Skip to main content
SpringerLink
Log in

Knowledge, evidence, and multiple process types

  • S.I.: Reliability
  • Published:
Synthese Aims and scope Submit manuscript

Abstract

The generality problem is one of the most pressing challenges for reliabilism. The problem begins with this question: of all the process types exemplified by a given process token, which types are the relevant ones for determining whether the resultant belief counts as knowledge? As philosophers like Earl Conee and Richard Feldman have argued, extant responses to the generality problem have failed, and it looks as if no solution is forthcoming. In this paper, I present a new response to the generality problem that illuminates the nature of knowledge-enabling reliability. My response builds upon the insights of Juan Comesaña’s well-founded solution to the generality problem, according to which relevant types are content–evidence pairs, i.e., descriptions of both the target belief’s content and the evidence on which the belief was based. While most responses to the generality problem, including Comesaña’s, only posit one relevant type for any given process token, I argue that knowledge-enabling reliability requires a process token to be reliable with respect to multiple content–evidence pairs, each with varying degrees of descriptive specificity. I call this solution multi-type evidential reliabilism (MTE). After offering a clear formulation of MTE, I conclude by arguing that MTE is sufficiently informative to rebut Conee and Feldman’s generality problem objection to a reliability condition on knowledge.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3

Similar content being viewed by others

Notes

  1. See Beddor and Goldman (2015). In their overview of reliabilist epistemology, they see the generality problem to be in the top six “problems,” or, “objections” to reliabilism.

  2. Feldman (1985) and Goldman (1979) are key figures who highlighted this important distinction for making sense of reliabilism. Most philosophers agree that types, rather than tokens, are the entities that can be measured for reliability. Although, Comesaña (2006) articulates a way in which tokens could be evaluated for reliability by taking a truth-ratio for that token across some class of possible worlds. However, Comesaña recognizes that framing reliabilism like this doesn’t get the reliabilist out of the generality problem (28–30). The reliabilist still would have to provide an account of which possible worlds were included in this truth-ratio.

  3. The specific language of “relevant” and “irrelevant” types was first introduced by Feldman (1985: p. 160).

  4. For instance, in Conee and Feldman’s canonical treatment of the generality problem, they describe the generality problem as undermining both “process reliability theories of justification and knowledge” (1998: p. 24). Interestingly enough, the growing literature on the generality problem does not feature any substantive discussion on the degree to which type-relevance for knowledge-enabling reliability correctly characterizes type-relevance for justification-enabling reliability, or vice versa.

  5. See Reichenbach (1949: p. 374) for a presentation of reference classes and their role in probability theory.

  6. Throughout the paper, I use the term “judgment” interchangeably with “belief.”

  7. See Wallis (1994) for a response to the generality problem that features a helpful discussion of the process type/reference class distinction. Note, Wallis calls reference classes “relevance classes” in his article.

  8. See Conee and Feldman (1998) for a criticism of most of the relevance theories presented up through that point in time, with a particular focus on Heller (1995), Alston (1995), Schmitt (1992), and Goldman (1979). For other theories of type-relevance, see Beebe (2004), Leplin (2007), Wallis (1994), Becker (2008) Adler and Levin (2002), Lepock (2009), Sosa (1991), Goldman (1986), Comesaña (2006), Greco (2010), Olsson (2016), and Wallbridge (2016). Also see Conee and Feldman (2002) for an updated criticism of Adler and Levin’s (2002) relevance theory. See Brueckner and Buford (2013) for a criticism of Becker (2008), and Dutant and Olsson (2013) for a criticism of Beebe (2004). See Matheson (2015) for a criticism of Comesaña (2006).

