Knowledge requires commitment (instead of belief)

Abstract

I argue that S knows that p implies that S is properly committed to the truth of p, not that S believes that p. Belief is not required for knowledge because it is possible that one could know that there are no beliefs. Being ‘properly committed’ to the truth of a proposition is a matter of having a certain normative status, not occupying a particular psychological state. After arguing that knowledge requires commitment instead of belief, I go on to demonstrate the theoretical utility of this hypothesis.

1 Introduction

Standard theories of knowledge take it that in order for one to know that p, one must believe that p. This is true of both internalist and externalist theories, and it is as true of knowledge-first theories as it is of those that take the concept of knowledge to be amenable to conceptual analysis. I would like to suggest, however, that one might know that a proposition is true without believing that it is true. Rather than believing that p, I shall argue that knowing that p requires being properly committed to the truth of p, where being properly committed to the truth of a proposition is a matter of having a particular normative status, rather than being in a particular psychological state.

In Sect. 2 I argue that believing that p is not a necessary condition on knowing that p. The problem with taking it that knowing that p requires believing that p is that it would render unknowable some propositions which, it seems, one might come to know. Consider eliminative materialism. Eliminative materialists argue that the explanations of human behavior provided by neuroscience represent dramatic improvements over those provided by psychology. Accordingly, it counsels that the ontological commitments native to psychology (most notably, but not exclusive, beliefs and desires) be dismissed. Since eliminative materialism implies that there are no beliefs, if knowing that p requires believing that p, it is not possible to know that eliminative materialism is true.

This conclusion does not seem to be well-supported. Eliminative materialism is an empirical hypothesis, which is to be established (or refuted) by scientific inquiry. And scientific inquiry, when it is conducted well and arrives at the truth, yields knowledge. So it is possible to know that eliminative materialism is true. Since it is, it follows that knowing that p does not require believing that p. Furthermore, I argue that the suggestion that knowing that p requires being properly committed to the truth of p does not face an analogous problem, and that commitments are well-suited to play the epistemic role usually reserved for belief. That beliefs can be both true and justified explains, in part, the appeal of assigning them a role in an epistemic theory. I shall argue that commitments can also be both true and justified; indeed, I shall argue that they can enjoy a specifically epistemic kind of justification.

In Sect. 3 I consider objections to my proposal. A theme common to several of them is that my hypothesis is subject to an argument similar to the one which shows that knowing that p does not entail believing that p. Such a challenge may be motivated by ontological naturalism—the thesis that the only things that exist are those posited by the sciences—or by reflection on epistemology itself. The problem with this suggestion, in whatever form it takes, is that we would not recognize something as a process of inquiry if it did not issue in cognitive commitments. A process of inquiry, be it naturalistic or otherwise, issues in an answer to a question. And, I shall argue, if one has answered a question, one has thereby taken up a cognitive commitment.

Finally, in Sect. 4, I show how my hypothesis resolves an apparent paradox generated by rationally compelling arguments. Consider that it is obvious that one can acquire knowledge from such an argument, but that it also seems clear that acquiring knowledge is a kind of achievement. If a rationally compelling argument compels belief, and if believing that p is necessary for knowing that p, it is hard to see how gaining knowledge by considering a rationally compelling argument can be an achievement. Being compelled to do something hardly seems to be an achievement. To resolve the paradox, I argue that rationally compelling arguments compel belief but provide reasons for commitment. Acquiring knowledge from a rationally compelling argument is an achievement because it involves deciding to take up the commitment that is supported by one’s reasons, not just being compelled to adopt a belief.

2 Belief and commitment

2.1 What commitments are

I will begin by discussing commitments generally, and then discuss the special case of commitment to the truth of a proposition. ‘Commitment’ is ambiguous; in one sense ‘being committed’ means having a particular kind of dedication. It is in this sense that environmental activists who chain themselves to a tree are committed to their cause. In another sense it means having a particular kind of normative status. Those who are married have undertaken a commitment in this sense. It is with commitment in the latter sense that I will be concerned.

In the general case, this is what I shall mean by ‘commitment’: a normative status that is properly taken up only if one has standing to do so, and which, if taken up, prohibits one from doing some of the things that one would otherwise be permitted to do.¹ Comment on both features is in order. Properly undertaking a commitment requires that one has already satisfied a set of conditions which gives one standing to undertake the commitment. Those who lack standing may nevertheless undertake a commitment, but it would be impermissible for them to do so. For example, selling a car involves undertaking a commitment (requiring, among other things, that one give the car to the person who bought it), and it is one that can be properly undertaken only if one owns the car that one is selling. Notice that lacking standing does not prevent one from undertaking the commitment. It is possible to sell a stolen car. Lacking standing does, however, make it impermissible to undertake the commitment in question.

