The “traditional” account of knowledge as justified true belief goes back at least to Plato’s time, over 2400 years. But recently, it has faced a significant challenge. So what happened? Well, a philosopher by the name of Edmund Gettier happened. He brought a number of counterexamples to the justified true belief theory into the center of attention of modern epistemology (that is, the study of knowledge). So-called Gettier cases are examples where a person believes something, which happens to be both true and justified, but where the belief nonetheless seems not to count as knowledge.
A number of examples of Gettier cases can be found in various articles on the problem, but here is a classic one. You glance at a clock (an analog one), and see that it reads 4:17. From this, you form the belief that it is 4:17pm, and in fact you are correct. But the clock actually broke 12 hours ago, and is just sitting there stuck at 4:17. So while you believe that it is 4:17pm, and your belief is true, and justified by your read of the clock, it seems wrong to say that you know it is 4:17pm.
I find the general problem in all these cases is that there is some kind of disconnect between the truth of the belief and the justification for it. Though in each case the subject holds a true belief, and is justified in doing so, the problem is that they would have still held the same belief, for the same justification, even if the belief were false. In many people’s minds, this disqualifies the belief from being knowledge.
The key insight revealed by Gettier cases, to me, is that the justification has to be valid for a true belief to count as knowledge, where validity has something to do with there being an appropriate connection between the justification and the truth of the belief. (I will flesh out this concept in a bit.) If I am right, the Gettier problem can be addressed by defining knowledge as validly justified true belief. This preserves the core intuition many people share that knowledge requires justification, while refining it to account for problematic cases.
And it preserves the argument I gave in this post that we can know things, and know that we know them. If we have justified belief, then not only do we think the belief is true, but we think our justification is valid. So again, justified belief is sufficient for a knowledge claim, and the claim will be correct if in fact the belief is true and the justification is valid.
My Response to the Gettier Problem
If you’re interested to learn more about the problem of defining knowledge, the YouTube channel Wireless Philosophy has an excellent introductory video series on epistemology, which covers many of the different theories of knowledge that have arisen in response to Gettier (starting on video 5 in that series). My own response is a variant of what is called the truth-tracking theory. I will get a little bit technical here, so feel free to skip this and come back to my next post, where I will start to explore more practically the different ways that we justify beliefs.
Here begins an aside on my own response to the Gettier problem:
The problematic situation common to almost all Gettier cases seems to be that the subject would still have had the belief, for the same reason, even if it were false. And I think it is also possible to generate cases where the problem is slightly different: the subject would not still have had the belief, even if it were still true (but the situation was modified in some way that should not have affected the belief). Given that, I believe it is possible to express conditions for the validity of some form of justification along these lines.
Let S be a subject (that is, a person). Let B be a potential belief (a proposition). And let J be a means of justification, by which S may or may not be able to justify belief in B. Then my first attempt at a suitable definition of knowledge can be expressed as follows:
S knows B on the basis of J if and only if:
- S believes B on the basis of J,
- J would justify B for S if B were true, and
- J would not justify B for S if B were false.
The first point says that S considers J, it seems to S on the basis of J that B is true, and S believes B. (The justified belief condition.) Here I am using “consider” in a very loose sense, since J could be anything from a rational argument to a subjective experience that S has.
The second point says that if it were the case that B were true and S were to consider J, it would seem to S on the basis of J that B were true. (The no false negatives condition.)
The third point says that if it were the case that B were false and S were to consider J, it would not seem to S on the basis of J that B were true. (The no false positives condition.)
The statements in the second and third points here are subjunctive conditionals, which take the form “if it were true that P then it would be true that Q,” where P and Q are propositions. It is an area of active debate among philosophers, logicians, and linguists as to how such conditionals are to be interpreted and evaluated. Determining what would happen were things different is not always an easy matter.
As an example, one way of handling conditionals of the form “if P were true then Q would be true” is by this process. Starting with what is actually the case, we mentally construct a set of counterfactual cases covering circumstances relevantly similar to the actual case; specifically, relevant to the way that the truth of P is connected to the truth of Q. Then we look at all of the counterfactual cases in this set where P is true, and if Q is true in all of them, then “if P were true then Q would be true” is true.
So dealing with subjunctive conditionals is a general difficulty with this definition. A more specific difficulty is how to trace the means of justification, J, between the actual case and the counterfactual cases being considered. This is related to the first difficulty, since specifying J will impact which counterfactual cases are relevantly similar to the actual case.
A final difficulty with this definition is that it might be too strong. Its strength makes it particularly elegant; together the three requirements actually entail that B is true (if I am reasoning about these subjunctive conditionals correctly, at least). So the requirement that B is true becomes redundant. But under this definition, no belief counts as knowledge unless it is almost infallible: if S knows B on the basis of J, then S would not be misled by J about whether or not B were true under any relevantly similar circumstance. This may put too great of a restriction on what we can call knowledge to accord with common epistemological intuitions. (It seriously calls into question what we can really know through inductive reasoning, for instance.) But that all depends on how the first two difficulties are addressed.
My second attempt at a suitable definition of knowledge weakens the first:
S knows B if and only if:
- S believes B on the basis of J,
- J reliably would justify B for S if B were true,
- J reliably would not justify B for S if B were false, and
- B is true.
Here the means of justification does not have to operate infallibly in order to be valid, but only reliably. (J’s false negative and false positive rates are only required to be low, rather than zero.) This deals with the problem of the first definition being too strong, but introduces the problem of how to characterize reliability, especially in the context of these subjunctive conditionals. And it could end up being too weak of a definition, since, depending on how reliability is characterized, it may fail to appropriately exclude the Gettier cases. Finally, it has to reintroduce the truth of B as a separate condition, so it is not as elegant. But maybe this definition accords better with our intuitions about what counts as knowledge, compared to my first definition.
There are certainly difficulties in making either of these precise. But, though I am no professional epistemologist, I think a good case can be made that one or the other of the definitions that I have offered is close to being the correct one.
To sum up, I believe the correct definition of knowledge is that it is validly justified true belief, where justification for a belief is valid if it tracks the truth (or, reliably tracks the truth) according to these subjunctive conditionals.
Here ends the aside.
In my next post, I will start to take a look at the different forms of justification that we use in practice. These forms of justification are generally considered to be valid ways of arriving at knowledge. This will give me a list of sturdy building materials, so to speak, out of which I can construct my belief system.