In my everyday life, I try to have intellectual humility. I try, in other words, to give views different from my own a fair hearing, even if I disagree with them. In this essay, I’m going to explain why I take this attitude.

My reason for trying to maintain some level of intellectual humility is grounded, ultimately, in a conflict between two philosophical views that I’m committed to. The first view is realism, which says, for any domain it’s applied to, that the facts of that domain — whatever it is — are objective and independent. The second view is certainty, which refers to the state of having infallible knowledge about something. The problem is that these two views are incompatible. When one accepts realism about any domain, one necessarily acknowledges that her understanding of that domain is fallible. This fallibility, however, contradicts the notion of certainty. Therefore, one must give up either certainty or realism, on pain of inconsistency, for whatever domain is in question. When it comes to domains that I am a realist about, such as the external world, for example, my response to this dilemma is to give up on certainty and to embrace humility.

In the next few paragraphs, I’m going to clarify the concepts of realism and certainty more fully, formulate and advance an argument in favor of my position, and discuss some of its implications. The reason I think this argument is important is that, in almost all matters of human life, we are concerned with realist domains. Questions of love, politics, and one’s bank account balance, for example, are all questions about realist domains. Therefore, if the adoption of realism, for any domain, necessitates a more modest epistemic stance toward that domain, then adopting an attitude of intellectual humility towards almost all questions of importance in human life becomes imperative.

Realism

In order to setup my argument, I need to foreground a few concepts. The first concept I’d like to discuss is what philosophers call realism, which consists in the adoption of the following two claims:

Existence: For any realist domain D, the objects and relations of which it is composed exist.

Independence: For any realist domain D, it is independent of minds, languages, theories, concepts, etc.

Basically, when I say of some domain that it’s real, what I mean is that it is the way it is regardless of what I think about it or what language I use to express facts about it. For example, two plus two is four even if I mistakenly think it’s three. My thinking that the answer is three doesn’t make it the case that it is three; it just makes me wrong.

The language I choose to use can make a difference in the way I express facts, but it doesn’t change the underlying facts themselves. For example, in binary, one plus one is ten, and not two. But whether I express myself in binary or decimal does not affect the fact in question, namely that one plus one is two in decimal and ten in binary. Such facts are, in this way, mind and language independent.

Similar remarks apply to the claim that George Washington was the first president of the United States of America or that the speed of light is one hundred eighty-six thousand miles per second. These facts are as they are regardless of my mental attitude toward them, or whatever language, theory, or concept I use to express them.

Not all domains are real in this sense. My taste in beer, for example, isn’t independent of my own mind. In fact, it just is what I think about it or what I experience it to be. This is, in part, because I have direct epistemic access to my own mind, my own experiences and thoughts, or at least to some of them.

Realism implies that, for any domain that I’m a realist about, I could be wrong about it. The primary reason for this is independence. Unlike the facts about my taste in beer, the facts of any realist domain are independent of what I think about them. I don’t have direct access to these facts and so my beliefs about them may fail to reflect the truth. The reason for this is that, unlike the cases of my own experiences and thoughts, there are possibilities that I cannot rule out. For example, I could be misremembering; I may have made a simple mistake; there could be something affecting my perception, like a thorn in my eye obscuring my view; or, more remotely, I could be dreaming right now; I could be a brain in a vat.

The more remote of these possibilities do not personally trouble me. But the nearer possibilities do. I suspect that most everyone has had an experience of misremembering something, of making a mistake when doing a computation, of seeing one thing at a distance and discovering later, when up close, that it is another.

To sum up, if I accept both existence and independence about some domain D, then it follows that

(1) Any of my beliefs about D could be wrong.

This is because, as I have already stressed, I do not have direct access to D. The best I can do from an epistemological perspective is have some justification of my D-beliefs based on my experience, reflections, reasoning, whatever. I am barred in principle, however, from being infallible about domain D, since, as I have already said, I do not have direct access to D.

Certainty

The next concept I’d like to discuss is certainty. Rather than trying to define the word “certainty” or to give an account of the concept itself, I’d like to emphasize just one entailment of the common understanding of certainty that I think any plausible account of the concept would yield.

(2) If I am certain that p, then that p could not be false,

where the letter “p” stands for any complete sentence. (2) asserts that certainty is infallible. Not all accounts of certainty yield this entailment, but I intend to focus on only those accounts that do.

Two points deserve emphasis and clarification. First is that I am interested in certainty from a first-person perspective, that is, from the point of view of any given agent or individual. I’m concerned with what is entailed when someone says, “I am certain.”

Second, (2) expresses what philosophers would call a modal aspect of the concept of certainty. This modal aspect can be interpreted in different ways. Here I will be interpreting this aspect epistemically. It will be instructive, however, to contrast this interpretation with a metaphysical one. Consider the following two sentences.

While the Jets lost the game last weekend, they could have won.
Mark must be in the room, since I hear him in there.

