Progress in philosophy: the Gettier case

by Massimo Pigliucci

As I’ve mentioned now a number of times, I’m in the midst of writing a book for Chicago Press on whether and how philosophy makes progress. The short answers are: yes, but in a way that is different from how science makes progress. Specifically, I will be arguing that philosophy is concerned with possibilities in logical (as opposed to empirical) space, and that therefore progress is measured by the elimination of logically flawed positions and the refinement of logically sound alternatives. Since empirical space broadly underdetermines logical space, this also means that in some sense doing science is actually easier than doing philosophy — or at the very least that there are good reasons why philosophers don’t usually settle on one “correct” theory of anything.

This is all well and good, you might say, but can you please give an example or two? Yes, I can. In this post I will explore a particular instance of what I consider progress in philosophy, and specifically in epistemology — the full book will contain several more, drawn from subfields as varied as ethics and philosophy of science.

The example in question has to do with nothing less than the very definition of knowledge, and in particular regards what happened in that area of epistemology in the aftermath of a short paper (three pages) published by now retired University of Massachusetts at Amherst’s Edmund Gettier back in 1963 (as it turns out, he wrote it in order to get tenure, and it is the only paper he published in his entire philosophical career!).

The following may become a bit technical (and you may want to read other stuff about it, like this recent short article by Fred Dretske at The Philosophers’ Magazine), but bear with me, it's worth it.

So, beginning on the left side of my conceptual space, we start with the standard definition of knowledge — going back to Plato — as Justified True Belief (JTB). This means that for something to count as knowledge, the epistemic agent’s (e.g., you) belief about a certain matter has to be both true and (rationally) justified. For instance, let’s say you believe that the earth goes around the sun, rather than the other way around. This belief is, as far as we can tell, true [1]. But can you justify it? That is, if someone asked you why you hold  that belief, can you actually give an account of it? If yes, congratulations, you can say that you know that the earth goes around the sun. Otherwise you are simply repeating something you heard or read somewhere else. (Which, pragmatically, is fine. It just doesn’t count as knowledge, that’s all.)

Now, the above approach was fine for about two and a half millennia, until some people — like Bertrand Russell — began questioning it and thinking about its limitations. But the big splash on the knowledge thing was the above mentioned short paper by Gettier. Because the problem posed by Gettier may not sound that impressive the first (or second) time you’ll actually look at the details (as we will do in a minute), let me remind you that, for instance, John L. Pollock and Joseph Cruz are on record as saying that it has “fundamentally altered the character of contemporary epistemology” and that it constitutes “a central problem of epistemology since it poses a clear barrier to analyzing knowledge.” So there.

Let’s get started with a typical “Gettier case,” that is a hypothetical situation that seems to be an exception to the JTB conception of knowledge. Let’s say I get letters, copies of utilities bills and other documents from my friend Phil, and they all refer to a residence in New York City, state of New York. I would be justified in believing that Phil lives in New York City. If Phil lives in NYC, then it is also true that Phil lives in the State of New York, and consequently I believe that too. Turns out, however, that Phil actually lives on Long Island. So my first belief about Phil was simply wrong. This presents no problem for the JTB account, since my belief only satisfied one of the two conditions (it was justified, but not true). The trouble comes when we assess my second belief, that Phil lives in the State of New York. I am correct, he does. That belief of mine is both true, and justified (logically, given the premise that Phil lives in NYC). But now we have a case of justified true belief that is actually based on false premises, since Phil does not, in fact, live in New York City. [You may want to re-read this before proceeding, first exposure to Gettier cases is often confusing, but once you get it, it hits you like a brick!]

Gettier cases have the general form of the example I just gave: they get off the ground because they are about logically entailed truths that happen to be derived via a belief that is justified but not true. The problem they pose is not with the first belief (the one that is justified but not true) but with the second belief (the one that is logically entailed by the first one, and which happens to be true). Now what?

The first response — the first move in logical space after Gettier’s own — was for epistemologists to seize on the fact that Gettier cases depend on the presence of false premises and simply amend the definition of knowledge to say that it is justified true belief that does not depend on false premises (the “no false lemma” solution, see figure). As it turns out, however, one can easily defeat this move by introducing more sophisticated Gettier cases that do not seem to depend on false premises, so called general Gettier-style problems.

Here is one possible (if a bit contrived) scenario: I am walking through Central Park and I see a dog in the distance. I instantly form the belief that there is a dog in the park. This belief is justified by direct observation. It is also true, because as it happens there really is a dog in the park. Problem is, it’s not the one I saw! The latter was, in fact, a robotic dog unleashed by members of the engineering team from Bronx High School. So my belief is justified (it was formed by normally reliable visual inspection), true (there is indeed a dog in the park), and arrived at without relying on any false premise [2]. And yet, we would be hard pressed to call this an instance of “knowledge.” It’s more like a lucky coincidence.

