I've been looking for an explanation of the meaning of knowledge and I've come across this video on Youtube : PHILOSOPHY Epistemology: Analyzing A Knowledge #1 (The Gettier Problem) [HD]

What I can't understand is how can they (Gettier and Russell) say if someone does/doesn't know something even though they don't have a rigorous definition of knowledge? What does it even mean to know something?

  • First off welcome to philosophy.SE. Who is "they" in your question?
    – virmaior
    Commented Oct 22, 2018 at 14:51
  • 1
    Its about proposing a technical definition and seeing if it lines up with our intuitions
    – Not_Here
    Commented Oct 22, 2018 at 16:39

6 Answers 6


The basic concept of "knowledge" is admittedly vague in contemporary philosophy; but it goes back to a much more crisp classical distinction between "opinion" (doxa) and "knowledge" (episteme). The basic idea was that there are lots of things you might think are true, like "Socrates is a coward" that you've heard from other people, or guessed based on trivial evidence, or that come down matters of interpretation, where (a) different people, and especially different groups of people, consistently disagree, proving (b) most people's opinions can't be right (because they are incompatible) and perhaps no one's are, as demonstrated by the fact that (c) insofar as we ever get a reality-check, these opinions frequently turn out to be false.

The contrast was supposed to be with knowledge-types like geometry, where you start from indisputable premises and move on to deduce equally-indisputable conclusions from them.

That's the starting point. But since Plato people have been asking the exact question you have: is it possible to say "I know X" or even to say "He doesn't know X, that's just his opinion" if you can't explain what knowledge is - and not just in the negative sense (i.e., explaining what mere opinion is) or an ostensive sense ("It's like geometry") but by defining it?

One (minimalist) answer to this is that knowledge is a belief that you have reasons to have (i.e. it's not a totally arbitrary opinion) and is also true (i.e. it's not an error). This answer seems plausible because it fits with the way we use the concept (or you could say: with our intuitions about it, with what we want to be able to do with it) but it's not a perfect fit. When someone like Gettier attacks justified-true-belief, he doesn't need to provide an alternative because he is criticizing it on the same terrain that its advocates defend it. You might as well say to them: "You say that knowledge is justified true belief, but how can we evaluate this claim before you first define knowledge, so we can see whether your proposal flows from the definition?" Well, that proposal is their definition, and they defend it not based on its fit not with some hypo-definition but its fit with our feelings about the meaning of the word; and that is exactly where Gettier says it fails.

(By the way, it's possible you might be a bit of a nominalist - that is, you don't care what knowledge "really means" so long as we define it and agree on how we're going to use it before we start arguing. That's perfectly fine, but you'll have to learn to translate these realist-framings of philosophical debates into something else: e.g. instead of "Does knowledge = justified true belief?", "Does the set of justified true beliefs have any special epistemological properties that differentiate them from unjustified false beliefs or justified true beliefs?")


The problem is to move from a negative concept of knowledge as non-accidental true belief to a positive concept that avoids accidentally acquiring true beliefs. Gettier problems show how various positive definitions fail such a test. We know the test is true to the issue because we can still say, "Knowledge is non-accidental true belief," i.e. this is essential to what knowledge is. And in particular cases we can know that our positive particular grounds for (truly) believing something, gave us true beliefs non-accidentally.

For example, if a mathematical theorem is true and we correctly deduce it from the relevant premises and our belief "tracks" this truth (in the deduction), then in that particular case, we know the theorem is true, i.e. we didn't accidentally believe the truth. So the specifics of the deduction and our sensitivity to truth-tracking in that context mean that we know the theorem's truth. (By "truth-tracking" is meant something like "wouldn't have believed the theorem anyway even without the deduction, or wouldn't have believed the theorem if it was false.")

Note that knowledge might be sui generis as a concept. After all, wouldn't we have to know that a definition of knowledge was true? So the positive condition of knowledge might be irreducible; discursively, that leaves us with the negative criterion even when evaluating particular cases.

  • I'm not understanding why accidental true beliefs are so awful. Isn't much knowledge gained in some part through trial and error? Should we have rejected the incandescent light bulb?
    – Scott Rowe
    Commented Oct 27, 2022 at 0:59
  • 1
    @ScottRowe Gettier's point is that if we state a belief and that belief, by mere chance, is true but our justification for this belief is totally off, we'd normally not call this "knowledge (about the world)". It would simply be a belief that happens to be true. It is indifferent to mere guessing. Like Mel Brook's blind outlook who "guesses" that there are no enemies closing in. Luckily for Robin and his friends that happened to be true. Calling that knowledge proper, or so Gettier contents, would not be befitting.
    – Philip Klöcking
    Commented Oct 27, 2022 at 7:13

I am not a philosophy major, so I can't quote anybody for you, but I am a full time research scientist in AI, so we think about thinking fairly often, and have to come up with "working definitions" (meaning we can implement them) of such things.