Furthermore, undertaking a commitment restricts one’s options, not in the sense that it prevents one from doing things that one could otherwise do, but in the sense that it prohibits one from doing things that one would otherwise be permitted to do. Having sold my car, it would be impermissible to drive it to Toledo. Having accepted a job, it would be impermissible for me to sleep in on Monday morning. These restrictions are normative. It would be wrong for me to drive the car, or to sleep in, but having undertaken the relevant commitments does not prevent me from doing so.² Driving a car that doesn’t belong to you and skipping work are both possible, the point is simply that they are not permitted.

Our interest here is not with commitments in general, but with (to borrow a term from Robert Brandom) cognitive commitments, commitments to the truth of a proposition.³ Cognitive commitments have the same normative features as do commitments of other kinds. Just as being married or accepting a job offer restricts one’s options, so too does committing oneself to the truth of a proposition. If one is committed to the truth of the proposition p, one’s options are restricted in that one is prohibited from acting as though p is false and required to act as though it is true. And as with the restrictions that accompany getting married or accepting a job offer, this restriction is normative: one is prohibited, but not prevented, from acting as though it is false, and required, though not caused, to act as though it is true.

Furthermore, just as one must have standing in order to properly undertake commitments of other kinds, so one must have standing in order to properly undertake a cognitive commitment. In the case of selling a car, having standing requires owning the car. In the case of cognitive commitments, to have standing is to have evidence that the proposition in question is true. In fact, the two aspects of cognitive commitments—that having standing to take up a cognitive commitment requires having evidence, and that having taken up a cognitive commitment one must act as though the proposition to which one is committed is true—are related. Acting as though a proposition is true involves many things: using it as a premise in one’s reasoning, refraining from undertaking projects that would be frustrated were it to be false, and so on. Notable among the ways in which we treat propositions as true is that we assert them to others.⁴ Asserting propositions for which one does not possess evidence is a kind of uncooperative behavior. Moreover, relying, as we each do, on cooperative behavior from our fellows, we each have a practical reason to refrain from behaving uncooperatively towards them; our uncooperative behavior at one time is likely to produce uncooperative behavior in others in the future. We can then explain why having standing to take up a cognitive commitment would require evidence that the proposition in question is true: if it was otherwise, properly taking up a cognitive commitment would undermine one’s own interests, by encouraging one to behave uncooperatively towards one’s fellows, and so encouraging them to behave uncooperatively in return.

Finally, commitments of all kinds can only be undertaken voluntarily. Marriage vows said at gunpoint are not binding on those who say them. It is likewise with cognitive commitments. Having undertaken a commitment to the truth of p, one is thereby obligated to act as though p is true. And just as it would be unreasonable to hold someone responsible for discharging the obligations involved in a marriage if they were undertaken accidentally or against one’s will, so it would be unreasonable to hold one responsible for discharging the obligations attendant upon a cognitive commitment, if they were undertaken accidentally or against one’s will.

2.2 Knowledge and eliminative materialism

I shall argue that it is possible to know that there are no beliefs. The argument, in short, is this: eliminative materialism is a scientific hypothesis, which could be revealed, by scientific investigation, to be true. When it is conducted well and arrives at the truth, scientific investigation produces knowledge. So, it is possible to know that eliminative materialism is true. Since eliminative materialism is, in part, the thesis that there are no beliefs, it is possible to know that there are no beliefs.

Eliminative materialists hold that psychology does a poor job of predicting and explaining behavior. Neuroscience, they claim, provides more accurate predictions and better explanations. As psychology has been superseded by neuroscience, they hold that the ontology of psychological states—including beliefs and desires—ought to be dispensed with, in much the same way that we removed phlogiston from our ontological theories when it ceased to serve a scientific purpose.⁵

According to nearly all accounts of knowledge, if one knows that p it follows that one believes that p. Now, if eliminative materialism was true one could not believe that it is true (since it implies that there are no beliefs). So according to nearly all accounts of knowledge, if eliminative materialism was true one could not know that it was.

