Abstract
This paper argues that reliabilism can handle Gettier cases once it restricts knowledge producing reliable processes to those that involve a suitable causal link between the subject’s belief and the fact it references. Causal tracking reliabilism (as this version of reliabilism is called) also avoids the problems that refuted the causal theory of knowledge, along with problems besetting more contemporary theories (such as virtue reliabilism and the “safety” account of knowledge). Finally, causal tracking reliabilism allows for a response to Linda Zagzebski’s challenge that no theory of knowledge can both eliminate the possibility of Gettier cases while also allowing fully warranted but false beliefs.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
If the Gettier problems are indeed solved by appeal to the presence of false lemmas (see Levin (2006, pp. 381–392) for an argument that the standard Gettier cases all involve inferences from false lemmas) then, why not accept the “no false lemmas” account of knowledge instead of reliabilism? Because there are other problems which it cannot handle. One such problem is the barn façade case, since Barny’s belief is not based on any false lemma. Another is the odds-based lottery problem in which I infer from the one in a million odds against my ticket that my ticket will lose. This belief involves no false lemmas, but is not generally regarded as knowledge.
Why not, for clause 2, simply require that in most nearby worlds in which the process produces the belief that p its distal cause is p? Why, that is, is the reverse order “when the process takes the truth-maker for p as input, it outputs the belief that p” mentioned? This is in order to forestall cases in which the truth-maker for p is inputted to the process, but some non-p belief is outputted (e.g., a distal scene involving a horse is taken as input to the process, which leads S to form the belief that there is a donkey nearby).
Note that individuating processes by means of facts external to the cognizer’s psychology allows such cases, since one can type process tokens as being of the same type on the basis of their taking either p or a non-p perceptual equivalent of p as input (see McEvoy 2005a for more on this way of typing processes).
CTR1 is neutral (as are all subsequent CTR modifications) with respect to how broadly or narrowly we should construe process types. (i.e., it is not committed to one or another solution to the generality problem for reliabilism). I have offered a reliabilist solution to the generality problem, in McEvoy (2005a).
A further, perhaps less far-fetched case: in a large room there are one million plants. All but one of these is a common, well known plant, the other is a very rare plant, extremely similar in appearance. Each plant is numbered. Outside the room is a lottery machine which spits out a numbered ball at random. Every day an expert botanist visits, and receives a numbered ball (and then replaces it in the device), and a plant corresponding to that number is brought outside, and she (correctly) identifies it as the common plant. It so happens that in all the years she does this, the number corresponding to the rare plant is never chosen. Were it to be chosen, the botanist would misidentify it as the common plant. Surely the never-actualized rare, but modally close, alternative does not imply that she, despite her expertise, never had knowledge of the plants.
For another case of knowledge where there is modally close error, consider Baumann’s (2008, p. 20) variation of Nozick’s (1981, p. 193) Jesse James case. Two rival bank-robbers, Nogood and Gottit, have their faces displayed on “wanted” posters; they never work together. When they work, they wear masks disguising themselves as one another. One day Frank sees what he takes to be Nogood exiting a bank with the loot. As it happens, it is Gottit, but his mask slips to reveal his true identity. Frank believes—and knows—that Gottit has robbed the bank, but his belief is not safe (since there are close worlds in which Frank believes that Gotitt has robbed the bank when it was really Nogood in his Gottit mask).
Of course, the question of exactly how rare the possibility of rare error must be is not one for which we can expect a precise answer.
Though they clearly involve some element of luck, neither the modified barn case, nor the Android Stephanie case are cases in which the subject’s belief is true merely as a matter of luck. The beliefs in each case are too secure to be just lucky; they are causally connected to the truth-makers for the facts, they are produced by reliable processes, and those processes take the relevant truth-makers as inputs. Moreover, if the Android Stephanie and Botanist cases (and Baumann’s revised Jesse James case) above are correctly diagnosed, whatever form of luck is at play here is compatible with knowledge.
One might object, as did a referee for Synthese, that the belief in the earthquake is caused by its truth-maker, in a way that might be problematic for CTR2. This can be seen most clearly using the counterfactual theory of causation. According to it, c is the cause of e provided that if c had not happened, then neither would e. And here, if we take c to be the earthquake, and e to be the belief in the earthquake, then, c indeed seems to cause e. This does not create a problem for CTR2, however, since it requires more than that a belief be caused by its truth-maker. It requires that there be a reliable causal link between belief and truth-maker, such that when the process takes an effect of the earthquake as input, it reliably produces the belief. And to satisfy this, we must include the content of the newspaper. To see this consider the world in which the shelf from which the newspaper falls contains a collection of newspapers from years past that reported local disasters. Though S, upon awakening, in fact sees a newspaper reporting an earthquake, had the impact been only slightly different, a different newspaper, reporting a different disaster, would have faced S upon awakening. If the mere presence of a newspaper on the floor is to be taken as the effect inputted to the process, then this would not cause the belief in the occurrence of the earthquake, violating clause 2b of the definition of causal tracking. Thus, the content of the headline (and not merely the presence of the newspaper on the ground) must be taken as (part of) the effect of the earthquake that is inputted to the process to yield the belief.
This is required in order to allow for empirical knowledge from a mixed set of premises, some of which are known empirically, some of which are known a priori.
The process of testimony, on this view, thus involves a causal chain most of which is external to the subject who forms the testimonial belief. This is another manifestation of CTR3’s departure from reliabilist orthodoxy.
