Epistemology
Epistemology is the branch of philosophy that asks what knowledge is, how it differs from belief, and what justifies holding a claim to be true, which makes it the source of every rule for testing one. The word comes from the Greek episteme, meaning knowledge, and the plainest synonym in English is the theory of knowledge, which is what the subject is often called in universities. It is the parent concept for everything else on this site: every rule about evidence, every named fallacy, every bias and every test for a claim is an answer, or an attempt at one, to a question epistemology asks first.
Yes, it is a branch of philosophy, and here is where it sits
Epistemology is one of the four or five traditional branches of philosophy, alongside metaphysics, which asks what exists; ethics, which asks what is worth doing; logic, which asks what follows from what; and aesthetics, which asks about art and beauty. Asking about epistemology versus philosophy is therefore a category question rather than a comparison: philosophy is the field, epistemology is one division of it, and the divisions leak into each other constantly. A question about whether a thing exists usually turns into a question about how anyone would know, which is how metaphysics hands work to epistemology several times a day.
What distinguishes epistemology from the sciences is that it studies the standards rather than the subject matter. A physicist asks what the evidence shows. An epistemologist asks what makes something evidence, why observation supports a conclusion at all, and how much support is enough. Those questions cannot be settled by running another experiment, because any experiment presupposes an answer to them.
Belief, truth and justification: the three parts of the old definition
The traditional definition holds that knowledge is justified true belief, and it has 3 conditions that all have to be met at once. The account traces back to a discussion in Plato's Theaetetus, where knowledge is examined as true belief accompanied by an account of why it is true, and it dominated the subject for a very long time.
- Belief. You have to actually hold the thing to be so. You cannot know something you do not believe, and reciting a fact you privately doubt is not knowing it.
- Truth. It has to be so. A confident, well argued, universally shared belief that is false is not knowledge, and no amount of agreement converts it into knowledge.
- Justification. You have to have some good reason for holding it. A lucky guess that happens to be right is not knowledge, because there is no connection between why you believe it and why it is true.
The third condition is where all the difficulty lives, and it is the one that most arguments are really about. Two people who agree entirely on the facts can still disagree about whether the reasons on the table are good enough, and that disagreement is epistemological even when it is conducted as if it were about the facts.
The Gettier problem in ordinary words
In 1963 the American philosopher Edmund Gettier published a paper of about three pages in the journal Analysis, titled with the question of whether justified true belief is knowledge, and the answer he gave was no. His argument was a pair of counterexamples in which all three conditions are satisfied and the result still fails to look like knowledge, because the belief turns out to be true by an accident unconnected to the reason for holding it.
The easiest version to hold in your head is the stopped clock, a case Bertrand Russell had discussed before Gettier wrote. You walk past a station clock, see that it reads nine o'clock, and form the belief that it is nine o'clock. It is nine o'clock. You have a reason: reading a public clock is a perfectly good way to find out the time, and you had no cause to doubt this one. So the belief is justified, and it is true, and you hold it. But the clock stopped exactly twelve hours ago, and if you had walked past four minutes earlier you would have formed a false belief with the same justification. Something is missing. What is missing is the connection: the fact that made your belief true had nothing to do with the process that produced it.
Philosophers have proposed repairs for more than 60 years, and none has become the standard. 3 families of answer are worth knowing because each one turns into a practical habit:
- Add a no-false-steps condition. Require that the reasoning contain no false assumption along the way, which disqualifies the stopped clock because it depends on the clock working.
- Require reliability. Ask not whether the reason sounds good but whether the method that produced the belief generally produces true ones, which shifts attention from the argument to the process.
- Require sensitivity. Ask whether you would have believed differently had the fact been different, which is the question that catches the four-minutes-earlier problem directly.
The second and third of those are not academic. They are the reasoning behind every methodological rule on this site. Asking whether a method is reliable is what a control group is for. Asking whether you would have noticed had the claim been false is what falsifiability asks. The Gettier problem is where those two habits come from.