But this seems to be wrong. Eliminative materialism is an empirical hypothesis, one whose truth or falsity is to be established by scientific means. Scientific inquiry is an improvement over other means of answering empirical questions, for example, guessing and astrology. Plausibly, when it is conducted well and arrives at the truth, the fact that it is an improvement over these other means of answering questions is marked by a difference in their products. For example, astrology, even when conducted well and even when it arrives at the truth, does not yield knowledge. Whereas scientific inquiry does. Indeed, given that eliminative materialism is a scientific hypothesis, to deny that one could know that it is true would reflect an unacceptable kind of skepticism about science.

It seems, then, that it is possible to know that eliminative materialism is true. If so, then it follows that one might know that p without believing that p. The intuition driving the thought that one must believe that p if one is to know that it is true is that knowing that p requires, in some way, that one endorse p, that one has, in some way or other, given the proposition one’s approval. Believing a proposition is probably the most common means by which this is done, and so it is natural to think that knowledge requires belief. But it is not the only way.

I am not the first to have suggested that knowing that p requires that one give one’s approval to p in some way other than by believing it. For example, Keith Lehrer and L. Jonathan Cohen have developed related accounts of acceptance, both of which allow it to play many of the same epistemic roles as have usually been reserved for belief.

Here is what Lehrer has to say about acceptance:

Acceptance is the sort of mental state that has a specific … functional role, in thought, inference, and action. When a person accepts that p, he or she will draw certain inferences and perform certain actions assuming the truth of p. Thus, if a person accepts that p, then the person will be ready to affirm that p or to concede that p in the appropriate circumstances. They will also be ready to justify the claim that p. If they accept information received from the senses or retained in memory, they will regard such information as correct and proceed accordingly in thought and action. (Lehrer 1990: 35.)

Cohen has a similar view:

to accept that p is to have or adopt a policy of deeming, positing, or postulating that p—that is, of going along with that proposition (either for the long term or for immediate purposes only) as a premise in some or all contexts for one’s own and others’ proofs, argumentations, inferences, deliberations, etc. Whether or not one assents and whether or not one feels it to be true that p. Accepting is thus a mental act … or a pattern, system, or policy of mental action (Cohen 1989: 368)

Lehrer argues against the belief condition on knowledge, suggesting that one must instead accept that p in order to know that p. (Lehrer 1981: 79) Although he draws a distinction between beliefs and acceptances, they are intimately connected. On Lehrer’s account, acceptances are higher-order cognitive states, which allow one to stand in judgment over one’s own beliefs.⁶

Notice that, by all accounts, acceptance is just as much a psychological state as is belief. Indeed, Pascal Engel suggests that it may be best to see acceptances and beliefs as differing in degree along various dimensions, but not differing in kind. (Engel 2012: 23) Those who want to draw a sharper distinction between beliefs and acceptances also insist that acceptances, like beliefs, are psychological states. For example, it is a part of Cohen’s project to show that cognitive science would be unable to provide satisfactory accounts of human behavior if it did not reserve a role for acceptances alongside beliefs. (Cohen 1989: 388)

The problem with the hypothesis that knowledge requires acceptance is fundamentally the same as the problem faced by the hypothesis that knowledge requires belief. The claim of the eliminative materialist is that there are no psychological states. And so eliminative materialism implies that there are no acceptances, just like it implies that there are no beliefs. Since it is possible to know that eliminative materialism is true, accepting a proposition must not be necessary in order for one to know that that proposition is true.

Now, I said above that the intuition behind the thought that knowledge requires belief is that one must, in some way, give a proposition one’s approval if one is to know that it is true. But this ‘giving of approval’ must not take the form of adopting a psychological attitude towards the proposition, for it could turn out that eliminative materialism is true, and hence that the attitude in question does not exist, and yet it would still be possible to know that eliminative materialism is true. In fact, the same argument can be run, with only slight modifications, to show that for one to know that p it is not necessary that one meet any empirical conditions (beyond those implied by p itself), because, for any such condition, empirical inquiry could show that nothing satisfies it. In such a case, we should be able to know that the condition in question is not met. So just as the fact that one knows that p cannot imply that one believes that p, it cannot imply that any other empirical conditions (again, beyond those entailed by p) are met, either.

And yet there must be some sense in which those who know that p give their approval to p. My suggestion is that those who know that p are committed to the truth of p, whether or not they believe or accept the proposition that p. Cognitive commitments are neither beliefs nor acceptances. Beliefs and acceptances are theoretical entities posited to predict and explain human behavior. By contrast, cognitive commitments are causally inert normative statuses. We do make sense of human behavior by appeal to normative statuses—as when we say that a chess player moved his king because he was in check—but such explanations differ dramatically from psychological explanations. They make sense of behavior without providing a causal explanation of it. What caused the player to move the king is not that he was in check, it was that he believed that he was in check. Similarly, beliefs about cognitive commitments, and affective attitudes that take commitments as their objects, may play a role in shaping behavior, but the commitments themselves do not.