One might wonder whether belief that p arising from reliable testimony might be regarded as an effect of p, thus obviating the need to revise CTR2. One reason to doubt the feasibility of this suggestion is that there may be cases of testimony that do not satisfy the definition of causal tracking. This definition would require that one reconstruct the relation between testimony and the event reported, or else that one have previous experience of similar events having brought about similar reports. But imposing this requirement on testimony would appear to rule out cases of knowledge where one is told that p only once, and where one does not actually reconstruct the causal relation between event and report, but rather simply accepts the report due to the reliability of one’s informant. The move to CTR3 circumvents these problems.
We might still hold that Smith’s belief (and the beliefs of others in Getter cases) is epistemically blameless, or even that it is (in some internalist sense) justified. It is, of course, not warranted, if we take warrant to be what is added to true belief to get knowledge.
And, of course, there is no causal relation, of the type required by clause 2, between the evidence for Jones owning the Ford and Brown being in Barcelona.
See McEvoy (2012) for more detail on this, and on CTR3’s treatment of the lottery problem.
Zagzebski subsequently came to favor the option that warrant implies truth, and that there therefore cannot be false but warranted beliefs. See Zagzebski (1999, p. 111).
CTR3 does allow that there can be knowledge in cases in which there is rare, but modally close, counterfactual error, but as I argued above, this is as it should be.
One objection that arose for CTK—that it could not account for mathematical knowledge, since there are no causal connections between our mathematical beliefs and the facts they reference—can be dealt with rather quickly. CTR3 is a theory of empirical knowledge. Mathematical knowledge is plausibly construed as a priori knowledge, and thus will not be captured by CTR3. For a reliabilist account of a priori mathematical knowledge see McEvoy (2004 December) and McEvoy (2005b).
Note that this does not allow me to know that my lottery ticket has lost merely by considering how the causal mechanism of the lottery will almost certainly result in my ticket losing. For here, unlike in the assassin case, the operation of the lottery mechanism is not sufficient to cause my ticket to lose (thanks to a referee for Synthese for raising this issue).
Consider a preemption case in which the preempted cause would not suffice to bring about the effect in question. Suppose I see Paul take a fast acting poison. I then leave the scene and, shortly afterwards, form the belief that Paul is dead. Unbeknownst to me, after I had left, Paul had taken an antidote for the poison, but, due to a rare allergy to the antidote, died instantly. Because Paul had access to an antidote, which absent the allergy, would have prevented the preempted cause from bringing about his death, I do not know that Paul is dead.
References
Baumann, P. (2008). Is knowledge safe? American Philosophical Quarterly, 45(1), 19–30.
Gettier, E. (1963). Is justified true belief knowledge? Analysis, 23, 121–123.
Goldman, A. (1967). A causal theory of knowing. Journal of Philosophy, 64, 357–372.
Goldman, A. (1976). Discrimination and perceptual knowledge. Journal of Philosophy, 73, 771–791.
Goldman, A. (1986). Epistemology and cognition. Cambridge, MA: Harvard University Press.
Greco, J. (2007). The nature of ability and the purpose of knowledge. Philosophical Issues, 17, 57–69.
Greco, J. (2010). Achieving knowledge: A virtue-theoretic account of epistemic normativity. Cambridge: Cambridge University Press.
Lackey, J. (2007). Why we don’t deserve credit for everything we know. Synthese, 158, 345–361.
Lackey, J. (2008). Learning from words. Testimony as a source of knowledge. NewYork: Oxford University Press.
Lackey, J. (2009). Knowledge and credit. Philosophical Studies, 142, 27–42.
Levin, M. (2006). Gettier cases without false lemmas? Erkenntnis, 64, 381–392.
McEvoy, M. (2004). Is reliabilism compatible with mathematical knowledge? Philosophical Forum, 35(4), 423–439.
McEvoy, M. (2005a). Belief-independent processes and the generality problem for reliabilism. Dialectica, 59, 19–35.
McEvoy, M. (2005b). Mathematical apriorism and warrant: A reliabilist-platonist account. Philosophical Forum, 36(4), 399–417.
McEvoy, M. (2012). Causal tracking reliabilism and the lottery problem. Grazer Philosophische Studien, 86, 73–92.
Nozick, R. (1981). Philosophical explanations. Cambridge, Mass.: Harvard University Press.
Pritchard, D. (2005). Epistemic luck. Oxford: Oxford University Press.
Sosa, E. (2000). Skepticism and contextualism. Philosophical Issues, 10, 1–18.
Sosa, E. (2007). A virtue epistemology. Oxford: Oxford University Press.
Williamson, T. (2000). Knowledge and its limits. Oxford: Oxford University Press.
Zagzebski, L. (1994). The inescapability of Gettier problems. Philosophical Quarterly, 44, 65–73.
Zagzebski, L. (1999). What is knowledge? In J. Greco & E. Sosa (Eds.), The Blackwell guide to epistemology (pp. 92–116). Oxford: Blackwell.
Acknowledgments
Thanks are due to two anonymous referees for Synthese, whose comments—too numerous to acknowledge individually (except for some particularly significant revisions )—occasioned a substantial rewriting of this paper. Thanks also to Peter Baumann for some helpful discussion of non-safe knowledge.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
McEvoy, M. Causal tracking reliabilism and the Gettier problem. Synthese 191, 4115–4130 (2014). https://doi.org/10.1007/s11229-014-0523-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11229-014-0523-1