The Folly of Knowledge

Why we should favor reasonable belief as the focus of our epistemology

The biggest problem with epistemology is knowledge. Issues like the nature of truth, the manner in which we mentally interact with the world, and the nature of doubt are all better addressed if we take our basic epistemological principles and remove knowledge from them, or at least remove it from the spotlight. Instead of discussing knowledge as our primary epistemological focus, we should instead discuss belief.

The issues with knowledge stem from the degree of absoluteness which the term carries. We either know something or we do not. Skepticism latches onto the tiniest weaknesses of the beliefs we’d like to claim as ‘knowledge’ and digs in its claws: how can we say we know anything about our physical world, for instance, if we cannot say that we know our senses to be accurate?

A contrast is often drawn between belief and knowledge. Knowledge is often defined as being a special, superior species of belief. Most commonly, the definition is “justified true belief” or some variant thereof: a belief for which we have some compelling reason for considering to be true and which reflects some deeper universal truth. Beliefs are often contrasted as weaker things, which we believe without the same sort of resounding reason. This, however, need not be the case. If knowledge is justified true belief, then surely we can have justified belief regardless of whether we can state authoritatively that our belief is true. Furthermore, we can consider that a particular belief might represent the best justified conclusion we can draw on a particular topic, again without need to add “truth” to our definition. If this is the case, we ought to hold the belief in question, , even if we are simultaneously acting in our best epistemological interest to hold on to some degree of doubt. We will refer to such a belief as a Reasonable Belief: the belief which, when we bring our powers of reason and logic to bear, is shown to be the best option. Reasonability is transitive, as well: if the premises upon which I base a belief are the best beliefs I can construct on their respective topics, then the belief upon which they rely is similarly reasonable. I would be behaving in a less logical manner if I were to accept different premises and a different resulting belief.

We lose very little from discarding knowledge in favor of reasonable belief. Whether George Moore¹ says “I know my hands exist” or “it is reasonable for me to believe that my hands exist”, he is still able to use that belief (keeping in mind that knowledge is a subset of belief) as a premise for further action. On the other hand, we stand to gain a great deal. The cracks upon which the arguments of skepticism find purchase are far less consequential if we simply seek the belief with the fewest cracks rather than perfect knowledge that we can state is absolutely true. Even a deeply flawed belief can be reasonable if we simply have little to base our beliefs off of, and a false belief can nonetheless be a reasonable belief that’s merely based on flawed premises.

We can be a bit more precise in our discussion of beliefs by introducing the concept of metabeliefs. A metabelief is a belief about a belief. For instance, if we have a justified belief, the justification itself is a belief about our original belief. If we have a justified belief in which we have some amount of doubt, we have the original belief, a metabelief describing our reasoning for holding that belief, and a metabelief describing how strongly we hold our original belief. Reflecting back on knowledge, knowledge is simply a belief with certain specific metabeliefs. Perhaps we believe it to be justified and true, or perhaps we prefer a more exotic epistemology with a different set of metabeliefs attached to the core belief we claim as “Knowledge”.

Just as we can ask what beliefs we ought to hold, we can ask what metabeliefs we ought to hold with regards to our beliefs. For a certain belief, the evidence we possess may have certain structural flaws, or there may be specific scenarios which would falsify that evidence. Determining what metabeliefs are most reasonable is a far more fruitful endeavor than determining whether a specific belief is worthy of the metabeliefs that collectively bestow upon a belief the title of “knowledge”.

Even for trivial cases, breaking down a statement of fact into its constituent metabeliefs can be a useful exercise. A philosophy professor of mine once asked me if I knew that George Washington crossed the Delaware. I responded that I didn’t think I could claim that as knowledge, but that I believed he did. If this belief is to be considered a ‘reasonable belief’, what is the structure behind that belief and what makes it reasonable? In this case, the event was observed by many different people and modern researchers who have studied that era are in consensus that the event happened. A great deal would have to go wrong, or a great number of people would need to collude in a very specific manner in order for my belief to be false². If the crossing had only been observed by one individual, or if Washington had crossed on his own in the dead of night and only told people afterwards, the resulting beliefs would be less reasonable to hold. Ultimately, whether or not Washington actually did cross is irrelevant for the construction of this belief: the evidence I have strongly supports a specific claim and it is most reasonable that  I believe that evidence. Additionally, I have no evidence suggesting that I have a particular reason for believing any of the premises upon which this belief rests to be inaccurate. While there is always a baseline of skeptical scenarios in which my belief could be incorrect, I can conclude on the basis of induction that any reasonable level of research I could do on the topic would be unlikely to result in different premises. It is reasonable not only to hold the belief that Washington crossed the Delaware, but to attach to that belief the metabeliefs that the core belief is a strong one which will not change in the future.