Given the hypothesis that knowing that p requires being committed to the truth of p, it is easy to explain why one might think that knowing that p requires that one believe that p. Typically, one is committed to all and only the propositions that one believes, and so there is rarely call to distinguish one’s commitments from one’s beliefs. Consider, opponents of doxastic voluntarism have long argued that our beliefs are keyed directly to our evidence. Watching the snow fall through my window, I am caused to believe that it is snowing; there seems to be no choice available to me, no option to do otherwise. And while I am not merely caused to undertake a commitment to the truth of the proposition it’s snowing outside by seeing the snow, seeing the snow does provide me with standing to undertake a commitment to that proposition. It is natural, then, that one would be committed to the propositions that one believes, and believe the propositions to which one is committed. The evidence which causes one to adopt a particular belief also typically provides one with grounds for undertaking a commitment to the same proposition. Running the two notions together is thus natural and, in most contexts, harmless.⁷

They are, however, distinct. And cognitive commitments seem to possess those traits that would allow them to play the role in an account of knowledge that was traditionally reserved for beliefs. Beliefs are states with content, which are capable of being both true and justified. Cognitive commitments can have all of these properties. To begin with, cognitive commitments, just like beliefs, have content. Just as I can believe the proposition that it is snowing, I can be committed to the truth of the proposition that it is snowing. Beliefs are candidates for truth because they have content. Since cognitive commitments have content, they are also candidates for truth.⁸ Finally, it is easy to see how a cognitive commitment could be justified. It is justified when those who hold the commitment have standing to undertake it. It would be good to notice that this is epistemic justification: having evidence that a proposition is true constitutes having standing to undertake the commitment in question. So not only are cognitive commitments subject to justification, they are subject to the right kind of justification to play the role in an account of knowledge that was traditionally reserved for beliefs.

However, not every cognitive commitment is a candidate for knowledge. In particular, cognitive commitments that are inappropriately undertaken are not. If I am committed to the claim that it will be warm tomorrow, but have undertaken this commitment simply due to wishful thinking, I do not know that it will be warm tomorrow. This is as it should be: failing to have standing to undertake the relevant commitment, I also fail to satisfy the justification condition on knowledge.

To summarize: in order for one to know that p, one must, in some way, give one’s approval to the proposition that p. This cannot be a matter of taking a psychological attitude towards p, since it could turn out that there are no psychological attitudes, and if there were no such attitudes, it would be possible for us to know that fact. Similar remarks apply to any state which empirical inquiry might show to not exist. Cognitive commitments are not posits of an empirical theory. The beliefs that people have about their commitments may play a role in scientific explanations, as might affective reactions to those commitments. But the commitments themselves find no place in the empirical sciences, and the reason to believe that there are commitments is not derived from their role in a scientific theory. Furthermore, properly undertaken cognitive commitments are well-suited to play the needed role: they have content, and they are subject to epistemic justification. They look, then, like likely candidates to play the role usually reserved for beliefs in the theory of knowledge.⁹

3 Objections and replies

3.1 One cannot (coherently) believe that there are no beliefs

Eliminative materialists take it that beliefs are akin to phlogiston: ontological commitments of false theories. But notice that when the phlogiston theory of combustion was rejected, scientists took it to be—believed it to be—false. It seems that there is no comparable attitude that one could adopt towards the claim that there are no beliefs.¹⁰ So perhaps eliminative materialism is not really an empirical hypothesis, as there seems to be no coherent way that one could discover that it is true. If eliminative materialism is true, then even if the empirical evidence unequivocally indicates that it is, the eliminative hypothesis is literally unbelievable. In fact, since eliminative materialism entails that there aren’t any psychological states, it seems that there is no attitude that the eliminative materialist can take towards her hypothesis that would constitute endorsing it.

And that’s right, but it does not follow that one could not discover that eliminative materialism is true. If eliminative materialism is true, there is no propositional attitude that one can take towards the hypothesis that eliminative materialism is true. But what one can do is enter into the neural state that will cause one to say ‘eliminative materialism is true’, to deny that there are beliefs, to feel the sense of conviction that we ordinarily (if mistakenly) associate with having a belief, and so on. The eliminative materialist need not believe her own hypothesis nor take any other attitude towards it, what she does need to do is enter into the brain state whose explanatory utility will (she alleges) render the hypothesis that there are beliefs obsolete.

