The Inescapability of Gettier Cases

Ever since Gettier famously highlighted a fault in the concept of knowledge by showing justified, true, belief (JTB) as jointly insufficient for knowledge, there have been many attempts to fix it. Arguably, none have. In Linda Zagzebski’s paper The Inescapability of Gettier Problems she claims that the approach to add or refine the “justification components” of JTB+/TB+ accounts are doomed to fail. So long as truth is not entailed by justification, which is to say it is possible to have justified false beliefs, these approaches will never be immune to Gettier cases. In this essay I reconstruct Zagzebski’s convincing argument and apply it to Nozick’s theory of knowledge which claims to have dealt with Gettier cases. I argue that one way to understand Zagzebski’s argument is as manifestation of a broader tension between our fallibility as knowers and our desire for knowledge to infallible. Finally, using ideas developed throughout this essay I suggest ways forward for the field of epistemology as it struggles to define its central object of analysis. 

Zagzebski begins her analysis of Gettier cases with a discussion of the role of luck. She writes, “[w]hat generates the problem for JTB, then, is that an accident of bad luck is canceled out by an accident of good luck. The right goal is reached, but only by chance.” (pg 66). Smith, our regular epistemic patsy, has an accident of “bad luck” that breaks the link between justification and truth. In the Gettier case of Brown in Barcelona, Smith’s justification for the proposition “John owns a Ford or Brown is in Barcelona” is that John told him he owns a Ford. Smith’s epistemic “bad luck” in this case is that John has lied, which is unlike his otherwise trustworthy nature. Smith forms a justified false belief. In the case of Fake Barns, Smith drives through the countryside, where unbeknownst to him, he is surrounded by fake barns. Although he points to a real barn and correctly says “this is a barn” he is unaware that his method of justification is not reliable in this situation. Therefore the justification and the truth have come apart. However after the accident of “bad luck” Smith gets a stroke of “good luck.” Brown happens to be in Barcelona, making the disjunction “John owns a Ford or Brown is in Barcelona” true. Smith luckily picks out the one real barn in a sea of fake barns. For reasons completely independent of the justification, the accident of “good luck” fixes the belief. In Brown in Barcelona the justified false belief turns into a justified true belief. Even though Smith reached the “right goal,” because it was “only by chance”  we do not want to grant these as cases of knowledge. Knowledge as we understand it normativley and descrpiptivley, is more robust and not as contingent. Yet, so long as there is space between justification and truth, Zagzebski claims that all JTB+ account can be hacked in this way. Because these accidents of “luck” cannot be fully systematically accounted for. 

Zagzebski’s argument turns on the claim that this “good luck” cannot be systematically accounted for. Even if the “justification component” blocked specif cases of “good luck” from consituting knowledge, we could always find another way to fix the belief. As long as there is space between justification and truth, further specification or additional criteria doesn’t help. Zagzebski demonstrates the generality of her claim by walking us through a recipe to generate Gettier cases for Alvin Plantinga’s new JTB+ account where he replaces the concept “justification” with the one of “warrant.” Plantinga’s component of justification, “warrant,” attempt to block Gettier cases where the environment is abnormal (i.e cases of lying or rouge modern artists with a predilection for fake barns). Yet this further refinement doesn’t help, since all we have to do is artificially construct a “warranted false belief” (which is possible by his admission) and creativley fix it with “good luck.” This is exactly what Zagzebski does. Mary has good eyesight and often sees her husband sitting in the chair inside the living room. Under the same conditions of lighting, regular functioning eyesight, no abnormal environment (no one is dressing up as her husband), and where she has every reason to believe her husband’s brother is in Australia, she forms the warranted false belief “my husband is sitting in the living room” when it in fact her husband’s brother in the char. Now let’s create some of the “good luck” that resists sytematization: her husband is actually sitting in the living room out of sight from Mary. We have now produced a Gettier case of warranted, true belief that is accidental. Zagzebski claims regrdless of the JTB+ account we always will be able to.

A helpful analogy to understand Zagzebski’s argument is to think of countable and uncountable infinite sets. A set is countably infinite if we can write it down in some ordering. There is a mapping from the natural numbers to the set. A set is uncountably infinite, such as the set of real number, if no such mapping is possible. In this case, the number of ways “good luck” can make fix a justified belief that has come apart from the truth is uncountably infinite. All possible necessary conditions/“justification components” is countably infinite. Even without this mathematical analog it is clear that given any arbitrarily number of “justification components” we can always produce another accident of “good luck.” So long as justified and truth can come apart occur due to “bad luck” there will always be uncountably “more” accidents of good luck we need only to get creative to find. Therefore, Zagzebski concludes Gettier cases are inescapable for JTB+ accounts. Gettier problems are not inescapble for all account of knowledge however. “Since justification does not guarantee truth, it is possible for there to be a break in the connection between justification and truth, but for the connection to be regained by chance.” (pg 65). Both Gettier and Zagzebski explicitly focus on accounts of knowledge where justification does not entail truth. With our back against the wall, one might ask why not close the gap between justification and truth, to raise the epistemic bar of justification or a similar concept to a degree that it guarantees truth. However, justification and its related concepts are precisely what helps us determine what is true and what is false. So to force p to be true all cases where S is justified in believing that p drains the concept of its use, and to render the “truth” condition on knowledge redundant. Such approaches is certainly possible, but these are not targeted by Zagzebski argument.

