Premium
The Generality Problem, Statistical Relevance and the Tri‐Level Hypothesis
Author(s) -
Beebe James R.
Publication year - 2004
Publication title -
noûs
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.574
H-Index - 66
eISSN - 1468-0068
pISSN - 0029-4624
DOI - 10.1111/j.1468-0068.2004.00467.x
Subject(s) - generality , relevance (law) , philosophy , epistemology , functionalism (philosophy of mind) , sociology , psychology , psychoanalysis , law , political science , psychotherapist
A cognitive process will be reliable just when it yields a sufficiently high ratio of true to false beliefs. If a belief is produced by a process with a high degree of reliability, then that belief will have a high degree of justification. If, however, a belief is produced by a cognitive process with a low degree of reliability, then that belief will have a low degree of justification. After two decades of debate, a few objections have emerged as the standard objections to reliabilism. The Generality Problem is one such objection, the most visible proponent of which has been Richard Feldman (1985; Conee & Feldman 1998). It is now cited as a serious problem for reliabilism in almost every introductory text on epistemology. In this article I offer a solution to the Generality Problem. The Generality Problem arises because reliabilists claim that it is process types rather than process tokens that are the bearers of reliability. A process token is an unrepeatable, causal sequence occurring at a particular time and place. Consequently, you cannot ask whether a process token is reliable (i.e., whether it would produce mostly true beliefs over a wide range of cases). Accordingly, reliabilists have claimed that only process types can be reliable or unreliable. We can revise (R1) to take this point into account.