That this is a possibility is essential to eliminative materialism, but it, or something like it, is also essential to psychology. Psychology aims to provide scientific explanations of human behavior in terms of internal states with representational properties. Now, scientific hypotheses must be falsifiable, so that we might discover that there are no psychological attitudes is necessary if psychology is to be properly scientific. Discovering that there are no psychological attitudes cannot, of course, entail believing that there are no psychological attitudes. So if psychology is to be properly scientific, something like the situation envisioned by the eliminative materialists must be possible.

3.2 A replacement for knowledge?

I take these reflections to indicate that it is possible for one to know that there are no beliefs. But one could draw a different lesson from them. One could insist that believing that p is a necessary condition on knowing that p, and argue that if it is shown that there are no beliefs, it is thereby shown that there is no knowledge. Now, there are some problems with this proposal. To begin with, it would provide the skeptic with an empirical argument: if human behavior is not best explained by adverting to psychological states, then the skeptic wins.¹¹

Having to face the threat of universal skepticism is too high of a price to pay for retaining the belief condition on knowledge. The claim that I know that I’m sitting at my desk ought not to be hostage to developments in neuroscience. However, one might argue that if eliminative materialism is true then a suitable replacement for knowledge can be found. This state would involve no scientifically dubious psychological states, rather it would make reference only to scientifically respectable brain states. So even if it turns out that knowledge is not possible (because it requires belief), the resulting epistemic consequences are mitigated by the possibility of an acceptable replacement for knowledge.¹²

How ought one to respond to this suggestion? Not, I shall argue, with equanimity. A crucial part of the eliminative materialist’s case against psychology is that the semantic properties of propositional attitudes do little explanatory work. So whatever brain states supplant beliefs will not have semantic properties. The brain state that one enters into when one looks at a red apple does not, for example, mean that the apple is red, although it may cause one to say that the apple is red. The problem with a materialist replacement for knowledge is that, because brain states do not have content, the replacement for knowledge will, in fact, be very little like knowledge.

Consider first that, although there are two kinds of epistemic justification, brain states admit of neither.¹³ An epistemic agent can be justified, in, for example, endorsing a well-supported proposition.¹⁴ And a state that constitutes endorsing a proposition (such as believing a proposition) can enjoy propositional justification. Such a state is propositionally justified for a person if and only if the proposition that is its object is likely to be true, given that person’s evidence. Call these kinds of justification ‘agential’ and ‘propositional’ justification, respectively.¹⁵

Because they lack content, brain states cannot be justified in either way. It is clear that propositional justification cannot attach to brain states, since they do not have propositions as their objects. Moreover, it seems that, for much the same reason, those who enter into a brain state are not thereby candidates for possessing agential justification. Traditionally, justification was taken to accrue to an epistemic agent due to the evidential basis on which she forms or maintains her beliefs. The problem is that brain states are the wrong kinds of things to enter into evidential relationships. Evidence for p indicates that p is true. But truth is a property of propositions. And so evidential relations bear between some set of evidence, and a proposition. Believers can derive agential justification for holding their beliefs because beliefs have propositional content, and so, given their evidence, holding a particular belief may be to a believer’s credit. But since brain states have no content, no evidence can tell for or against them. And so it cannot be to one’s credit for being in a particular brain state, in the way that it can be for holding a particular belief.¹⁶

So it seems that whatever state stands to replace knowledge cannot have a justification condition. Perhaps this is not very troubling. After all, reliabilists argue that knowledge does not have a justification condition. But on any conception of knowledge, the truth condition is non-negotiable. Notice, however, that since brain states have no propositional content, just as they cannot be justified, so they cannot be true.

Thus the replacement for knowledge must have neither a belief, nor a justification, nor a truth condition. These three conditions were traditionally each taken to be necessary for knowledge. I submit that any property which requires none of these conditions is not a replacement for knowledge. The proposal mooted above suggests not replacing epistemic concepts with slightly different but scientifically respectable concepts, it suggests rejecting epistemology entirely.

3.3 Maybe there aren’t any cognitive commitments

The strategy that I have pursued here is to argue that if empirical inquiry could show that there are no φs, then there being a φ is not a necessary condition for S to know that p (unless that there are φs is implied by p), because empirical inquiry is a path to knowledge, and if it were to show that there are no φs, then one would know that there are no φs.¹⁷ More generally, one could say that if it is possible for a knowledge-yielding process of inquiry to show that there are no φs, then there being φs is not a necessary condition for one to have knowledge.