In contrast with general conceptions of knowledge, whether or not a belief is reasonable is not contingent on truth. Suppose, for example, that there was a compelling reason for the Revolutionary Army to spread the falsehood that Washington had crossed the Delaware, and that the current belief I possess is ultimately rooted in a revolutionary disinformation campaign and not on the movements of Washington himself. This ultimate reality does not change the evidence from which I can draw my conclusions: I have read in numerous places and heard from numerous independent people that Washington crossed the Delaware. I have not seen any contrary evidence or evidence that the crossing narrative was falsified, nor do I have any other reason for believing as such. It would be unreasonable to conclude on the basis of my evidence that the disinformation narrative was true. There are possible scenarios in which my belief might be wrong, but all of them would be unreasonable to believe.

And what would we gain from going from stating that the Crossing narrative is a reasonable belief to stating that it is Knowledge? I can use it as a premise for additional beliefs, add it to my world view, and take actions on its basis regardless of whether I possess reasonable belief or knowledge. If I’m taking a history test, it is irrelevant whether or not that belief can be counted as ‘knowledge’: it represents the position I ought to hold and should guide my answers on that test.

The corollary to this is that we must remain open to new evidence. Our beliefs represent only the best conclusions we can draw from the information we have available. The temptation to engage in confirmation bias and reject evidence not because it is flawed but because it contradicts our current beliefs is a powerful one and a poisonous one if we seek the worldview that cleaves closest to truth. We should not be afraid to question our beliefs.

Skepticism and Reasonable Belief

The elephant in the room for any construction of epistemology is strong skepticism. For virtually any knowledge-based epistemology, there is a criticism which in some way reduces to “there exists a skeptical hypothesis we cannot disprove for which our knowledge claim is inaccurate.” “There is a tree in front of my house”, for instance, can be countered by the skeptical hypothesis of “what if you are actually a brain in a vat and your house doesn’t exist?” Proving knowledge in the face of skepticism is incredibly difficult, and the question of skepticism has sat at the forefront of epistemology for thousands of years. Approaching epistemology from a position of Reasonable Belief instead of from knowledge allows us to confront skepticism head on. Rather than proving that the skeptical hypothesis is false or finding some way to bypass it completely, we will address it with the goal of stating not that we can disprove skepticism but that believing skepticism is not reasonable. To state this differently: we can’t disprove skepticism, but believing it is not reasonable.

(As an aside, the remainder of this section will be a fumbling attempt at something resembling a proof. If you don’t like excessive use of acronyms or untrained attempts at philosophical rigor, please skip to the next bold header.)

We can begin our attempt at demonstrating that skepticism is unreasonable by first considering the purpose of our beliefs. We construct beliefs as a basis for action, and to serve as propositions for additional beliefs. Even if our goal is simply to describe our external world, our beliefs have a purpose.

For any possible world, we can describe the degree to which it is concordant to our beliefs. In particular, we can consider the degree to which it is concordant to our evidential beliefs; the direct beliefs which represent our internal representation of the external world. Our senses and memories, for instance, are direct representations in our minds of a small slice of the external world.

For the most extreme Skeptical hypotheses (we will refer to these as strong skeptical hypotheses, or SSHs), all of our evidential beliefs are non-representative of the actual external world. For instance, if we are deceived by Descartes’ demon³, everything we experience and everything we remember is a curated illusion. As an additional constraint on our definition of a Strong Skeptical Hypothesis, let us assume that induction is also false in these hypotheses: our memories are neither a descriptor of the external world nor even a predictor of the sensations we will experience as a result of future action. If an SSH is true, then we do not have a rational basis for interacting with the world.

We do not have evidence for an SSH, and definitionally, all of our memories and senses would be inaccurate if an SSH were true. Under an unknown SSH, any belief set is equally useful as a basis for action because we have no evidence from which to construct rational beliefs. If I am a brain in a vat in which the scientists have manipulated my memories into existence and where my memories aren’t even a reasonable basis for action in my perceived world, I have no idea how to go about interacting with my surroundings, nor do I have any evidential basis for either constructing beliefs about my surroundings.

Let us contrast a Strong Skeptical Hypothesis with its opposite: a world in which every one of our memories and sensations is representative of reality. In this world, believing our evidence gives us the ability to interact rationally with reality. Any deviance we accept from our evidence, anything we believe which conflicts with the evidence we have from our senses and memories reduces our ability to predict the outcome of our actions or produce additional reasonable beliefs. We will describe this belief set as the Least Skeptical Hypothesis.

