Toward a Better Epistemology: Why Do We Care About What We Know?

A reaction to the ‘Gettier problem’

Digital art by Tucker Lieberman.

Why should we assume, with Socrates, that it is useful or even possible to give one single account of the many instances of knowledge? Why should we expect that each instance of knowledge is a branch off the main tree of Knowledge? What if, instead, each instance of knowledge is loosely bound to others by a Wittgensteinian “family resemblance”?


The study of the definition and meaning of knowledge has, since Socrates, been known as epistemology. And the project of epistemology, as traditionally conceived, is too restrictive. It is restrictive in at least three ways.

First, it tries to make a clean taxonomy of two broad types of knowledge: “propositional” knowledge, which is a statement we make about something, and “non-propositional” knowledge, which includes subtypes like ability (e.g. riding a bicycle, learning to swim) and familiarity (e.g. recognizing a person, navigating a neighborhood). It’s not actually possible to do this cleanly. We can’t always distinguish when our knowledge is propositional or not.

Bicycling and swimming are only truly learned by “doing,” yet we call out support to those who are learning, like: “Once the bicycle has momentum, you won’t fall” or “Look how the water supports your weight as you hold onto the wall with one hand!” When I immediately recognize a person who waves to me, I may have no need to articulate this, but if it takes me a few moments to recognize them, I may find myself constructing an argument or making some kind of mental exclamation, especially if I have to explain this process to someone else. The same goes for finding my way around my neighborhood.

Now, by contrast, consider writing and math, which we are often taught to think of as following a set of propositional rules. How much of that turns out to be less propositional and more about ability and familiarity — something we learn by doing! One needs non-propositional knowledge to be able to express what one knows propositionally. Just because something can be expressed propositionally does not mean that the proposition captures all the information there is to know, nor that it is the best way to convey the information.

A second common failure of epistemology is that it tries to cast all propositional knowledge in the same mold. It seeks a single set of necessary and sufficient conditions by which we can be certain that we know something propositionally. This is a misguided approach, because, in reality, there are many kinds of propositional knowledge. The category of “propositional knowledge” needs to be more fluid and less tightly encapsulated. Maybe “justified true belief” is indeed, as Socrates had it, their common definition — but then again, maybe it isn’t.

Third, traditional epistemology tends to neglect to emphasize the context in which we ask about knowledge.

The project of epistemology should not be narrowly focused on listing sufficient conditions for knowledge. Instead, it should strive for a more complex, integrated picture of human knowledge that represents our intelligence on a multi-dimensional level.

Gettier’s Argument

In a 1963 paper titled “Is Justified True Belief Knowledge?”, reprinted years later in Epistemology: An Anthology, Edmund Gettier (1927–2021) answered his own title question: No. A belief’s mere conditions of being “justified” and “true” were insufficient, he thought, to deem it knowledge. He questioned this received Platonic definition of knowledge because he could think of hypothetical examples of justified true beliefs that would not rise to the level of what we would commonly call knowledge.

In Gettier’s first example, two people apply for the same job. Candidate A has two justified beliefs: Candidate B will get the job and Candidate B has ten coins in their pocket. From these beliefs, Candidate A infers another belief: the successful candidate has ten coins in their pocket. As it turns out, Candidate A receives two surprises: they are offered the job, and they discover ten coins in their own pocket. It does turn out to be true, then, that the successful candidate has ten coins in their pocket. And, again, Candidate A was justified in holding this belief. But they didn’t know it. Their supporting information was that Candidate B would get the job (which turned out to be false) and that Candidate B had ten coins in their pocket (which turned out to be irrelevant). Such false and irrelevant information cannot be adequate justification for knowledge. Candidate A’s belief was only accidentally true.

In Gettier’s second example, a hypothetical person has a justified belief that their friend owns a car. Therefore, any proposition that begins with Either my friend owns a car, or… will be a justified belief, purely on logical grounds. No matter how outlandish the second option that completes the statement — say, that another friend has traveled to Barcelona, when there is no particular reason to believe that— the entire sentence nevertheless represents a justified belief. But let’s imagine that the friend turns out not to own a car, and the other friend just so happens to be in Barcelona. The proposition Either my friend owns a car, or my other friend is in Barcelona was a justified true belief, but, again, it was hardly knowledge, as the belief was justified in the wrong way. There was evidence supporting the idea that proved false; there was no evidence supporting the idea that proved true. These two unrelated ideas were combined into a single proposition only by a trick of logic: the “either…or…” formulation. And only by accident did that formulation happen to be, as a whole, true.

Again: In Gettier’s first example, the person has two justified beliefs, one of which (who will get the job) turns out to be wrong and the other of which (how many coins are in that person’s pocket) has a flexible meaning since, in the rephrased sentence, it describes whoever does get the job. In the second example, the person has only one justified belief, and the sentence is constructed such that any additional statement — even a false one — can be tacked onto it without falsifying the statement as a whole. If it turns out that the justified belief is false and the unjustified belief is true, the sentence as a whole remains true.

What’s Weird About These Examples?

I can’t imagine anyone formulating these propositions.

The successful candidate has ten coins in their pocket is something one might say to be coy. It’s a poetic representation that deliberately obscures the identity of the job candidates. One might refer to a king’s bodyguard as “The Sword of the King,” but in that case the image of the “Sword” is at least relevant to the job description, whereas pocket change has nothing to do with the outcome of a job interview.

An acquaintance might ask, “Do you think you got the job?” and the reply might be, “Oh, I’m sure the successful candidate will make that information public next week.” That’s primarily a way to deflect the conversation. It also happens to be a justified belief if the new hire is indeed expected to start the job next week. But it doesn’t represent knowledge of who is going to be hired. It just means the aspiring candidate doesn’t want to talk about it yet.

Referring to a person who has ten coins in their pocket is sort of like that. When you don’t want to think about the person by their name or other identifying characteristics, you think about them in a detached way such as by your information about what they have in their pockets. If you also happen to have ten coins in your pocket, that is very far down the road to irrelevance. The question was who would get the job, and your answer fixated on the other person’s success. If you get the job instead, your prediction turned out not to be true. So, in retrospect, you didn’t know.

Similarly, the proposition Either my friend owns a car or my other friend is in Barcelona isn’t a real-life proposition. Why would anyone ever say this? We would have to develop a story behind it. For example, the second friend may wish to go to an opera in Barcelona, and the question may be, if they are not already in Barcelona, whether the first friend will kindly drive them there. Suppose you have reasons to believe that your first friend owns a car, and you are wrong; still, your other friend does attend the opera because, happily, they were already in that city. You didn’t have knowledge that it would turn out this way. Again, your solution fixated on the wrong person. The possibility that, well, maybe the second friend is already in Barcelona was a throwaway speculation that you made, perhaps as a joke. It is funny that it turned out to be correct.

Objections like these are emblematic of philosophy’s disengagement from the world in favor of logical puzzles. The two propositions seem not to be in the real world of human thought. They arise for no meaningful reason and serve no purpose. If we do not come up with examples that are more human, we will not be able to evaluate what such an imagined person might have been hypothetically thinking, nor will we have the authority to hold this fictional person to our own standards. We can’t really say whether they are instances of human knowledge.

A Version of the ‘Ten Coins’ Proposition That is More Real-World

When I was considering “the Gettier problem” in college in 2001, I complained that Gettier’s “ten coins” example was “distractingly irrelevant.” I came up with my own version: structurally similar but with a little more emotional weight.

George H. W. Bush was elected President of the United States in 1988. Suppose a person had a justified belief that George H. W. Bush would be re-elected in 1992. That proposition would have entailed the following proposition: “In the near future, the United States will have a President named Bush.” That proposition turned out to be true, but only because George W. Bush, the son, was elected in 2000. George H. W. Bush, the father, was never re-elected.

It seems to me that anyone in 1992 who might have made a statement like this — in the future, the President will be named Bush — would have been poetic (intending to represent George H. W. Bush) or coy (deliberately creating ambiguity to heighten suspense). Either way, they would have had a specific Bush in mind. They were probably thinking of the father. If they were aware of other Bushes, had they been considering them as potential future candidates, and if they had wanted to share that information, they could have named them specifically. Or they might have said something like: There are several people in the Bush family who might plausibly become President someday. But the question wasn’t whether some Bush would someday be elected. The original question was whether the father would be re-elected and whether a particular person knows this.

A statement of belief like “The winning candidate has ten coins in their pocket” or “Someday, there will be another Bush presidency” may not be explicit and specific enough even to allow us to debate whether it is knowledge. Just because information can be phrased propositionally does not mean it can be known (propositionally or otherwise). It may have been given the appearance of a proposition, but it is not actually informative. There may not be enough substance there to be known and, therefore, it may be unable to change us.

A Version of the ‘Barcelona’ Proposition That is More Real-World — Or, Well, Not

We don’t usually make sentences in which we are aware that one of the disjuncts is irrelevant. Therefore, it is hard to revise this example to make it more meaningful.

