Abstract
The safety principle is the view that, roughly, if one knows that p, p could not easily have been false. It is common for safety theorists to relativize safety to belief-formation methods. In this paper, I argue that there is no fixed principle of method-individuation that can stand up to scrutiny. I examine various ways to individuate methods and argue that all of them are subject to serious counterexamples. In the end, I conclude by considering some alternative ways to preserve the insight behind safety without invoking a fixed principle of method-individuation.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Pritchard (2005) characterizes safety along these lines. The nearness relation of possible worlds is typically understood according to Lewis/Stalnaker possible world semantics. Distance between worlds is a function of similarity between them, such that the more similar a possible world is to another possible world, the nearer (or closer) it is. (Some have suggested that safety theorists’ understanding of nearness relation is problematic. See, e.g. Baumann (2008). For a helpful survey on the methodology of nearness relation in safety, see Bogardus (2014, pp. 294–297)).
Hawthorne brings up a similar example to refute sensitivity: “Suppose I observe a real dog that obscures an excellent facsimile of a duck. I believe a dog is in front of me. That belief is sensitive. I deduce that there is an animal in front of me. That isn’t sensitive: if there weren’t an animal in front of me, I would see the fake duck and believe there was an animal there” (Hawthorne 2004, p. 42). In a footnote on same page, Hawthorne admits that such a counterexample can be fixed by Nozick’s talk of ‘method’ of belief.
This formulation is adapted from Alfano (2009, p. 279). .
Goldman (1976, 2009) presents a similar case involving a dachshund and a wolf: “I look at a nearby dachshund and truly believe that what I see is a dog. Had I not been seeing a dachshund, I would have been seeing a wolf, and would have falsely believed myself to be seeing a dog” (Goldman 2009, p. 80). Goldman thinks that in this case the existence of the wolf does not prevent the subject from knowing what she actually sees—the dachshund—is a dog, for a wolf does not really look like a dachshund. This motivates Goldman to develop method-individuation according to a subject’s internal experience. I argue in the next subsection that this approach fails as well.
For Goldman (1976) this method-individuation serves his relevant alternative theory of perceptual knowledge.
For Nozick this method-individuation is designed for his sensitivity principle. Here is how he specifies the method: “Usually, a method will have a final upshot in experience on which the belief is based, such as visual experience, and then (a) no method without this upshot is the same method, and (b) any method experientially the same, the same “from the inside”, will count as the same method.” (Nozick 1981, pp. 184–185).
See Goldman (1976).
Incidentally, the implication of Artistic Barn goes beyond the current context. The case also renders Goldman’s (1976) famous discussion of perceptual knowledge problematic. According to Goldman, S perceptually knows that p only if there is no perceptual equivalent where p is false. The idea here is in line with Same Method (Internal Experience). ‘Perceptual equivalent’ is a possible situation where S has sufficiently similar perceptual experience that she cannot discriminate from her actual experience. In the original fake barn case, Henry has perceptual equivalents of his actual barn-experience, which are induced by those fake duplicates—duplicates that give sufficiently similar visual experiences. But once we move on to Artistic Barn, clearly there will be no perceptual equivalents for Henry, although we would still judge that he has no perceptual knowledge. Clearly, something is lacking in Goldman’s analysis.
See, e.g. Pritchard (2005).
Thanks to an anonymous referee for pressing on this point.
My discussion here has focused on Lucy’s animal-belief, which is initially formed at t1. What about her dachshund-belief formed at t2? Presumably, the current proposal would claim that the method for dachshund-belief is ‘dachshund-classification’, which is based on perceptual appearance that gives rise to ‘dachshund gestalt’ (as opposed to ‘animal gestalt’). According to the current proposal, this is not the same type of method as the method in the weasel scenario, which is ‘animal-classification method’ involving ‘animal-gestalt’. But be that as it may, this does not affect that at t2 Lucy’s animal-belief is still held on the basis of the same method as in t1. Forming a new belief does not mean the method of one’s previously formed belief is changed. Since that method of animal-belief at t1 involves ‘animal gestalt’—which, as argued, can be similar to the method in the weasel scenario—Lucy’s animal-belief is still unsafe at t2.