Where justification actually comes from
Justification arrives through 4 routes, and each has a characteristic way of failing. Knowing which route a belief came down is often more useful than debating the belief itself.
| Route | What it gives you | How it fails |
|---|---|---|
| Perception | Direct contact with the thing | Perception is constructive, and expectation shapes what is noticed and stored |
| Memory | Access to what you already established | Memory reconstructs, and it reconstructs toward the story now held |
| Testimony | Almost everything you know | The chain from the original observation is long and each link drops conditions |
| Inference | Conclusions beyond what was observed | Valid form with a false premise, or a conclusion wider than the premises support |
The third row deserves emphasis, because it is the largest and the least examined. You did not personally verify the distance to the sun, the existence of countries you have not visited, or the date of any historical event. Nearly the entire contents of an educated mind arrive by testimony, and the practical epistemological question is therefore not how to verify everything, which is impossible, but how to allocate trust between sources: who was in a position to know, what would have happened to them if they had been wrong, and how many retellings sit between the claim and the observation.
The fourth row is where inductive reasoning lives, and it carries the oldest unsolved problem in the subject. David Hume, in his Enquiry Concerning Human Understanding of 1748, observed that every argument from past regularity to future regularity assumes that the future will resemble the past, and that this assumption cannot itself be established by observation without circularity. Nobody has produced a clean answer. What working inquiry does instead is treat the assumption as a working commitment and build in error correction, which is why hypotheses in research are set up with an explicit alternative and a null hypothesis stated in advance: the design gives the data a genuine opportunity to leave the null standing, rather than asking whether the favored explanation can be made to fit.
Why this is the parent of every other page here
Every practical tool on this site is an epistemological position in working clothes, and the connection is worth making explicit because it explains why the tools have the limits they have.
- What counts as evidence is a ruling on which perceptions and records can serve as justification, and on the difference between empirical and anecdotal support.
- The burden of proof is a rule about who has to supply justification, derived from the asymmetry between confirming and refuting.
- Correlation and causation is the inference row of the table above, examined in detail with its four standard failure modes.
- Occam's razor is a rule for choosing between explanations when justification runs out, and it is a preference rather than a proof precisely because epistemology cannot license anything stronger.
- Confirmation bias is the systematic corruption of the justification process from inside, in which evidence is gathered and weighed unevenly without the person noticing.
This is also why a fallacy label is not a rebuttal. Identifying that an argument fails to justify its conclusion tells you the justification is absent, and it says nothing whatever about whether the conclusion is true. Truth and justification are separate conditions in the definition, and collapsing them is the most common mistake made by people who have recently learned to spot bad arguments.
Certainty, doubt and the useful middle
Total skepticism and total certainty are both unavailable, and epistemology spends most of its energy on the territory between them. The argument for doubt is easy to state and hard to escape: any belief can be questioned, any faculty can be deceived, and Rene Descartes made the point vividly in the 17th century by imagining a deceiver capable of producing all his experiences without any of them being real. The argument cannot be refuted from the inside. It also cannot be lived, and treating it as a reason to abandon inquiry converts a serious problem into a pose.
The position that most working scientists and philosophers occupy is fallibilism: beliefs are held with confidence proportioned to their support, all of them are revisable, and none is immune. Fallibilism is not the claim that everything is equally uncertain, which is the caricature and the lazy version. It is the claim that certainty is not the standard, and that a belief can be well justified, worth acting on, and still wrong.
5 questions turn all of this into a habit you can run on a claim you meet tomorrow, including one of your own.
- Separate the three conditions. Do I believe this, is it actually so, and what is my reason, listed as three answers rather than one?
- Trace the route. Did this reach me by perception, memory, testimony or inference, and what is the characteristic failure of that route?
- Test the connection. If this were false, would my reason for believing it look any different to me?
- Count the links. How many people stand between the original observation and the version I was given?
- Name the standard. What would be enough to change my mind, stated before the evidence arrives?
The third question is the stopped clock, asked about yourself, and it is the one worth keeping. A belief you would hold just as firmly whether or not it were true is not knowledge, however carefully it was argued and however true it happens to be.