epistemic way to distinguish between models. If models are epistemically on par then moral considerations function as a tiebreaker. Second, we shall argue that, in some cases, if there is a loss of epistemic value, it is only true from a static perspective. If ones takes a dynamic perspective, some ethical choices can in fact be both epistemically and ethically superior. The second claim applies in cases of prevention, this is cases in which the choice of a model can have a performative effect on the individuals.