My present concern is that there might be a knowledge-yielding process of inquiry which shows that there are no commitments. If so, then one could know that there are no commitments, and, hence, being committed to the truth of a proposition is not a necessary condition on having knowledge. There are a couple ways in which this argument might be developed. It could be that epistemic inquiry will show that there is no place in epistemology for cognitive commitments.¹⁸ Alternatively, perhaps cognitive commitments can be naturalized. In which case, empirical inquiry could show that they do not exist, and the same argument which shows that believing that p is not necessary for knowing that p would show that being committed to p is not necessary for knowing that p.¹⁹

I doubt, however, that any process of inquiry could show that there are no cognitive commitments. The fundamental problem faced by both of the possibilities mentioned above is that what inquiry produces, when it is successful, is a cognitive commitment. So, I shall argue, no process of inquiry could coherently lead to the conclusion that there are no cognitive commitments, as no inquirer can coherently be committed to the truth of the proposition that there are no cognitive commitments.

Now, what makes something a process of inquiry is that it is aimed at arriving at a rational answer to a question.²⁰ Consider first what is involved in answering a question. Answers close inquiry. If one consults the schedule and concludes that the train will arrive at four, one no longer has the option to act as though the time of the train’s arrival is still in question, or that it is any time other than four. Now of course one could act as though the question is still open, or that the train arrives at, say, three, but doing so puts the lie to one’s earlier conclusion. Reaching the conclusion of a process of inquiry resolves the inquiry in a particular direction; doing so makes it objectionable to treat the inquiry as still open, or as closed in a different direction. It is objectionable in part because it reveals a flaw in one’s character: it shows one to be flighty or indecisive, both of which are epistemic vices. But also, and more importantly, it is objectionable as a part of the practice of inquiry. Those who answer a question and then go on to treat the question as still open are not conducting inquiry well, they are failing to achieve the central purpose of the practice.²¹ Notice that there are two elements to answering a question: taking the matter to be closed (by endorsing a particular answer to the question), and it being objectionable to treat the matter as if it is still open.²²

A process of inquiry aims not simply at answering a question. The answer must, in some relevant respect, be a rational one. This is what is needed to differentiate processes of inquiry from other ways of answering questions (for example, guessing). That said, we should not demand too much from the rationality requirement. Many processes of inquiry produce only rudimentary reasons, and they are often not consciously available. But if I look out of my window and see snow, that I see snow under apparently normal conditions may be a reason to think that it is snowing. If I were merely to guess that it is snowing, any belief that results would not have even a rudimentary reason like this one behind it.

Moreover, if the reason that one acquires is not an epistemic reason then the process in which one is engaged is not one of inquiry. Pascal argues that everyone should believe in God because the expected utility of doing so is higher than the expect utility of not doing so. Now, while Pascal recognized that we do not have the ability to believe what we want at will,²³ let us assume that there is someone who has this capacity, and so reading about Pascal’s wager is sufficient to induce belief in him. For this person, undertaking a cost–benefit analysis is one way of answering the question ‘does God exist?’. But it is clear that undertaking a cost–benefit analysis is not a way of engaging in inquiry. The reasons that cost–benefit analyses produce are of the wrong kind. They provide one with practical reasons, not evidence.

It seems, then, that when inquiry into the question whether p is conducted successfully, its product will have three properties: (1) it will provide one with an epistemic reason to (2) endorse a particular answer to the question which (3) one is then obligated to treat as the answer to the question. And these are precisely the conditions on having a cognitive commitment. An epistemic reason to endorse an answer to a question constitutes standing to take up the corresponding commitment, and if one endorses a particular answer one is prohibited from treating it as if it is false, that is, one is obligated to treat it as the answer to the question. So it seems that what makes something a successfully conducted process of inquiry is that it leads to the adoption of a cognitive commitment which one has standing to undertake. I conclude, then, that a process of inquiry could not coherently lead one to the conclusion that there are no cognitive commitments.