Limiting ourselves to discussion of the SSH and the LSH, it is clear that we are better served by believing in the LSH. If we believe in the LSH and the SSH is true, we still have the same amount of evidence and the same ability to interact with our world as if we believe the SSH. If the LSH is true, then we are able to epistemologically interact with our world only if we believe in the LSH, in which case all of our evidence is a rational basis for action and for the construction of belief. It is most rational to believe in the LSH because we gain nothing from changing our beliefs to the SSH, even if it is in fact reality. So long as the LSH is possible, there is a possibility that our beliefs are predictive and useful if and only if we believe in the LSH. Even if only a portion of the LSH is possible, if we believe in the LSH, our beliefs will have some degree of ability to describe the world and predict our actions, while belief in the SSH gives us nothing.

This is not contingent on how likely we perceive the existence of a SSH as being. Whether we think it’s unlikely, equally likely, or unknowably likely that true reality is an SSH, our conclusion is still the same because the SSH is fundamentally unactionable. There is nothing we can gain from believing in a SSH, no reasonable course of action we can draw from such a belief, and no way that believing in it can give us a reasonable course of action, regardless of how likely it is to exist

It is important to note that, even if I believe in the ‘correct’ SSH, I still do not have a rational basis from which to construct beliefs about the external world. Belief in an SSH entails believing in a reality which contradicts my memories and sensations, which are singular in nature. Even if I can believe in an SSH, I cannot will myself to have different experiences or different memories. If I believe in an SSH, I ought to believe that I have no evidence from which to draw conclusions about the external world. Even if I were to by chance believe in an SSH which is representative of reality, because I lack any perceptual or mnemonic evidence or other means to draw conclusions my actions with regards to reality would be just as baseless as if I were to believe in the LSH.

We must, of course, consider other skeptical hypotheses: partially skeptical hypotheses in which some, but not all, of our beliefs are incorrect. There are two possible ways in which we can construct such hypotheses: in a non-falsifiable or in a falsifiable manner.

A non-falsifiable partially skeptical hypothesis (NFPSH) is one in which some portion of our beliefs are incorrect, but in which we do not have the means to identify them as incorrect. For instance, suppose that Descartes’ demon falsified all of the external world beyond our front door, but that our evidence of our home (e.g. senses and memories) was representative of reality. Every panorama we see through the windows was falsified by the demon and the moment we stepped out the door, the demon replaced all of our senses with a fake world, while in reality our house sat adrift in a screaming void.

In such a NFPSH, we can make the same value-based judgment as we did for the SSH: we cannot interact with the truth that corresponds to our false belief, so we gain no value from believing the NFPSH. We should choose every aspect of the LSH over the NFPSH, assuming that the LSH is possible.

This leaves falsifiable partially skeptical hypotheses (FPSH). If I look at a clock which says “2:00 PM”, I could believe that it’s two o’clock or that the clock is broken and inaccurate. Unlike the previous hypotheses, I can easily test and see if the clock is broken by either waiting for it to change (ruling out the possibility that the clock is stopped) or by using some other means of determining the time.

This particular FPSH also represents something that happens with some regularity. I have encountered broken and inaccurate clocks before. Beliefs are not binary in nature: I don’t have to choose between believing that the clock is broken or that the clock is accurate. I can believe, for instance, that the clock is probably functional and is reasonably accurate. I can make this claim on the basis of my past experience with clocks: most clocks are kept in a working state and are generally accurate, provided that Daylight Savings Time hasn’t recently started or ended. Unless I have particular evidence that this clock is broken, it’s reasonable for me to believe that it offers a decent estimate of the current time.

There are scenarios in which I would like to reinforce my belief, though. For instance, if I’ve got a critical meeting to attend that I cannot be late for, checking to make sure the clock is accurate might be the best course of action. Particularly consequential decisions might need particularly robust premises, and increasing the strength of my belief is something I am certainly capable of.

Other falsifiable hypotheses might not have the same property of possible but estimable failure. The belief that my guitar case is full of bees that will swarm me if I open it is one such example. In this case, I can conclude on the basis of induction that there is virtually no possibility that my guitar case is full of bees. I’ve opened it numerous times and never had bees emerge, nor have I heard of any other guitar players suffering from such bee-related calamities. I can conclude on the basis of my evidence that my guitar opening experience will almost certainly be bee-free without needing to listen closely for buzzing to confirm my belief.

While these falsifiable skeptical hypotheses differ from my prior skeptical hypotheses in that I interact with them on the basis of evidence, past experience, and probability, the core mechanism I use to discard them as reasonable beliefs is still the same: it is unreasonable for me to disbelieve my evidence without compelling counter-evidence. I should believe my evidence and hold beliefs supported by that evidence, while at the same time seeking to form metabeliefs as necessary to describe my degree of confidence.