I would get rid of this problem by just accepting that there is no such thing as absolute knowledge. Our brains construct models of reality. There may be absolutely true things in reality, but using a model (or simulation) of reality, we cannot know for certain that we know anything. Our model may not be complete, or accurate. Even in something like physics, we have a dozen phenomena we still cannot explain (including the nature of gravity).

So in terms of a human "knowing" something about the real world or universe, my working definition would be what they "know" is true is something consistent with all the evidence they are aware of, and they cannot imagine any other explanation that fits and explains all the evidence they are aware of. This is provisional: It is something the human knows until they learn something that contradicts it.

However, you can "know" something is true about a manufactured system: 1+1=2, I know that and nothing else I learn will change it, because the system of mathematics is not part of nature, it is something we created. Now, agreed, we created it to model nature, and we have continued to devise mathematics to more and more closely approximate or predict nature, but it is all an invented system of rules, a system of rules we came up with. Likewise you can know that a Full House beats a Pair of Aces in poker, if it does not, the game you are playing is not poker!

You can know the rules of grammar or law or any other artificial system we humans have created. But outside these finite, artificial systems of rules, I would say it is not possible to know anything for certain, what we know is provisional and subject to change by new evidence or insight.

I'd guess I'd have to add that we can know we don't have knowledge of something. Like our models of physics above (general relativity and quantum mechanics), which are both artificial models we have constructed of reality. We know they do not capture all of reality, because they are incompatible, and do not explain a dozen anomalies. So we can be certain our artificial systems, as they stand now, are incomplete models of reality.

  • 3
    Even though you are not a philosophy major (neither am I), would you have references to people in the AI field that share your views? Quoting them would give the reader a place to go for background material on your position. Welcome to this SE! Commented Oct 22, 2018 at 21:31

The two questions you have asked interested Plato to the extent that he invented the inquiry of epistemology in order to answer them.

The first question: "How can one know any characteristics of a thing if one does not know what that thing is?" appears right at the start of Meno when Meno nearly accosts Socrates, demanding to know how virtue is acquired by men.

Socrates chides Meno for presuming to know what virtue itself is, claiming that he, himself, has no clear idea and so cannot answer how it is acquired. This astonishes Meno, who chides Socrates for ignorance on a matter of basic gentlemanly knowledge.

But Socrates drives his point home by asking if anyone could know whether Meno were rich or good looking without knowing who Meno, in fact, is. Meno admits that a person lacking acquaintance with him could not know such characteristics as whether he were rich or handsome.

Socrates seems to prove to Meno that knowledge of a thing is necessary for knowledge of the characteristics of that thing. So, Socrates would echo your question, "How can we make claims about knowing things when we do not know precisely what knowledge even is?"

To ameliorate that difficulty, Socrates suggests that he and Meno search together for knowledge of virtue in order to determine how it is acquired by men.

A definition of virtue is more elusive than expected and results in a separate inquiry into the distinction between true opinion and knowledge. Socrates claims that when true opinion is based on causal reasoning, it is knowledge. This seems to be the origin of the justified-true-belief definition of knowledge.

In his Theaetetus, however, Plato has Socrates explore the JTB definition in great detail, eventually rejecting it with what I consider the presentation of a Gettier case.

So, Plato seems to have had similar questions to the ones you pose here, and seems to have struggled to more than one answer to them in the case of "What is knowledge?"

Your first question: "How can we make claims about knowledge when we do not know what knowledge is?" seems to highlight our ability to intuit the nature of a thing without a fully articulated definition of it. We intuit that Gettier cases are not knowledge from some internal conviction we have of what real knowledge is. Neither Plato nor Gettier completely accounts for this.

Your second question: "What does it mean to know something?" seems to be an ongoing one. Some additional criterion seems to be required to augment JTB, or some refinement of what is meant by justification.

Karl Popper argues that knowledge is not a belief at all, but an object in the world. We do not seem, however, to have reached the end of epistemology, so apparently, Sir Karl has not definitively solved the problem, as he claims.