What about the naturalist’s challenge? Ontological naturalism is, perhaps, the most prominent contemporary metaphysical program. The ontological naturalist’s program is to show that the only things that exist are those that find a place in the sciences,²⁴ by showing, for each purportedly non-natural part of the world, either that it simply does not exist, or that, appearances to the contrary, it is identical to some natural part of the world. The ontological naturalist’s targets for elimination or reduction are many and varied, but normative properties feature prominently among them.²⁵Being committed is a normative property, and so, presumably, among those that are targets for naturalization. This would seem to be a problem. If commitments can be naturalized, then empirical inquiry could show that they do not exist. And so, if they can be naturalized, then, by my own reasoning, being committed to the truth of p is not necessary for knowing that p.

The naturalist is surely right that anything that can be naturalized can be eliminated. But I have argued that cognitive commitments cannot be eliminated, or, at least, no process of inquiry can show that there are no cognitive commitments. If that argument was any good, we should conclude that, since cognitive commitments cannot be eliminated, they cannot be naturalized. Hilary Putnam has long argued that scientific practice involves appeal to the normative: that judgments of, for example, coherence, are irreducibly normative, and that if one is to engage in scientific inquiry one must take some such judgments to be true.²⁶ The argument recently reviewed suggest an amendment to Putnam’s project. It is not only that the process of scientific inquiry involves the appeal to irreducible norms, the product of such inquiry does as well.²⁷

The next section provides an example of the utility of the belief/commitment distinction. There is a puzzle about rationally compelling arguments. It is hard to see how rational influence could be compelling. There might, it would seem, be a causal story to tell about how a rationally compelling argument could compel belief, but it does not seem that this would be a story of rational influence. I argue that rationally compelling arguments are rational because they provide reasons for commitment, and compelling because they compel belief. Drawing this distinction also allows me to explain how the knowledge gained through consideration of a manifestly sound argument might be to a knower’s credit. The knower has the option to not take up the commitment recommended by the argument that she considers. However, it is to her credit if she does take it up, because in doing so she recognizes her reasons for what they are, and acts accordingly.

4 Rational compulsion

Consider a manifestly sound argument. For example: Books have pages and trees have leaves, therefore books have pages. The conclusion of such an argument is, in a certain way, non-optional for rational agents. The argument compels assent from rational agents. Obviously sound arguments like this one are not the only things that can make a proposition non-optional. Kidnapping someone and brainwashing them can have the same effect. But there is clearly a difference between the two cases. The influence of an obviously sound argument is a rational influence, whereas the influence of brainwashing is not.

Now, there appears to be a problem. If obviously sound arguments have non-optional conclusions, rational agents have no choice but to accept them. Notice that the way in which they are non-optional is causal, it is not as though rational agents are merely blameworthy if they fail to believe the conclusion of an obviously sound argument; rather, their appreciation of the fact that the argument is obviously sound causes them to believe the conclusion of the argument. This is at odds with the suggestion that the influence of an obviously sound argument is a rational kind of influence. Standards of rationality govern behaviors that are optional, those that an agent can perform, or can refrain from performing.

To say as much is not to endorse libertarianism. There are many strengths of modality that might be picked out by the word ‘can’, some of them will render the foregoing consistent with compatibilism. But on no account is compulsion consistent with rational influence. Consider the following two thought experiments. In the first, you are approached by a robber with a gun, who says ‘your money or your life!’ Seized by fear, you hand him your money. Not giving him the money was not an option for you, not only in the sense that the threat was sufficient to induce you to surrender your money, but also in the sense that, at the time, you did not see multiple courses of action as available. At the sight of the gun, your fear automatically led to you surrendering your money. There was not any decision to be made. Contrast this case with the next one. In this case, you are approached by a robber with a gun, who says ‘your money or your life!’ You examine the gun, consider what would happen if you were you to be shot, determine that there is only a very low chance that you could wrest the gun from your assailant without being shot, in short determine the expected outcome of refusing to surrender your money. And then you consider the expected outcome if you were to surrender your money. Deeming that the latter is better than the former, you give the robber your money.

In both cases the threat of the gun is sufficient to induce you to give the robber your money. But in the second scenario, you have the option not to, at least in that a choice is involved. Notice that you have a choice, even if the choice that you will make is causally determined by the reasons as you see them. (Hence there is no need to subscribe to libertarianism.) It makes sense to say that, in the former case, your action was simply compelled by the threat of violence, in the latter case it was a product of a rational decision.

What is puzzling about rationally compelling arguments is that they are more like the scenario in which you surrender your money out of sheer terror than the one in which you soberly weigh the pros and cons of giving up your money. The obviously sound argument reviewed above is a simple case of conjunction-elimination. Given that the only inference rule invoked is so simple and obviously valid (and that the premise of the argument is so obviously true), anyone with any familiarity with logic (or even clear thinking) will not face a decision about whether or not to believe the argument’s conclusion. The argument does not provide reasons on the basis of which one can deliberate and arrive at a decision, it is like the threat in the first thought experiment: it simply compels belief. And so it does not seem that the process is a rational one, it is merely and brutely causal.