One may be tempted to argue here that two method tokens are of the same type only if they are caused by objects of the same kind. Furthermore, the dachshund and the fake weasel are not of the same kind. Therefore, Lucy’s animal-belief formed by perceiving the dachshund is safe.
Here is my response. Granting that the above reasoning is plausible, what shall we say about the fake barn case? Are the genuine barn and the fake barns of the same kind? I think the answer is negative. Given that fake barns are paper facsimiles which share very few similar properties with the genuine barn, it seems quite a stretch to say that they are of the same kind. Most plausibly, just as a dachshund and a fake weasel belong to different kinds, a genuine barn and fake barns also belong to different kinds. As a result, appealing to the talk of ‘kinds’ doesn’t help to predict that Lucy’s animal-belief is safe whereas Henry’s barn-belief is unsafe.
Additionally, metaphysically determining whether different objects are of the same kind is itself a matter of significant controversy. It would be an undesirable result for an epistemological principle of method-individuation to rely on it.
Thanks to Katherine Sweet for helpful discussion on this point.
See, e.g. Bogardus and Marxen (2014, p. 325).
Perhaps the only lesson (2*) could tell us is that a method should not be individuated in a way that involves exclusively the specific belief that is formed by a subject. But this is hardly a useful lesson since it still leaves wide open the question of how to individuate the method.
Grundmann argues in his paper that his two constraints make it the case that in the fake barn example, Henry’s method should not be “seeing that specific barn” because individuating method this way would violate both of his constraints. However, he does not go ahead and discuss what Henry’s method should be. My discussion here shows that once we try to tackle this latter question, it becomes obvious that his constraints are implausible.
One may insist that although the environment is the same, this same environment is epistemically unfavorable relative to Lucy’s animal-belief formed by perception. Relative to her animal-belief formed by competent deduction, however, the environment is epistemically benign.
My response is that, admittedly, the same environment could be unfavorable to a belief formed by one faculty, X, but not unfavorable relative to the same belief formed by another faculty, Y. But presumably this is because the environment would affect one’s performance of X but wouldn’t affect the performance of Y. For instance, imagine that two subjects, A and B, form the same belief that there is a bird. A forms the belief by vision whereas B forms it by hearing the bird’s chirping. Further suppose that both A and B are situated in the same dark environment. Plausibly, this dark environment is unfavorable relative to A’s believing that there is a bird, but it is not unfavorable relative to B’s believing the same proposition by hearing. Why? Because the darkness affects performance of A’s cognitive faculty, i.e. vision, but it does not affect B’s faculty, i.e., hearing. However, there is no such difference between Lucy’ animal-belief in Dachshund-Weasel I and the same belief in Dachshund-Weasel II. In the latter example, the existence of the weasel facsimile doesn’t affect Lucy’s seeing the dachshund clearly and forming a true belief that there is an animal. Similarly, in the former case, the weasel facsimile doesn’t prevent Lucy from competently inferring that there is an animal. Thus, as far as the epistemic environment is concerned, I see no good reason to treat these cases differently.
Thanks to John Greco for pressing me on this issue.
Determining the correct version of closure is a tricky issue, but this need not bother us here. The case can be easily adjusted in order to meet a form of closure that one favors. For an overview of different versions of closure, see Baumann (2011).
See, e.g. Nozick (1981).
This can certainly be true if one considers these two propositions: “there is a dachshund” and “there is a mammal.” Although the former entails the latter, for many people these two may not be quite related to each other, in the sense that while believing the former proposition one may fail to infer the latter one.
This is an interesting empirical question, though.