Suppose I am expecting a dinner guest, and I have a justified belief that this guest likes tofu. I don’t say to myself: Either this guest will want tofu or some other guest will want carrots. The point is, I’m going to make tofu. That is what my properly invited guest wants to eat. Carrots (or whatever hypothetical, mysterious, uninvited guests do like to eat) are irrelevant. So I am just not going to construct a sentence like this.


After tinkering with Gettier’s argument about the insufficiency of the formulation of knowledge as a “justified true belief,” I conclude that we may not be missing a fourth sufficient condition for knowledge. Rather, it is just that some statements are so unimportant or irrelevant that they do not even prompt us to ask whether we “know” them.

When we attempt to cast “propositional knowledge” in a rock-solid mold defined by a single set of necessary and sufficient conditions, we place too much focus on the simple fact that a person is engaging with a proposition. We ought also to stop and reflect: Why are we asking whether that person knows that proposition? Why do we care? There may be no basis on which to judge our reasons for our inquiry, and perhaps no reason is better or more correct than any other, but nevertheless we ought to have some reason. This is the meta-epistemological question. It is the question that makes epistemology matter. Our reasons for asking about knowledge — that is, our motives for verifying facts and for assessing the quality of reasoning — give force and weight to our claim to know something.

“Propositional knowledge” is not unified by a Platonic Form of justification for believing a true statement. Epistemologists should not spend very much time pigeonholing one instance of known information into the same category with another instance of known information.

This category of propositional knowledge is best understood as a loose association whose members are somehow related to each other but may not all be governed by exactly the same set of rules. People’s instances of knowledge are, instead, related by how that knowledge (or its absence) impacts our world. Epistemologists should spend more time on examples in which a person’s knowledge affects their course of action.

Knowledge seems, to me, to be the response to an inquiry about whether we know something in a certain way for certain purposes. Beyond this, there is no meaning to asking whether any specific justified true belief is an instance of knowledge.

What makes the Gettier examples an intractable problem is that the chosen example propositions are empty of meaning. They cannot be known, not primarily because of their logic, but because their content and context is mostly meaningless. We can try to find a better set of rules by which to govern our cache of epistemic information, but that won’t solve the problem. We still won’t know anything about who has ten coins or who is in Barcelona. The content of the known turns out to be more important than whether it can be neatly expressed in propositions that stand in logical relationship to each other.

We can begin to see why Socrates identified knowledge so closely with virtue. If a belief encourages a political, academic, or other virtue-related reaction, or if the belief is further expressed in some non-propositional way, we have a reason to ask whether that belief is knowledge. But if the belief never has any impact beyond its own self-contained proposition, whether it amounts to knowledge is unimportant — literally, inconsequential. It has been so divorced from the knowledge that matters to us — an understanding of the world that allows us to take action — that we do not have the resources even to ask whether it is knowledge. It is uninvestigatible.

We should not assume, independently of a proposition’s content, that we can ever answer whether a person knows that proposition. The proposition may not be the best representation of the sort of knowledge in question. it may not represent anything “knowable” at all.

Consider how a teacher decides what type of exam to give the students. A multiple-choice quiz suggests that true belief, at a rate higher than would be expected from random guessing, is sufficient to prove knowledge. An oral defense suggests that a justified belief that engages opposing justified beliefs is the better proof of knowledge. The teacher decides what type of exam is appropriate for these specific students in this specific course; in doing so, the teacher isn’t taking a stand on “the definition of Knowledge itself” everywhere and for all time.

Gettier’s article intends to show that a “justified true belief” is insufficient to constitute knowledge. But, perhaps accidentally, it reveals something more exciting and important: the existence of justified true beliefs is not a sufficient motivation for the discipline of epistemology. Just any old proposition (coins in the pocket, who is in Barcelona) — even if justified, true, and believed — does not make one virtuous or allow one to live the good life. And it is exactly those sorts of skills and virtues that make knowledge so attractive. The meta-epistemological questions need to be present to keep the debate in focus. Just because p is a proposition does not mean it is knowable or worth knowing. Just because I know something does not mean I can fully, clearly, or appropriately express it in a propositional format without other sorts of knowledge. That we cannot be said to know some of our own justified true beliefs suggests that our real knowledge is bigger and more thrilling.

This article was based on my final paper for my Epistemology class at Brown University. I wrote it in May 2001 and happened to stumble across a copy just now, exactly 20 years later. I rewrote it substantially to make it more Internet-friendly. Edmund Gettier died two months ago at age 93.

Writing on dignity, democracy, and the pursuit of truth. Author: ‘Ten Past Noon: Focus and Fate at Forty’

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store