The first sentence expresses a metaphysical possibility. It is not making a claim about how the world actually is, for it is known that the Jets lost; it is making a claim about how the world could or could not have been. It is saying that the Jets could have won. In contrast, the second sentence is making a claim about how the world really is. It is saying that, since I hear Mark in the room, he must really be in the room. In other words, there is no other epistemic possibility relative to current information other than that Mark is in the room.

(2) above is to be interpreted in this second, epistemic, way. (2) is making a claim about how the world really is.

The Argument

Here’s the argument. The acceptance of (1) and (2) about the same domain D yields the following challenge.

(1) Any of my beliefs about D could be false.
(2) If I am certain that p, then that p could not be false.

Let us assume that

(3) p expresses a belief about D.

So, by (1),

(4) p could be false.

Therefore, by modus tollens against (2),

(5) I am not certain that p.

Since p can stand for any D-belief or D-proposition, I cannot be said to be certain about anything about D. Realism and certainty are, I think this argument shows, incompatible.

To clarify the upshot of my argument a bit further, it may be helpful to run through a concrete case. Let’s say, for example, that I’m certain that I’m in Seattle right now. By (2), it follows that it could not be false that I am in Seattle. But, since my belief that I’m in Seattle is about the domain of the external world, which is a realist domain, on my view, it also follows that I might not be in Seattle. I might in fact be in New York and am only dreaming or suffering from some momentary confusion that I’m in Seattle. So, there’s a coherent, if low probability, possibility that I’m in New York and not Seattle. So, by my own lights, it could not be the case that I’m certain that I’m in Seattle, since, clearly, it could be the case that I’m not in Seattle.

In ordinary life, few would be troubled by such a low probability possibility. For most intents and purposes, one need not concern oneself with such oblique possibilities in expressing certainty about some or other claim. However, the fact that the probability that I’m wrong is low in this example is not relevant. What’s relevant is that, in principle, I cannot rule out these recalcitrant possibilities to attain infallible certainty. Being a realist about the external world — or, for that matter, any other domain — necessarily imposes limits on what I can know about it.

To put the point in more philosophical terms, it will be helpful to distinguish the following two classes of question, based on the distinction drawn above.

Metaphysical questions: For any domain, how are things in that domain? What are the actual facts?

Epistemological questions: For any domain, how does one come to know about how things are in that domain? What are the methods for determining the facts? Can one come to know all the facts about this domain?

My argument shows that when one adopts a realist position in response to the metaphysical question about a given domain D, then one cannot admit that certainty is attainable in her answer to the epistemological question.

Degrees of Belief

In Tolstoy’s novel War and Peace, he writes,

All we can know is that we know nothing. And that’s the height of human wisdom.

The position I’ve just argued for sounds dangerously similar to this quip. It sounds as if I’m saying that I’m certain that I can’t be certain, which would be a contradiction. My position, correctly stated, however, is not that I’m certain that I can’t be certain of anything. It’s that I believe I can’t be certain of anything about any realist domain.

The meaning of the term “believe,” in this context, deserves a bit of clarification.

In a word, I judge various claims to be true or false and I hold some claims with more confidence than others. More specifically, for any claim that I’m competent enough to evaluate, I have some degree of confidence in it. I’m very confident in some statements, like one plus one is two, for example — such a claim is as close an approximation of certainty as one could hope for. But there are other statements, at the other end of the spectrum, that I believe to be true but which I’m not so confident about, such as the claim that there isn’t an afterlife. I come down, on that particular question, saying yes, but I’m not very confident that this belief is true. I don’t really know.

When I say that I believe I can’t be certain of anything, I am expressing a certain confidence in the proposition. I am expressing a positive epistemic attitude toward it, but I am not expressing full certainty in it.

Objections

There are two primary ways of pushing back against the argument I gave above. One is to revise what is meant by certainty in such a way that it does not entail infallibility, rejecting the second premise of my argument. In a word, the objection is to say that (2) — the claim that p can’t be false — isn’t really part of the common understanding of certainty. When one says that she is certain that p, she isn’t saying that p must be true (in the epistemic sense); she is merely saying that p is true. If this is the case, then certainty would seem to be achievable, even with respect to realist domains.

Let me begin my answer to this objection by restating it more exactly in two different ways. I’ll start by couching this objection terms of knowledge, then more generally. I’ll explain the difficulties with each version in order.

To put the objection in terms of knowledge, I’ll need to explain the distinction between first and second-order knowledge. Consider the following statements.

(6) I know that p.
(7) I know that I know that p.

The first statement (6) expresses first-order knowledge, whereas (7) expresses second-order knowledge. If I believe, for example, that the cat is on the mat and, in fact, the cat is on the mat, then it may be claimed that I know that the cat is on the mat. In this example, I have first-order knowledge that the cat is on the mat. However, I may not know that I know that the cat is on the mat. So, I may have first-order knowledge that p without having second-order knowledge of the same.