Zagzebski’s claim is abstract but operates concretely. Give me your JTB+ account and I’ll give you a Gettier cases that beats it. Nozick’s famous (failed) TB+ account of knowledge that ostensibly blocked Gettier cases presents an interesting example of this dialectic. Nozick’s account of knowledge as “belief that tracks the truth” linked truth and justification through counterfactual dependence. Using the subjunctive condition we ask if it were not the case “John owned a Ford or Brown was in Barcelona” (p), would Smith believe it to? In other words is Smith’s belief sensitive to the truth (stability condition) and falsity (sensitivity condition) of the proposition believed. In the the Brown in Barcelona case, Smith would continue to falsely belief p because his justification was based on John’s faulty testimony. Smith’s belief is insensitive to the truth of p, therefore not knowledge. However, Saul Kripke’s takedown is as devastating as it is illustrative. One of the the most famous of his recipes that broke down Nozick’s account was to vary an accidental feature of the visual experience. In the case of the Fake Barns, if all real barns are painted red then under Nozick’s account Smith knows that the barn is red but doesn’t know that it is a barn. This ridiculous failure of closure of knowledge under known entailment is known as the abominable conjunction. Although the “good luck” in this case operates differently than in the examples Zagzebski uses, it shows how “good luck” circumvents the “justification component” of Nozick’s account. Smith had the good luck that the barns were color coded, and this hacked the condition It made his belief sensitive so that in nearby possible worlds where it was false, he wouldn’t believe it and in nearby possible worlds where it is true, he would. Had Nozick come up with another justification component, we would be able to generate a different Gettier case, with luck playing a different role. Even though Nozick’s justification looks very different to other TB+ accounts, Nozick’s still failed to escape Gettier cases because the additional senstivity and stability conditions that tested the relation between justification and truth, didn’t guarantee truth. Luck intervened.

If we accept Zagzebski’s claim about the inescapability of Gettier cases in TB+ accouts, what are we to do with them? I believe these attempts are valuable even if they failed to by their own standard. Different “justification component” systematize an intuition we want knowledge to have. Even if we give up specifying a set of jointly sufficient conditions on knowledge there is still value in specifying necessary conditions. Kripke never clarified his thoughts on if counterfactual dependence should remain a necessary condition on knowledge. I think the silence on the point says something. Nozick captured a powerful normative intuition about knowledge, which even if it breaks in unacceptable ways when it tries to claim sufficency for knowledge it may still be useful when assessing knowledge claims as necessary conditions. Similarly other failed attempts such as reliabalism or “warrant” may work better in different contexts.

Zagzebski argument captures a very broader tension in epistemology between our desire for knowledge to be infallible but recognition that our methods for acquiring it aren’t. David Lewis discusses falliabalism in his Elusive Knowledge, where he lays out a lucid context-dependent account of knowledge. He writes “… it seems as if knowledge must be by definition infallible. If you claim that S knows that P, and yet you grant that S can not eliminate a certain possibility in which not-P, it certainly seems as if you have granted that S does not after all know that P. To speak of fallible knowledge, of knowledge despite un-eliminated possibilities of error, just sounds contradictory.” (pg) Yet in spite of saying this and his later claim that contextualism provides a way out of both skepticism and fallibilism, his theory of context is one of proper ignoring. In almost all contexts, we are allowed to properly ignore certain possibilities, and therefore by Lewis’ own words allowing knowledge to be falliable. Only in the strictest context (i.e an epistemology seminar) do all not-P possibilities have to be eliminated including the skeptical hypothesis. In this context, knowledge is infaliable, but we also know nothing about the contingent external world. Lewis write, “maybe ascriptions of knowledge are subtly context-dependent, and maybe epistemology is a context that makes them go false. Then epistemology would be an investigation that destroys its own subject matter.” His tacit captiulation to falliabalism as context dependendent opens up ways in which JTB+ analyses my be helpful. What if contexts could be specified by necessary conditions for knowledge claims within them. If contexts grant a degree of fallibalism to knowledge it is the task of the epistemologist to systematize and understand that degree of falliability best they can.

Zagzebski argument demonstrates how the “luck” generates Gettier cases so long as there is any space between justification and truth. This luck resists being systematized. Even if an aspect of it was systematized and therefore blocked by a “component of justification” in a JTB+ account, we could always produce another that circumvents it. Looking forward then in our analysis of knowldege I suggest focused on a change of tact. Instead of focsuing on sufficient conditions for knowledge we focus on the neccessary ones, there is space for useful analyses of knowledge and context. Our creativity to construct cases of “good luck” and specify necessary conditions on knowledge that specify links between justification and truth that are important, can be immensly useful when tied to context. If we accept what Lewis tacitly does, namely that knowledge is fallible on the basis of context, Gettier cases become useful to generate concepts of spurious justification which we can decide as relevant or not in given contexts. Gettier cases aren’t escaped, but at least they aren’t “problems” anymore.