From your assumption that Gettier does not have a rigorous definition I conclude that you did not read the paper and the video does not mention it. Gettier, as a matter of fact, starts his paper with the following:

Various attempts have been made in recent years to state necessary and sufficient conditions for someone's knowing a given proposition. The attempts have often been such that they can be stated in a form similar to the fol1owing:
(a) S knows that P IFF (i.e., if and only if)
(i) P is true,
(ii) S believes that P, and
(iii) S is justified in believing that P.
For example, Chisholm has held that the following gives the necessary and sufficient conditions for knowledge:
(b) S knows that P IFF (i.e., if and only if)
(i) S accepts P,
(ii) S has adequate evidence for P, and
(iii) P is true.
Ayer has stated the necessary and sufficient conditions for knowledge as follows:
(c) S knows that P IFF
(i) P is true,
(ii) S is sure that P is true, and
(iii) S has the right to be sure that P is true.

These are three formal definitions of what it means to know something. And what Gettier does is nothing more (or less) than giving us situations (the famous Gettier problems, which are two distinct cases) in which the criteria given in these definitions are fulfilled and yet the situation would not be judged as a case in which the person actually does know what they think they know.

We are not speaking about knowing that a given mathematical theorem is true here, at all. The problems are about knowledge of external states of affairs, ie. empirical knowledge. Thus, the problems are about epistemic uncertainty in pragmatic contexts and ultimately show that no matter how rigorous your definition, you will end up with two alternatives:

  1. You need to be so restrictive that the definition of knowledge "proper" becomes meaningless as it would mean we could not speak of knowledge in most cases. Basically, you would have knowledge in this sense only about abstract objects, like mathematical theorems or truisms. Empirical knowledge would have to be excluded and the normal use of the word meaningless.
  2. The definition would be more lenient and include cases like the Gettier problems, ie. we need to accept the fact that any use of the term "knowledge" in empirical context, as being about specific states of affairs we build beliefs about, is fallible.

In any case, I suggest you read the original paper. It is very short. And considering he wrote these two pages just because he was forced to publish something as part of his academic position, it is quite a feat to accomplish this amount of discussion being solicited.


Knowledge defined to overcome the Gettier problem
When knowledge is defined as a justified true belief there is a gap in the degree of connection between the justification of the belief and its truth. When we define knowledge as a justified true belief such that the justification proves that the belief is true this gap is eliminated.

"Gettier claims to have shown that the JTB account is inadequate because it does not account for all of the necessary and sufficient conditions for knowledge." https://en.wikipedia.org/wiki/Gettier_problem

A simple example of a sentence that cannot possibly be false:
This sentence is comprised of words.

The assertion that semantic meanings are expressed using words proves itself to be true on the basis that it is an example of semantic meanings expressed using words.

It is also irrefutable because every rebuttal would be an example of semantic meanings expressed using words.

If we simply ignore the limitations of empirical knowledge such as:
the Problem of induction
the Brain in a vat and
the Five-minute hypothesis

Then we can address how empirical "knowledge" can be defined to overcome the Gettier cases. A belief about our current model of the world that is fully justified by direct physical sensations or memories of direct physical sensations.

"I am seeing my shoes on the floor of my living room right now."
"I remember seeing my shoes on the floor of my living room."

  • If you translated the sentence into Chinese it would not be true. Ideograms do not stand for words, and any of 400 dialect speakers could still read the printed symbols. Two people reading the same Chinese text might utter different words.
    – Scott Rowe
    Commented Oct 26, 2022 at 23:10
  • @ScottRowe My example still stands because it is not written in Chinese.
    – polcott
    Commented Oct 26, 2022 at 23:34
  • Right, but: "The assertion that semantic meanings are expressed using words" is only true for non-Chinese statements, so it is not unconditionally true. A horse's assertion that you are about to get bitten is not even a statement, but true nonetheless, as someone I know found out.
    – Scott Rowe
    Commented Oct 27, 2022 at 0:54
  • 1
    @ScottRowe The key point is that I did solve the Gettier problem by reformulating the notion of knowledge.
    – polcott
    Commented Oct 27, 2022 at 1:21
  • 2
    Shifting the goalposts does not solve anything, it only shifts goalposts and produces triviality. Also, the Gettier problems are pragmatic problems, not formal ones. The pragmatic dimension is that you can still err on that which you think to be a proven fact.
    – Philip Klöcking
    Commented Oct 27, 2022 at 6:15

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .