Is There a Statistical Solution to the Generality Problem?

Abstract

This article is concerned with a statistical proposal due to James R. Beebe for how to solve the generality problem for process reliabilism. The proposal is highlighted by Alvin I. Goldman as an interesting candidate solution. However, Goldman raises the worry that the proposal may not always yield a determinate result. We address this worry by proving a dilemma: either the statistical approach does not yield a determinate result or it leads to trivialization, i.e. reliability collapses into truth (and anti-reliability into falsehood). Various strategies for avoiding this predicament are considered, including revising the statistical rule or restricting its application to natural kinds. All amendments are seen to have serious problems of their own. We conclude that reliabilists need to look elsewhere for a convincing solution to the generality problem

A New Statistical Solution to the Generality Problem

“…To loosen homogeneity is to impose a condition that keeps subtypes like T+ and T from being "admissible subtypes" (p. 1358). Dutant and Olsson note Wesley Salmon's (1984) definition of homogeneity, which rules out subtypes that cannot "be 'determined without reference to' the outcome we are considering" (Dutant andOlsson, 2013, p. 1358 …even if we conceded that [the process type] "occurring in the vicinity of maple trees" does make reference to the process's outcome, there are many other features that could less easily be claimed to do so, but that are equally irrelevant to the justification of Smith's belief and yet statistically relevant to his forming a true belief. Candidates are, for instance, "occurring in an area where there were maple trees a few seconds ago", "occurring while perceiving an object that many people to take to be a maple tree", or "occurring in the vicinity of bits of DNA of type X", where X is a chemical description of DNA that is maple-tree specific, and so on (p. 1359).…”

“…Dutant and Olsson (2013: 1354–5) raise a similar problem: call it the Trivialization Problem . Consider process type T , and consider the proper subtype of T (call it T+ ) that contains all and only those tokens of T that produce true beliefs.…”

“…To loosen homogeneity is to impose a condition that keeps subtypes like T+ and T − from being “admissible subtypes” (p. 1358). Dutant and Olsson note Wesley Salmon's (1984) definition of homogeneity, which rules out subtypes that cannot “be ‘determined without reference to’ the outcome we are considering” (Dutant and Olsson 2013: 1358). While specifying homogeneity in this way is supposed to help prevent trivialization, there are issues with Salmon's proposal.…”

“…There's just one desideratum it lacks: being possibly true. In their sustained treatment of the Statistical Solution, Julian Dutant and Erik Olsson (2013) successfully refute it. They go a whit too far, however, in maintaining that all Statistical Solutions are dubious.…”

“…Solutions have been proposed by Goldman (1986: 49–51), Schmitt (1992), Alston (1995), Wunderlich (2003), Beebe (2004), Becker (2008), and Olsson (2016). But as Dutant and Olsson (2013) note: “although much work has gone into solving the Generality Problem, none of the proposals that have emerged so far has gained widespread acceptance” (p. 1351). While a full treatment of the aforementioned solutions is beyond the scope of this paper, certain of their shortcomings may become apparent along the way.…”