  9. Conee and Feldman’s demands “to know what determines [type] relevance” and insistence that this determining factor is “principled” rather than “ad-hoc” indicate that they are assuming the following claim: if there are facts about type relevance, then these facts must be explained or determined by some set of general principles (1998: pp. 3–4, emphasis mine). While this is the dominant assumption throughout the generality problem literature, not everyone agrees that type-relevance must have some principled and general theoretical account or explanation. Klemens Kappel defends what he calls the “no determination view,” according to which “there are no facts that determine relevant types” for particular tokens (2006: p. 256). In other words, the characteristics of a token’s relevant type are unexplained by any broader theoretical principles. It’s beyond the scope of this paper to offer a full rebuttal to Kappel. In what follows, I’ll agree with the majority of contributors to the generality problem literature and assume that if there are facts about which types are relevant for any given token, then there exists some is a reliability condition on knowledge, then type-relevance has a principled account or analysis.

  10. The way that I reconstruct Conee and Feldman’s reasoning here represents the inductive evidence (of failure to solve the generality problem) as supporting a metaphysical conclusion: there probably are no facts of the matter about type relevance and knowledge-enabling reliability. But one could interpret the inductive evidence to have more of an epistemic upshot: we could never know or reasonably judge any facts about type relevance and knowledge-enabling reliability. This interpretation of the argument would still (if cogent) undermine our reasons for accepting a reliability condition on knowledge. After all, if we could never know or reasonably determine any facts about knowledge-enabling reliability, then one should worry about our ability to apply reliabilism to particular (actual or hypothetical) cases of belief formation in order to test reliabilism for explanatory power—i.e., to test whether reliabilism accommodates our epistemic intuitions on a variety of cases.

  11. For example, Conee and Feldman (1998: pp. 6–11) argue that the “common-sense types” and “natural kinds” responses to the generality problem both fail to deliver a verdict on which types are relevant for any given token.

  12. For instance, Conee and Feldman show that Alston’s theory types processes too narrowly (1998: pp. 13–15) and that Adler and Levin’s theory (2002) types processes too broadly (Conee and Feldman 2002: pp. 102–103). Moreover, Brueckner and Buford (2013) show that Becker’s (2008) theory types processes too narrowly, and Dutant and Olsson (2013) show that Beebe’s (2004) theory types processes too narrowly. To clarify, narrower type descriptions build in more detail from the token process, and broader type descriptions build in less detail from the token process.

  13. Erik Olsson (2016) offers a similar explanation for why the generality problem objection is so pressing. Given Conee and Feldman’s arguments, it seems as if responses to the generality problem will make process types either “too broad” or “too narrow” for delivering justification verdicts that match our “everyday concept of justification” (181).

  14. Also, a single-type approach to epistemic reliability appears to be framing Richard Feldman’s original (1985) presentation of the generality problem:

    • Let us say, then, that for each belief-forming process token there is some "relevant" type such that it is the reliability of that type which deter mines the justifiability of the belief produced by that token. Thus, the reliability theory can be formulated as follows:

      • (RT) S's belief that is justified if and only if the process leading to S's belief that is a process token whose relevant process type is reliable. (1985: p. 160, emphasis mine)

  15. That said, some multi-type responses to the generality problem have been proposed. The earliest multi-type approach comes from Mark Wunderlich, as he presented his theory of vector reliability (2002). Wunderlich claims that every type exemplified by the token process plays some role in determining the degree of justification that a resultant belief has (243–245). While Wunderlich explores various approaches for aggregating all of the truth ratios from all of a token’s types in order to determine whether that token generates knowledge, he ultimately refrains from defending any particular aggregation approach, claiming that such a project is “beyond the scope of this paper” (249). Next, Wallbridge (2016) defends a multi-type approach to knowledge-enabling reliability. Like my own theory of type-relevance, it too draws inspiration from Comesaña’s well-founded solution to the generality problem. See fn. 24 for further discussion on Wallbridge’s proposal.

  16. In my (2018), I briefly entertain such a multi-type approach to knowledge-enabling reliability. I argued that if such a multi-type approach to reliability were correct, then we’d have good reason to believe that a proper function condition on knowledge is explanatorily dispensable. However, there I offered no substantive arguments in defense of this multi-type answer to the generality problem nor any thorough discussion of how to formulate this reliability condition on knowledge. Here, I take up both of these tasks in what follows.