Does certainty operate the same way knowledge does? In my view, the meaning of the term certainty, in contrast to knowledge, implies both first and second-order knowledge. To say that I am certain that p is not to say only that I know that p, but also that I know that I know that p. If this is right, then one cannot be certain of p without knowing that one knows that p, which entails that p must be true in the epistemic sense. So, (2) obtains on this understanding of the term certainty and the argument above succeeds.

The objection, now expressed more exactly, is to claim that only first-order knowledge of p is required to be certain that p. One need not know that one knows that p in order to be certain that p; rather, one may merely know that p in order to be certain that p. The upshot of this position is that knowing p in this way does not entail that p must be true; it only entails that it is true, which enables rejecting premise (2) above, causing my argument to fail.

My reply is that such an account would allow for accidental certainty, which is not consistent with the term’s ordinary usage. It would follow, on such an account, that I’m certain of all kinds of things that I am plainly not certain of, since all that would be required for me to be certain that p would be the accident of my belief that p coupled with the truth that p. It would follow, for example, that I’m certain that there’s life on Mars, whether or not I have any justification or any attitude of certainty about this claim, if I believe that there is life on Mars and, in fact, there is. This does not reflect the ordinary way that the word "certainty" is used. When someone says she is certain that p, she is saying more than that p is true; she is saying that p is true, that she knows it, and that she knows that she knows it. The attitude of certainty itself, in other words, demands second-order knowledge, which, in turn, entails (2) that p could not be false in the epistemic sense.

No everyone will follow me in the belief that certainty requires second-order knowledge. Still others may reject my formulation of this objection in terms of knowledge in the first place. What this objection really amounts to, if construed more generally, is that one may specify any theory of certainty one likes while stipulating that it denies (2), the claim that p could not be false. Such a theory would be coherent and could be plausible; and it would prevent my argument from succeeding. For example, one theory that fits this schema says that certainty consists of justified and true belief held with maximal confidence, without accepting (2).

I have two replies to this more general objection. First, many accounts of certainty require, in addition to the truth that p, that one have a justification for the claim that one is certain of. But, to my mind, what such justifications or other epistemic requirements do is rule out epistemic possibilities in which that p is false. That is, they end up entailing (2), or something close enough to it that the argument given above would still obtain.

My second reply is that, if one’s preferred account of certainty really does not yield (2), so that one could be certain that p, while at the same time admitting that p could be false, then we are not talking about the same concept. In this case, I would be willing to admit that the argument I gave above does not succeed. I do not find this troubling, however, since my aim is to show that the adoption of realism about any domain entails that one is fallible about that domain, a conclusion that remains intact, even if I have to give up on framing it in terms of certainty.

The second main way of pushing back against my argument is to revise or reject realism, specifically, its claim of independence, in order to reject the first premise of my argument. Many philosophers do reject independence. Berkeley, for example, famously claims that all the world is in the mind. And other philosophers make the same assertion, which enables some of them to say that we have direct epistemic access to the world and so we can’t be wrong about it.

The problem with this objection is that it violates my starting assumption. It goes beyond what I promised in this paper, which is to show that realism and certainty are incompatible. Thus, while I acknowledge that there is a rich literature of anti-realist philosophy and many ways of understanding realism besides, this is not the venue to pursue such reflections. Here, I take independence as a foundational assumption. And I have shown, on the basis of these assumptions, that realism and certainty are indeed incompatible.

Humility

I would like to close with an observation about human psychology.

In recent years, I have been stunned to see so many people who seem to hold their opinions in such confidence that they believe there is no possibility whatever that they could be wrong. Such certainty in one’s beliefs frightens me. It frightens me because, in my view, confidence to this extreme is likely to lead to misery.

In considering any practical question about what to do, in order to know what to do — that is, in order to know what actions or which policies would produce the best outcome — one needs to know the truth about the matter in question, or, at least, to come close to the truth. But gross levels of confidence prevent one from getting close to the truth since they make one unwilling to revise one’s beliefs in the light of new evidence. They make one unwilling to admit new evidence into one’s deliberations because one thinks one already has all the answers. Indeed, in my experience, I have often found such confidence to be indicative of ignorance for this very reason.

The argument I have given above, if successful and if they would hear it, should induce sobriety among people with such confidence. I have argued that everyone is in principle fallible about any realist domain. What follows from this argument, if it is to be believed, is that certainty about any realist domain is impossible. What can be achieved is confidence in one’s beliefs.

I want to caution that these reflections do not entail that one should abandon the pursuit of knowledge on the grounds that it is futile. Despite the fact that certainty may not be realizable, human knowledge advances. What these reflections do entail is that one should bring a degree of humility to her epistemic considerations. There is a degree of humility that respects the epistemological situation one finds herself in but which does not prevent due confidence in one’s beliefs.

I am urging one to give up on the fiction of certainty, but not on independent truth. In one’s epistemic considerations, even if certainty in it may always elude one’s grasp, truth is the only thing worth striving for.

Thanks to Said Saillant for comments on an earlier draft.