There is a move that can be made by supporters of the no false lemma solution to repair their argument, for instance adding that the epistemic agent needs to (consciously or even unconsciously) consider the possibility of both deception and self-deception, claiming knowledge only when those have been ruled out. The problem with that solution is that if we accept this then it turns out that we hold to a lot fewer justified beliefs than we think.

A related, but distinct, move, is to say that Gettier cases are not exceptions to JTB because it does not make sense to say that one can justify something that is not true. That may be, but this moves the discussion away from the concept of knowledge and onto the concept of justification, which turns out to be just as interesting and complicated (and outside the scope of this post).

A completely different take is adopted by philosophers who have tried to “dissolve” rather than resolve the Gettier problem (lower portion of the concept map). Here there are at least two areas of logical space that can be reasonably occupied: the minimalist answer is to bite the bullet and agree that all cases of true belief, including accidental ones, count as knowledge. The good news is that we end up having much more knowledge than we thought; the bad news is that it seems we are now counting as “knowledge” the sort of lucky coincidences (see the dog example above) that are really hard to swallow for an epistemologist. A second way of dissolving the Gettier problem is to say that it gets wrong the concept of justification (again, shifting the focus of the discussion). For instance, one could say that justification depends not just on the internal state of the epistemic agent, but also on how such state relates to the state of affairs in the external world (the dog is really a robot!). This means that we are now owed an account of why there may be a mis-alignment between internal and external states, or what makes a belief appropriate or inappropriate.

The center of my concept map refers to two broad categories of replies, one that adopts the strategy of revising the JTB approach itself, the second that aims at expanding it with a further, “G” (for Gettier) condition. Let’s start with possible modifications of JTB. One option was suggested by the above mentioned Dretske and separately by Robert Nozick, and is known as the “truth tracking” account: it basically says that the epistemic agent wouldn’t believe proposition P if P were not true. This immediately leads to the question of what accounts for agents having this or that belief, of course. A second modification of JTB is known as Richard Kirkham’s skepticism, and it is an acknowledgment of the fact that there will always be cases were the available evidence does not logically necessitate a given belief. This move in turn leads to a split: on the one hand one can simply embrace skepticism about knowledge and be done with it. On the other hand one can adopt a fallibilist position and agree that a belief can be rational even though it doesn’t rise to the lofty level of knowledge.

We now move to explore the last area of logical space opened up by discussions of Gettier problems: the so-called “fourth condition” family of approaches. One is represented by Alvin Goldman’s causal theory of belief, which says that it is the truth of a given belief that causes the agent to hold to a belief in the proper manner (an improper manner would fall back into Gettier-style cases). This again raises the issue of how we account for the difference between appropriate and inappropriate beliefs (the very same question we have seen raised by one of the dissolution approaches, the one that says that Gettier cases involve a wrong concept of justification, as well as by the Dretske-Nozick response). Goldman himself was happy to proceed by invoking some form of reliabilism about justification, a discussion of which will definitely lead us outside the territory being charted here.

Keith Lehrer and Thomas Paxson have advanced the possibility of defeasibility conditions: knowledge gets redefined as “undefeated” justified true belief. The problem here is that it is hard to get a good grasp on the concept of a defeater in a way that it doesn’t rule out well established instances of a priori knowledge (like logical and mathematical knowledge).

Finally, we have the pragmatic move: since truth is defined by pragmatists like Charles Sanders Peirce as the eventual opinion reached by qualified experts, we get that in most ordinary cases of “knowledge” we simply need to embrace a Socratic recognition of our own ongoing ignorance.

Now, you may be thinking: so, after all this, what is the answer to Gettier-style problems? What is the true account of knowledge? If so, you missed the point of the whole exercise. Unlike science, where we do seek answers to questions determined by empirical evidence, philosophy is in the business of exploring possibilities in logical space. There are often many such possibilities, since the constraints imposed by logic are weaker than those imposed by empirical facts. At the end of this sort of discussion we are left with the following: a) A much better appreciation for the complexities of the deceptively simple question: what is knowledge? b) An exploration of several possible alternative accounts of knowledge and related concepts (such as justification and belief); c) A number of options still standing, some of which may be more promising than others; and d) A number of possibilities that need to be discarded because they just don’t work when put under scrutiny. And that, my friends, is how philosophy makes progress [3].

__________

[1] Yes, yes, I know that it’s not really the case that the earth rotates around the Sun, as much as that the Sun occupies one of the foci of the elliptical orbit of our planet (the Newtonian account). Or better yet, that both Sun and Earth interact in a complex fashion because of their respective gravitational fields (the relativistic account). But all of this is largely irrelevant to the point I’m making here.

[2] Notice that the premise that my observations are reliable under normal conditions is not false.

[3] It’s also possible that you may be asking yourself which of the above options I think is the best. Well, my thinking at the moment is that JTB is ok as a first approximation — that is, any time you are not talking to professional epistemologists. Of the others, I’m ok with responses that invoke either fallibilism or reliabilism, which means that I can go either with Kirkham’s skepticism or with Goldman’s causal theory. Then again, philosophers of knowledge may not be done just yet exploring the relevant logical space...