  17. See Comesaña (2010: pp. 584–593) for further explanation and defense of this solution to the generality problem.

  18. Alston explicitly invokes these kinds of experiences as the grounds or “inputs” to relevant belief-forming processes (1995: pp. 17–18).

  19. Comesaña argued for a “diagonal reliability” requirement on knowledge (2002: pp. 261–262). As he makes clear, diagonal reliability is determined by truth-ratios that are counterfactual in nature (258).

  20. For example, Goldman (1988: p. 63) argues that epistemic reliability measurements are taken across possibilities that that are sufficiently close to the actual case. Interestingly enough, Duncan Pritchard’s account of the safety condition on knowledge is rather similar to RC:

    If a believer knows that p, then in nearly all, if not all, nearby possible worlds in which the believer forms the belief that p in the same way as she does in the actual world, that belief is true (2005: p. 163).

    It appears as if, according to Pritchard, a belief is safe just if there’s a sufficiently high truth-ratio taken across a reference class of possibilities—that are sufficiently nearby—in which the subject forms a judgement “in the same way” that she does in the token case. See fn. 27 below for a further discussion on the relationship between safety and knowledge-enabling reliability.

  21. See Wallbridge (2016: p. 347) for a different sort of counterexample to the sufficiency of Comesaña’s theory of justification. Wallbridge uses an altered version of the famous fake-barn case to construct this counter-example. See fn. 24 for further discussion of Wallbridge’s theory of type-relevance.

  22. In Matheson’s presentation of this problem, he uses stable propositions that are contingently (yet eternally) true, like the gravitational force is proportional to the inverse-square of the distances between two objects (2015: pp. 465–467). While the use of contingent stable propositions suffices for highlighting this narrowness worry for the well-founded solution, I think that judgments on necessary truths (which manifest actual and counterfactual stability) illustrate this point even more clearly.

  23. Much thanks to an anonymous referee who insightfully encouraged me to frame Comesaña’s well-founded solution as one of the reliability necessary conditions on knowledge.

  24. Importantly, COLOR 1 highlights a way to construct a counter-example to the sufficiency of Kevin Wallbridge’s (2016) account of knowledge-enabling reliability. Wallbridge uses a slightly-altered version of the famous fake barn case to point out that, at least in some cases, certain portions of the target belief’s propositional content are based on only certain portions of the total evidence used for basing the belief. Furthermore, these different portions of evidence work together in a “combinatorial structure” to support the entire target proposition (2016: p. 349). For instance, consider a token tr, in which one visually comes to believe that object x1 is a red 14 barn. Token tr has the following relevant type components. Component C1: [forming a judgment on whether x1 is red 14, on the basis of a visual experience representing x1 to be red 14]; Component C2: [forming a judgment on whether x1 is a barn, on the basis of a visual experience representing x1 to be barn-shaped] According to Wallbridge, C1, C2, and the combination type C1 and C2 must all three be reliable in order for tr to generate knowledge (349–350). That said, imagine someone who, counterfactually, is very good at recognizing things to be red 14, but terrible at (accurately) visually representing every other color shade. Such a person’s tr token could be counterfactually reliable with respect to C1, C2, and C1 and C2, but since it is also counterfactually unreliable with respect to Cb (much like Susie’s token tc1), it seems like tr would fall short of having knowledge-enabling reliability.

  25. The notion of counterfactual “distance” as being determined by a similarity/ordering relation of possible worlds plays a central role in both Lewis (1973) and Stalnaker (1968) semantics for counterfactual sentences.

  26. Most importantly, I hold that MTE is one of the necessary conditions on knowledge. Moreover, I think MTE crystalizes the sort of multi-type structure that characterizes knowledge-enabling reliability. However, for all I’ve said here, there could be additional necessary conditions on knowledge that pertain to knowledge -enabling reliability. For example, Duncan Pritchard has recently suggested that, in order to generate knowledge, tokens must manifest reliability across multiple reference classes each characterized by different counterfactual distances from the actual token case (2012: pp. 179–180). MTE could be augmented to incorporate such a reliability requirement across numerous counterfactual distances, although exploring this option is beyond the scope of this paper. Furthermore, one might think that there are multiple similarity relations in addition to L that determine a token’s relevant types. On this picture, a token would have multiple concentric sphere type structures, each corresponding to a different similarity relation.