Other details should be added to Closure in order to have a sufficiently plausible form of closure. For example, (3) does not exclude the possibility that one’s belief is overdetermined by both deduction and improper reasons. Therefore, another premise such as “no defective reason” should be added to Closure. Since in our case Lucy’s beliefs are not going astray in this way, I will set these subtleties aside. For more discussion on Closure, see David and Warfield (2008, pp. 149–150).
Here is an example: “…I know that there are at least nine people in the room (I started counting and stopped at nine), and I know that this entails that there are at least seven people in the room, and I believe that there are at last seven people in the room but not because of the inference or the counting but rather because I always believe that there are at least seven people in the room when I am in the room. Though the example is somewhat frivolous, it should be clear that, in this sort of situation, where one does not base believing q on proper grounds for believing q, it looks like one does not know q” (David and Warfield 2008, p. 147).
It may also be helpful here to consider the following plausible form of closure:
Closure**:
(1) Kp
(2) K(p entails q)
So,
(C) S is in a position to know q.
It has been notoriously difficult to interpret the phrase “being in a position to know.” But this much seems obvious:
Other things being equal, a subject who satisfies (1) and (2) is in a significantly positive epistemic status with respect to knowing q.
In our counterexample, the fact that Lucy knows there is a dachshund, plus her background knowledge that dachshunds are animals, renders her a significantly positive epistemic status with respect to knowing that what she sees is an animal. This, coupled with the fact that Lucy properly bases her animal-belief on perception, seems enough to guarantee our intuition that at t2 Lucy knows what she sees is an animal. Our intuition doesn’t hinge on whether she actually undergoes the inferential process or not.
Greco (2016) interprets this form of safety to be the one that Sosa argues for in a number of recent works, including Sosa (2007, 2015). For my purposes here, I shall put aside the issue of interpreting Sosa’s safety and focus on Greco’s formulation of virtue-theoretic safety. Thanks to John Greco for discussion on this point.
See Greco (2016, p. 52).
See Williamson (2009).
For Williamson, safety does lots of theoretical work, such as rejecting the KK principle. However, this is at most the consequence of his safety principle. Such a consequence does not exempt Williamson from the burden of explaining in the first place why it is safety, instead of other theories (e.g. process reliabilism, sensitivity, etc.), that should be considered as a necessary condition of knowledge. It seems to me that Williamson does not say much in this regard.
Thus, we may have the same idea about similarity orderings, but disagree on which of those similar worlds are relevant for determining safety. For example, in Dachshund-Weasel II, presumably all the parties would agree that the possible world where Lucy sees the weasel facsimile is a close/similar one, since compared to Lucy’s actual situation, very little needs to be changed in that world. But because of difference of context, we may disagree about whether the weasel facsimile world is relevant for determining the safety of Lucy’s animal-belief.
The proposal here is inspired by Heller’s (1995) solution to the generality problem for process reliabilism.
References
Alfano, M. (2009). Sensitivity theory and the individuation of belief-formation methods. Erkenntnis,70(2), 271–281.
Baumann, P. (2008). Is knowledge safe? American Philosophical Quarterly,45, 19–30.
Baumann, P. (2011). Epistemic closure. In Duncan Pritchard & Sven Bernecker (Eds.), The Routledge Companion to Epistemology (pp. 597–608). Abingdon: Routledge.
Baumann, P. (2014). No luck with knowledge? On a Dogma of Epistemology. Philosophy and Phenomenological Research, LXXXIX (3), 523–551.
Baumann, P. (2016). Epistemic Contextualism: A Defense. New York: Oxford University Press UK.
Bogardus, T. (2014). Knowledge Under Threat. Philosophy and Phenomenological Research,86(1), 289–313.
Bogardus, T., & Marxen, C. (2014). Yes, safety is in danger. Philosophia,2, 1–14.
Broncano-Berrocal, F. (2014). Is safety in danger? Philosophia,42(1), 1–19.
Comesaña, J. (2005). Unsafe knowledge. Synthese,146, 395–404.
David, M., & Warfield, T. (2008). Knowledge-closure and skepticism. In Quentin Smith (Ed.), Epistemology: New essays. Oxford: Oxford University Press.