So we face a paradox. It seems obvious that there is a difference between giving someone a good argument and brainwashing them. But rationally convincing someone to do something requires providing them with a choice about which a decision is possible. And in the case of obviously sound arguments, the agent cannot make any decisions about what to believe. If rational, their belief in the conclusion of the argument will follow immediately from being presented with the argument, and so will not depend on any decision that they make. The solution to this paradox, I would like to suggest, is to distinguish beliefs from commitments, and see that rationally compelling arguments compel belief, but provide rational grounds for commitment.

The literature on doxastic voluntarism provides independent reasons to think that beliefs are not under the voluntary control of believers.²⁸ So when considering the doxastic result of rationally compelling arguments, it seems advisable to take appearances at face value, and conclude that those beliefs that result from these arguments really are compelled. But commitments cannot be compelled. If Andy forces Beth to marry him at the point of a gun, she is not thereby under any obligation to love, honor, and cherish him, nor is she justly subject to criticism if she lacked standing to marry him. So we can make sense of the idea that obviously sound arguments are both compelling, and exert a rational influence. Beliefs are subject to many kinds of compulsion—seeing the snow falling I am compelled to believe that it is snowing, remembering my high school civics class I am compelled to believe that the United States is not a colony of England—and exposure to obviously sound arguments is simply one of the ways in which they are compelled.

But they also exert a rational influence on us. Properly taking up a cognitive commitment requires that one have evidence that the proposition in question is true. And this is what obviously sound (and so rationally compelling) arguments provide. Exposure to such an argument thus gives us a reason to take up a cognitive commitment; in particular, it allows us to see that we would not be subject to criticism for lacking standing if we took up the commitment, and we would be subject to criticism if we were to take up any commitment to any incompatible proposition.

One final issue demands attention. Virtue epistemologists make much of the fact that gaining knowledge is a kind of achievement. Sosa compares it to an archer hitting a target through an exercise of skill.²⁹ It is worth noting that firing an arrow is something that is done on purpose, and over which archers have discretion. Now, encountering an obviously sound argument is one way to acquire knowledge. If knowing that p required believing that p, and since obviously sound arguments compel belief, it is hard to see how acquiring knowledge in this way amounts to an achievement. The knower is fortunate that she encountered the argument and was such as to be compelled by it, but there is no achievement here, it is as though a skilled archer fired her arrow involuntarily. She may be such as to hit the target even under such circumstances, and may be praiseworthy in that respect, but praise for the shot itself would be misplaced. But if knowing that p requires that one be committed to p, rather than believe it, there is an explanation available of why acquiring knowledge from a rationally compelling argument is an achievement creditable to the knower. The argument gives the knower a reason to undertake the relevant commitment. Now, agents do not always act in accordance with their reasons. When they do, it is to their credit. And so it is to the knower’s credit when she decides to act in accordance with her reasons, and undertakes commitment to the propositions for which she has evidence. It is the degree of optionality that comes with commitments, but not beliefs, which explain why knowledge acquired through consideration of rationally compelling arguments represents a kind of achievement creditable to the knower.

5 Conclusion

That S knows that p does not entail that S believes that p. Consider the proposition that there are no beliefs. If this were true it would be possible to know that it was true. If so, then knowing that p does not entail believing that p. Similar reasoning shows that knowing that p does not entail that any other empirical conditions, besides those entailed by p itself, obtain.³⁰ However, in order to know that p one must, in some sense, have endorsed it. Taking up a commitment to the truth of p is one way of endorsing it.

Furthermore, cognitive commitments are well-suited to play the epistemic role that has traditionally been reserved for beliefs. Cognitive commitments have content, and so they can be true or false. They are also subject to epistemic justification, as having standing to undertake a cognitive commitment requires that one have evidence that it is true. Finally, there is some theoretical utility to taking it that knowledge requires commitment rather than belief. It allows us to explain how the knowledge that one gains from considering rationally compelling arguments is to one’s credit. Rationally compelling arguments compel belief, but provide reasons for commitment. If knowing that p requires being committed to the truth of p, then those who acquire knowledge from rationally compelling arguments act in accordance with their reasons. And when one acts in accordance with one’s reasons, it is to one’s credit.