  27. See fn. 20. More recently, Duncan Pritchard has clarified that a belief that p is safe if and only if there is a sufficiently high truth-ratios measured across nearby possible judgments regarding p made “on the same basis” as the actual belief itself (2012: p. 176, p. 179). If MTE is correct, then we can reasonably read Pritchard’s safety condition on knowledge as constituting one part of the reliability condition on knowledge. In particular, Pritchard’s safety condition expresses the reliability requirement with respect to just the token’s precise content-evidence pair (CEt).

  28. Thanks to an anonymous referee for raising the example of CV and encouraging me to clarify the difference between relevant types according to MTE and irrelevant types like CV.

  29. That being said, according to MTE it is still possible for someone’s visual incompetence at representing violet shades to undermine his knowledge in cases where one visually ascribes orange 47 to some object. It all depends on the distribution of possible belief-forming events that are counterfactually close to the token process. Suppose that, rather than living on Planet X, Jim lives on Planet Y where 80% of all things are colored some shade of violet. In this scenario, when Jim undergoes process token tJ2 in which he visually identifies an object as being orange 47, he is surrounded by mostly violet-shaded objects. Here, it does seem that Jim’s inability to visually represent violet shades undermines the knowledge-enabling reliability of tJ2. MTE can straightforwardly accommodate this result, because tJ2 is counterfactually unreliable with respect to the concentric type Cb in this altered scenario. Relative to tJ2, the majority of counterfactually nearby color shade judgments would be judgments about violet objects (given that tJ2 occurs on Planet Y). But given that Jim is disposed to misrepresent violet shades, the truth ratio across all counterfactually nearby Cb judgments would be low.

  30. Thanks to an anonymous referee for identifying the current limitations of MTE and work that remains to be done in order to solve the generality problem.

  31. As Korcz illustrates, there already exists a vast and highly contentious literature on the nature of the epistemic basing relationship (2015). Furthermore, at this stage of the inquiry, one could reasonably foresee various approaches that might be taken to further explicate concepts (c) and (d). Take similarity relation L for instance. The contemporary work on similarity/ordering relations in the counterfactual semantics literature might provide insights into the nature of L as well. David Lewis argues that counterfactual similarity relations are determined by features of the context in which the counterfactual sentence is uttered (1979: p. 465). Analogously, one might argue that relation L functions in a similar way, thus making for an interesting sort of contextualism about knowledge attributions that arises due to the contextual nature of process type relevance. Another option worth exploring would be to analyze relation L in terms of understanding. Plausibly, in order to base some belief that p1 on evidence e1, the subject must possess some sort of understanding of the relationship between p1 and e1. Let U1 denote this cognitive state of understanding that corresponds to content-evidence pair [p1, e1]. Presumably, there’s a unique cognitive state of understanding that corresponds to each distinct content-evidence pair. For example, let U2 denote the state of understanding corresponding to [p2, e2]. Perhaps the epistemically relevant degree of similarity (according to L) that holds between content-evidence pairs [p1, e1] and [p2, e2] is determined by the degree of similarity between cognitive states U1 and U2. The challenge for this approach lies in explicating the kind of understanding involved in epistemic basing. Lastly, after further exploration, one might uncover good reason to conclude that L is conceptually primitive, admitting of no further informative analysis or explanation. Certainly, this is a result that we’re in no position to rule out at this point in time. Of course, some other approach to explicating L may emerge—these are just preliminary suggestions. But what’s important to notice at this point, regarding the generality problem objection, is that we have no reason to doubt that further theoretical progress on notions like L can occur.

  32. Interestingly enough, Comesaña, in his presentation of the well-founded solution, seems to assume the same account of evidence as Conee and Feldman: evidence consists of mental states (Conee and Feldman 2008: pp. 84–88).

