Sopha 2009

1. Commentaire de Julien Dutant - le 1 juillet 2009 à 13:16

Here´s the general idea of the paper. One can have the following picture of the current landscape in epistemology: there is a widespread acceptance of Unger´s idea that knowledge is a matter of having a non-accidentally true belief. Most of the debate on the analysis of knowledge boils down on how to cash out the notion of ´non-accidentality´ at play. In particular, we could distinguish:
- modal accounts, according to which (roughly) something is « non-accidental»  if it is necessary, that is if it holds throughout some relevant sphere of worlds. (Dretske´s « conclusive reasons» , Nozick´s « sensitivity» , Sosa´s and Williamson´s « safety» , and so on, all fit here.)
- explanationist accounts, that involve a « because»  relation. Roughly: either S knows p only if S believes p because p (Jenkins), or S knows p only if S´s belief that p is true because of the way in which it is formed (Sosa´s and Greco´s « true-because-virtuous» ).
Since modal notions are best understood and formalized that explanationist ones, one should first explore modal accounts – and in fact they have been the most popular. It seems clear nowadays that the « safety»  conditions are doing the best job.

This paper is about puzzling types of case that seem to raise problem for the safety account. You could call them « barely-Gettiered cases» : there are cases in which a Gettier-style process could easily have intervened between the facts and the agent’s belief so that the agent could easily have formed a false belief, but where we still have the intuition that the agent knows. For instance: there´s an evil demon who, in 1900, threw the dice to decide whether he would fool everybody in a Cartesian fashion. The result of the dice made the demon give up on that project. Do people afterward fail to know anything about the external world, since massive deception was so close to happen? many people say that no. Still their beliefs are unsafe, and so would count as non-knowledge by a standard safety account. (NB, with a bit of effort you can easily construct ordinary cases with the same structure. The intuitive judgments are the same.)

In a way this is about rescuing the whole « modal account»  approach, since safety looks like its last viable option. I´m arguing that that these cases raise problems for a safety account IF you take closeness of worlds to be, roughly, closeness on the branching-time tree. (What´s close is what could easily have happened, that is what lies on a close branch of the tree to yours.) A different notion of closeness, based on the idea that « close»  worlds are variants of the actual world that are at least as normal as the actual world, deals with the case, but it´s unclear how to interpret that notion of closeness. (e.g. does it reflect bona fide interesting metaphysical facts about knowledge, or does it reflect some heuristic for knowledge attribution.)

2. Commentaire de Julien Dutant - le 29 août 2009 à 13:38

I’ve uploaded a draft of the paper; the « aptness»  side is not covered in the new draft, but I will mention it in the presentation.