La conférence fait partie du séminaire hedbomadaire de l'Epistemology Group de Duncan Pritchard.

Résumé

ABSTRACT. Several authors have recently defended a safety condition on knowledge according to which one knows only if one's belief couldn't easily have been wrong. The basic understanding of this clause is that S's belief that p is safe iff in all close worlds in which S believes that p, p is true. Here I raise four problems for basic safety: (1) it cannot deal with fake barn cases, (2) it cannot deal with borderline ignorance resulting from vague predicates or concepts, (3) it treats symmetrical cases in an asymmetrical way when necessary truths are involved, (4) it cannot deal with cases of knowledge based on accidentally-gathered evidence. I argue that all worries point out in the same direction: that safety has to be formulated with reference to methods. The consequences are (1) safety is not so different from reliabilism after all; in particular it faces the generality problem just as reliabilism does; (2) safety thus generalized may be sufficient for knowledge.

Une première version du papier, "Four Worries for Safety" est en ligne.

Je me réjouis de présenter devant les épistémologues edinburgeois, et espère que mon papier en sortira vivant!