  33. For example, many philosophers, including Price (1969) and Audi (1972), think that belief is a dispositional concept. On this view, whether one believes that p in situation W depends on whether she has dispositions to think or act in certain ways in W. The dispositional theory of belief has broad appeal and seems reasonable as stated. To the point in question, it would be odd if the reasonability of the dispositional theory of belief were in some way undermined simply due to an absence of further theoretical work to explicate belief’s corresponding similarity relation and relevant maximum degree of difference.

  34. Evidentialist Kevin McCain has recently offered a substantive and compelling defense of a dispositional account of evidence possession (2014: pp. 31–55). I defend a dispositional account of evidence possession as well (2017: pp. 1949–1950). According to both McCain and myself, if evidence possession (at a given time t) only depends on one’s occurrent mental states, then evidentialism won’t be able to account for the vast majority of the justified beliefs that we plausibly have.

  35. Additionally, I argue that there’s no reason to view these sorts of dispositions as importantly different or less complex than the dispositions invoked by reliabilism (1951–1953).

  36. Much thanks to Ted Warfield, Blake Roeber, Tom Senor, Liz Jackson, Ting Cho Lau, Andrew Moon, and Peter Finocchiaro for helpful conversations and comments at various stages of drafting this paper.

References

  • Adler, J., & Levin, M. (2002). Is the generality problem too general? Philosophy and Phenomenological Research, 65(1), 87–97.

    Article  Google Scholar 

  • Alston, W. (1995). How to think about reliability. Philosophical Topics, 23(1), 1–29.

    Article  Google Scholar 

  • Audi, R. (1972). The concept of believing. Personalist, 53, 43–62.

    Google Scholar 

  • Becker, K. (2008). Epistemic luck and the generality problem. Philosophical Studies, 139(3), 353–366.

    Article  Google Scholar 

  • Beddor, B., & Goldman, A. (2015). Reliabilist epistemology. In E. N. Zalta (Ed.), The Stanford Encyclopedia of Philosophy (Winter 2015 Edition). http://plato.stanford.edu/archives/win2015/entries/reliabilism/.

  • Beebe, J. R. (2004). The generality problem, statistical relevance and the tri-level hypothesis. Noûs, 38(1), 177–195.

    Article  Google Scholar 

  • Brueckner, A., & Buford, C. T. (2013). Becker on epistemic luck. Philosophical Studies, 163, 171–175.

    Article  Google Scholar 

  • Comesaña, J. (2002). The diagonal and the demon. Philosophical Studies, 110, 249–266.

    Article  Google Scholar 

  • Comesaña, J. (2006). A well-founded solution to the generality problem. Philosophical Studies, 129(1), 27–47.

    Article  Google Scholar 

  • Comesaña, J. (2010). Evidentialist reliabilism. Noûs, 44(4), 571–600.

    Article  Google Scholar 

  • Conee, E., & Feldman, R. (1985). Evidentialism. Philosophical Studies, 48, 15–34.

    Article  Google Scholar 

  • Conee, E., & Feldman, R. (1998). The generality problem for reliabilism. Philosophical Studies, 89(1), 1–29.

    Article  Google Scholar 

  • Conee, E., & Feldman, R. (2002). Typing problems. Philosophical and Phenomenological Research, 65(1), 98–105.

    Article  Google Scholar 

  • Conee, E., & Feldman, R. (2008). Evidence. In Q. Smith (Ed.), Epistemology: New essays (pp. 83–104). Oxford: Oxford University Press.

    Chapter  Google Scholar 

  • Dutant, J., & Olsson, E. J. (2013). Is there a statistical solution to the generality problem? Erkenntnis, 78, 1347–1365.

    Article  Google Scholar 

  • Feldman, R. (1985). Reliability and justification. The Monist, 68(2), 159–174.

    Article  Google Scholar 

  • Goldman, A. (1979). What is justified belief? In G. S. Pappas (Ed.), Justification and knowledge (pp. 1–23). Dordrecht: D. Reidel Publishing.

    Google Scholar 

  • Goldman, A. (1986). Epistemology and cognition. Harvard: Harvard University Press.

    Google Scholar 

  • Goldman, A. (1988). Strong and weak justification. Philosophical Perspectives, 2, 51–69.

