Abstract
In “Why the generality problem is everybody’s problem,” Michael Bishop argues that every theory of justification needs a solution to the generality problem. He contends that a solution is needed in order for any theory to be used in giving an acceptable account of the justificatory status of beliefs in certain examples. In response, first I will describe the generality problem that is specific to process reliabilism and two other sorts of problems that are essentially the same. Then I will argue that the examples that Bishop presents pose no such problem for some theories. I will illustrate the exempt theories by describing how an evidentialist view can account for the justification in the examples without having any similar problem. It will be clear that other views about justification are likewise unaffected by anything like the generality problem.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Strictly speaking, this is oversimplified. Belief preserving processes can be of equal importance to process reliabilists. They can hold that a reliability that pertains to belief preservation determines the justification of continuing beliefs. They can also hold that some combination of the formation and preservation processes is what counts. But the Generality Problem issues that affect belief forming processes can be readily seen to affect belief preservation in the same way. So a full solution to the Generality Problem for process reliabilism would have to cover belief preservation as well. For the sake of simplicity, only belief formation will be explicitly discussed here.
Perhaps in the case of demonstratively formulated perceptual beliefs, e.g., “I see that tiger,” the same BFP token could have produced a different belief content, when qualitatively the same scene included a numerically distinct demonstrated referent. If so, and if that referent were to have been a visually indiscernible fake tiger, then the same BFP token would have produced a false belief. Any tendency to this sort of falsehood is slight enough not to affect the conclusion that such BFP tokens would have an overly high truth frequency. The reliabilism would still fail to classify properly output beliefs that are too unreasonably formed to be justified. In any case, beliefs with no apparent demonstrative or indexical content, e.g., that carbon is an element, also exemplify the problem when they are unreasonably believed.
The “discernible and credible extension” terminology is intended to require something that is vital for a solution, but not susceptible to precise characterization. “Discernible” marks the fact that the specification must be usable by us to find the extension of the theory in some reasonable variety of examples. As for “credible,” clearly a successful solution cannot ruin the theory. So a proposed solution that gives the theory a preposterous extension definitely fails. A proposed solution that gives the theory a radically revisionary extension is a failure too, unless it is accompanied by an adequate defense of such a revision. On the other hand, some revisionary implications might be, on reflection, instructive rather than problematic. Also, process reliabilism might be mistaken because of some flaw other than a problem with the types selected—as would be the case if, for instance, it were shown that nothing causal affects the justification of belief. In light of such possibilities, the standard for successful solution is best described rather loosely as giving the theory a “credible” extension, an “acceptable,” “defensible,” or “tenable” one. Of course, doing better for the theory than this would be better. But something of this minimal sort is most plausible as a requirement for solving the problem.
V has the advantage over R that we have some pre-theoretical understanding of the intellectual virtues and vices. The range of eligible types in R is not limited in any such way. A consequentialist thesis that is more closely analogous to R uses an utterly schematic notion of “the relevant action type” to say what feature of an alternative act determines its moral permissibility. The consequentialist thesis says that an act is permissible if and only if its relevant type is good enough on average. Clearly this proposal lacks any extension in the absence of any further specification of “its relevant type”.
E is a paraphrase of the account of “well-founded” belief that is given in Feldman and Conee (1985).
Even where the vagueness makes us unsure how E applies, this need not be problematic for E. It might be that the justificatory status of the beliefs in examples of unclear evidence possession has a matching indeterminacy. If so, then E makes no classificatory mistakes in these cases.
Such proposed solutions are criticized in sections II and III of Conee and Feldman (1998). To describe trouble for the proposals in a few words: several common sense types of a BFP are frequently tied for salience. In any case, common sense types are too broad to account for all variations in justification. The narrowest natural kind of a BFP is narrowed by all causally relevant factors in a belief’s production. The result is that often, if not always, the selected type of BFP produces just one belief content. In consequence, unjustified beliefs the content of which tends to be true by natural necessity or the like get too much reliability for their actual strength of justification in some instances.
See Conee and Feldman (1998) for a critical survey of several proposed solutions.
Perhaps the attribution of reliability in the propositions also implies some tendency toward truth by virtue of membership in the class. Any such modal association with truth would only strengthen the contents of the justified beliefs as evidence for the target propositions.
As the example is described, Paula counts the time interval type as “reliable”, presumably owing to the very high actual ratio of the truths to the falsehoods that are believed by Paula during the time. It is doubtful that this truth ratio suffices for this particular type to enjoy any intuitive sort of reliability. The actual truth ratio seems too precariously associated with the interval to reveal the “reliability” that the type has, in the sense of its tendency toward truth. In any case, it can be readily seen that the E account would apply to what really was a reliability belief as well as it applies to the discussed belief about a high truth ratio.
References
Bishop, M. A. (2010). Why the generality problem is everybody’s problem. Philosophical Studies, 151(2), 285–298.
Breiman, L., Friedman, J., Olshen, R., & Stone, C. (1984). Classification and regression trees.
Conee, E., & Feldman, R. (1998). The generality problem for reliabilism. Philosophical Studies, 89, 1–29.
Feldman, R., & Conee, E. (1985). Evidentialism. Philosophical Studies, 48, 15–34.
Goldman, A. (1979). What is justified belief. In G. S. Pappas (Ed.), Justification and Knowledge (pp. 1–24). Dordrecht: D. Reidel.
Lackey, J. (2006). Introduction. In J. Lackey & E. Sosa (Eds.), The epistemology of testimony. Oxford: Oxford University Press.
Acknowledgment
I am grateful for comments from Michael Bishop and Richard Feldman.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Conee, E. The specificity of the generality problem. Philos Stud 163, 751–762 (2013). https://doi.org/10.1007/s11098-011-9843-x
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11098-011-9843-x