Freitag, W. (2014). Safety, Sensitivity and “Distant” Epistemic Luck. Theoria,80, 44–61.
Goldman, A. (1976). Discrimination and perceptual knowledge. Journal of Philosophy,73(November), 771–791.
Goldman, A. (2009). Williamson on knowledge and evidence. In Patrick Greenough, Duncan Pritchard, & Timothy Williamson (Eds.), Williamson on knowledge (pp. 73–91). Oxford: Oxford University Press.
Greco, J. (2007a). Worries about Pritchard’s safety. Synthese,158, 299–302.
Greco, J. (2007b). The nature of ability and the purpose of knowledge. Philosophical Issues,17(1), 57–69.
Greco, J. (2010). Achieving Knowledge: A Virtue-Theoretic Account of Epistemic Normativity. Cambridge: Cambridge University Press.
Greco, J. (2016). Knowledge, Virtue and Safety. In Performance Epistemology (Ed.), Miguel Ángel (Fernández ed.). New York: Oxford University Press.
Grundmann, T. (forthcoming). Saving safety from counterexamples. Synthese. https://doi.org/10.1007/s11229-018-1677-z.
Hawthorne, J. (2004). Knowledge and lotteries. Oxford: Oxford University Press.
Heller, M. (1995). The simple solution to the problem of generality. Noûs,29(4), 501–515.
Hiller, A., & Neta, R. (2007). Safety and epistemic luck. Synthese,158, 303–313.
Kelp, C. (2009). Knowledge and safety. Journal of Philosophical Research,34, 21–31.
Kelp, C. (2016). Epistemic Frankfurt Cases Revisited. American Philosophical Quarterly,53(1), 27–37.
Lewis, D. (1973). Counterfactuals. Oxford: Blackwell.
McEvoy, M. (2009). The lottery puzzle and Pritchard’s safety analysis of knowledge. Journal of Philosophical Research, 34, 7–20.
Nozick, R. (1981). Philosophical explanations. Oxford: Oxford University Press.
Pritchard, D. (2005). Epistemic luck. Oxford: Oxford University Press.
Pritchard, D. (2007). Anti-luck epistemology. Synthese,158, 277–297.
Sosa, E. (1999a). How must knowledge be modally related to what is known? Philosophical Topics,26(1/2), 373–384.
Sosa, E. (1999b). How to defeat opposition to Moore. In J. Tomberlin (Ed.), Philosophical Perspectives 13: Epistemology (pp. 141–154). Oxford: Blackwell.
Sosa, E. (2000). Skepticism and contextualism. Philosophical Issues,34, 1–18.
Sosa, E. (2002). Tracking, competence, and knowledge. In P. Moser (Ed.), The oxford handbook of epistemology (pp. 264–286). Oxford: Oxford University Press.
Sosa, E. (2007). A Virtue Epistemology: Apt Belief and Reflective Knowledge. Oxford: Clarendon Press.
Sosa, E. (2008). Skepticism and perceptual knowledge. In Quentin Smith (Ed.), Epistemology: New Essays (p. 121). Oxford: Oxford University Press.
Sosa, E. (2015). Judgment and agency. Oxford: Oxford University Press.
Williamson, T. (2000). Knowledge and its limits. Oxford: Oxford University Press.
Williamson, T. (2009). Reply to Goldman. In Duncan Pritchard & Patrick Greenough (Eds.), Williamson on Knowledge (pp. 305–312). Oxford: Oxford University Press.
Acknowledgements
For helpful comments and discussions, I am grateful for Peter Baumann, Quan Jin, Minghe Li, Joe Salerno, Katherine Sweet, Chong Yuan, Yiling Zhou, two anonymous referees for Synthese, and especially, John Greco.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Zhao, H. Knowledge without safety. Synthese 197, 3261–3278 (2020). https://doi.org/10.1007/s11229-018-1881-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11229-018-1881-x