When Moore states “I believe my hands exist”, he is perhaps in opposition to more than one of these. Certainly, if the world as a whole is an illusion, his hands don’t exist, so he must contend with the non-falsifiable fully skeptical hypothesis. Descartes’ Demon could be falsifying only Moore’s hands, in a partially skeptical hypothesis, and a myriad of other skeptical hypotheses of various flavors could similarly be constructed. For all of these, we can dismiss them as unreasonable on the basis of the above arguments against general skeptical hypotheses. We have no evidence that any of these could be true, even if we can’t absolutely disprove them, and thus no reason to believe them. Moore ought to believe in his hands because there is no non-skeptical hypothesis he can concoct in which it is likely that his hands do not exist.

Rejecting Skepticism: Belief vs Knowledge

We address Russel’s complaint about the psychological impossibility of skepticism by offering a sound basis for believing that skepticism is false, even in absence of ability to disprove skeptical hypotheses. In a sense, we accept the possibility of a skepticism of knowledge, but not one of belief. It may be possible that I know nothing and cannot possess true knowledge of anything, but that isn’t especially relevant to the construction of my beliefs: there are nonetheless things I ought to believe and which I ought to use in the same manner in which I’d generally strive to use knowledge. For all practical purposes, a reasonable belief is a reasonable substitute for knowledge.

On the flip side of Russel’s epistemological observation, skepticism of belief is logically unsound. I ought not believe in skeptical hypothesis, or that my memetic and sensory evidence is fundamentally unfit to serve as a basis for further belief.

These patterns of belief also reflect our day-to-day epistemological processes. I cannot say, for instance, that I know I will not be hit by a meteor while walking my dog. However, it would be unreasonable for me to believe that I will be hit by a meteor, and most reasonable for me to go about exercising the sullen canine staring at me expectantly from across the room. I believe, reasonably, that walking outside will be safe, even if I also believe reasonably that there’s some small amount of uncertainty in that belief.

Beyond Skepticism

While it is important for philosophical reasons to wrestle with the stronger forms of skepticism, we do not address them or include them in our ordinary beliefs on a regular basis. Perhaps this is an important lesson we can take from Moore: when we reason about our hands or any other manner of things on a day to day basis, we do not first consult our arguments for and against skepticism. For many beliefs, such as our core evidential beliefs in our senses, we simply trust that the evidence is so overwhelmingly irrefutable that there is no need to question or delve  further.

Our epistemology should match this thought process, not require that we jump through hoops to state that certain epistemological constructs are “knowledge” as opposed to belief. Moore’s argument may be an unconvincing way to prove knowledge of the external world, but it represents a prototype of how we construct reasonable beliefs: we construct whatever beliefs are reasonable on the basis of the evidence at hand.

We don’t question our fundamental evidential beliefs. We don’t consider the Brain in a Vat hypothesis as a basis for finding the best route to our office or comforting a crying child. The beliefs we do question are those for which we believe, on the basis of our evidence or lack thereof, that there is reason to doubt.

Our assessment of doubt and degree of trust in flawed evidence epitomizes most of the interesting and consequential beliefs we consider. If someone tells us something that contradicts our beliefs, how do we assess whether we should believe it or stay the course? What if we hear the same thing from a few people, or read something in a book? These sorts of questions lie outside the realm of what would normally be termed “knowledge”, but form perhaps the most crucial component of a day-to-day belief-based epistemology. Whether we seek to better understand ourselves or to understand others, the most important job of our epistemology must be to dig into these sorts of beliefs. We should guard against ossifying our beliefs into the bedrock of "knowledge". Especially in an era where different people possess different, conflicting accounts of knowledge, burying our beliefs where the light of introspection does us no favors. It is critical that we not only be able to shift our beliefs in the face of new evidence, but that we remain aware of the premises that support those beliefs and the strengths and weaknesses that those premises bring.

1. George Edward Moore was a late 19th century English epistemologist whose four-line proof of the existence of the external world can be summarized as “My hands exist, therefore the external world exists.” He did not hold skepticism in high regard.

2. This alludes to a deeper epistemological topic under the umbrella of belief-based epistemology: socially confirmed belief. Socially confirmed beliefs are those for which I rely on information obtained secondarily through others, rather than through direct experience. The transparency and structure of the networks along which such information travels can tell us a great deal about their reliability.

3. Descarte’s Demon was a philosophical though experiment from Renee Descarte’s Meditations in which he envisioned a powerful demon that could deceive us into believing in a falsified world. The consequences of such deception are similar to if we’re brains in vats, but for lovers of fantasy instead of science fiction.