    Article  Google Scholar 

  • Greco, J. (2010). Achieving knowledge. Cambridge: Cambridge University Press.

    Book  Google Scholar 

  • Heller, M. (1995). The simple solution to the generality problem. Noûs, 29(4), 501–515.

    Article  Google Scholar 

  • Kappel, K. (2006). A diagnosis and resolution to the generality problem. Philosophical Studies, 127, 525–560.

    Article  Google Scholar 

  • Korcz, K. A. (2015). The epistemic basing relation. In: E. N. Zalta (Ed.), The Stanford Encyclopedia of Philosophy (Fall 2015 Edition). https://plato.stanford.edu/archives/fall2015/entries/basing-epistemic.

  • Leplin, J. (2007). In defense of reliabilism. Philosophical Studies, 134, 31–42.

    Article  Google Scholar 

  • Lepock, C. (2009). How to make the generality problem work for you. Acta Analytica, 24, 275–286.

    Article  Google Scholar 

  • Lewis, D. (1973). Counterfactuals. Cambridge: Harvard University Press.

    Google Scholar 

  • Lewis, D. (1979). Counterfactual dependence and time’s arrow. Noûs, 13(4), 455–476.

    Article  Google Scholar 

  • Matheson, J. D. (2015). Is there a well-founded solution to the generality problem? Philosophical Studies, 172, 459–468.

    Article  Google Scholar 

  • McCain, K. (2014). Evidentialism and epistemic justification. New York: Routledge.

    Book  Google Scholar 

  • Olsson, E. J. (2016). A naturalistic approach to the generality problem. In B. P. McLaughlin & H. Kornblith (Eds.), Goldman and his critics (pp. 178–199). Chichester: Wiley-Blackwell.

    Chapter  Google Scholar 

  • Price, H. H. (1969). Belief. London: George Allen and Unwin.

    Google Scholar 

  • Pritchard, D. (2005). Epistemic luck. Oxford: Oxford University Press.

    Book  Google Scholar 

  • Pritchard, D. (2012). A defense of modest anti-luck epistemology. In K. Becker & T. Black (Eds.), The sensitivity principle in epistemology (pp. 173–192). Cambridge: Cambridge University Press.

    Chapter  Google Scholar 

  • Reichenbach, H. (1949). The theory of probability. Oakland: University of California Press.

    Google Scholar 

  • Schmitt, F. F. (1992). Knowledge and belief. London: Routledge.

    Google Scholar 

  • Sosa, E. (1991). Knowledge in perspective. Cambridge: Cambridge University Press.

    Book  Google Scholar 

  • Stalnaker, R. (1968). A theory in conditionals. Studies in Logical Theory, American Philosophical Quarterly, 2, 98–112.

    Google Scholar 

  • Tolly, J. (2017). A defense of parrying responses to the generality problem. Philosophical Studies, 174(8), 1935–1957.

    Article  Google Scholar 

  • Tolly, J. (2018). Swampman: A dilemma for proper functionalism. Synthese. https://doi.org/10.1007/s11229-018-1684-0.

  • Wallbridge, K. (2016). Solving the current generality problem. Logos and Episteme, 7(3), 345–350.

    Article  Google Scholar 

  • Wallis, C. (1994). Truth-ratios, process, task, and knowledge. Synthese, 98, 243–269.

    Article  Google Scholar 

  • Wunderlich, M. (2002). Vector reliability: A new approach to epistemic justification. Synthese, 136, 237–262.

    Article  Google Scholar 

Download references

Author information

Authors and Affiliations

  1. Department of Philosophy and Religion, University of Indianapolis, 1400 E. Hanna Ave., Indianapolis, IN, 46227, USA

    Jeffrey Tolly

Authors
  1. Jeffrey Tolly
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Jeffrey Tolly.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Tolly, J. Knowledge, evidence, and multiple process types. Synthese 198 (Suppl 23), 5625–5652 (2021). https://doi.org/10.1007/s11229-019-02146-4

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11229-019-02146-4

Keywords

Search

Navigation