Introduction

1In the first appendix to On What Matters I, Derek Parfit takes up a much-discussed problem in moral psychology and meta-ethics, that of the “right kind of reason.” [1] This problem is crucial for a theorist who intends, as Parfit does, to defend a “reasons fundamentalism”—that is, the thesis that the fundamental normative concept is neither axiological nor deontic, but is instead based on reasons. This concept is itself ambiguous, because it seems reasons can be both objective reasons (facts) and attitudes (beliefs and desires). Parfit dispels the ambiguity by distinguishing between “object-given” reasons, which are based on the objects of agents’ desires and beliefs, and “state-given” reasons, which are based on states (or attitudes), reasons we may have for having a particular desire or belief. Parfit is an objectivist with regard to reasons, and this thesis is essential to his meta-ethics. It means that his position is opposed both to Humeans, for whom all of our reasons must be based on desires, and contractualists like Rawls, who claim that “there is no way to get beyond deliberative rationality,” that is, beyond reasons based on our ideal rational attitudes. [2] But Parfit’s position runs up against the problem of whether the only good reasons are object-given ones, or whether there can be reasons that are based on attitudes. In this paper, I discuss Parfit’s approach to the problem, and defend a version of objectivism close to his own. [3]

Nihil appetimus nisi sub ratione boni

2The problem of the “right kind of reason” is a descendant of those questions raised by a scholastic formula: Nihil appetimus nisi sub ratione boni, nihil aversamur nisi sub ratione mali. [4] On an “objective” reading of this formula, it is good and evil that guide and direct our desires. In other words, evaluation determines motivation. It is objective goodness and badness that determine the content of our attitudes, which are intrinsically directed towards this bad, or this good (what we desire is by definition good, and what we do not is by definition bad). So it is difficult to understand how we can desire something bad, or that we judge to be bad, as in the cases of akrasia, acedia, and perversity so familiar to any reader of the Divine Comedy. On a “subjective” reading of the formula, by contrast, it is the attitude or motivational state (i.e. the desire itself) that determines the goodness or badness of what is desired. The formula becomes a tautology: what is desired is eo ipso good because it is desired. While we do evaluate it as good or bad, we do not do so independently of our desires and motivations. In other words, there is no external criterion for how well adapted the objects of our desires are to our motivation. As a consequence, criticism of our desires is impossible: they are by definition good as long as they are indeed our desires. The only mistakes we can make in our actions involve our beliefs (we may be wrong about the facts, or we may be inconsistent), not our desires, which alone can motivate us. Our beliefs (what Hume calls “reason”) cannot in themselves motivate us. This is often called the Humean theory of motivation. [5]

3The opposition between the objective and the subjective conception of the relationship between desire and values is repeated when we formulate the motivation in terms of the reasons that agents have. This formulation has become a classic in the Humean tradition, where reasons for acting are viewed as causes of our actions. [6] One objection to this conception of reasons is that, by treating reasons as psychological states (desires and beliefs), it does not explain the objectivity of reasons and values. We can be driven by beliefs about what we should do or what it is good to do without having the corresponding desires. [7] The Humean theory seems to have nothing to say that can account for the normative character of reasons. To take a famous example from Williams, [8] suppose I want a gin and tonic, and believe the glass in front of me is full of gin whereas, unknown to me, it is actually full of petrol. I certainly have a reason, in the causal, motivating sense, to put tonic in the glass and drink it. But in the normative sense I have no reason to drink it, because doing so would make me seriously ill. This objection is fundamentally the one made by meta-ethical cognitivism. But while it involves positing values and norms that are real and external to agents, it runs up against both the classic ontological problems and also that of precisely how normative facts, independent of our desires, can motivate us. There is a position that seems to avoid the pitfalls of both the Humean and the cognitivist conceptions. This is the view, inspired by Brentano’s conception of value, [9] that a thing X has value if and only if the agent has the well adapted or correct attitude towards X. The “buck-passing” conception of normativity is based on the same idea: something has value if the agent has justified reasons for having positive attitudes towards the thing, and for acting in a certain way towards it. [10] On this conception, value (and every other normative concept, like duty, or permission) can be reduced to having reasons. On this view, value is neither subjective in the non-cognitivist sense, nor objective in the cognitivist sense: it depends on our reasons, but these are objective.

4The whole problem is knowing what makes a good kind of reason and what does not. What makes the two kinds different? Does the difference come from the nature of the states or attitudes governing the use of reason in any particular domain? Following a distinction made famous by Anscombe, [11] our beliefs and cognitive attitudes are aimed at truth and have a direction of fit that goes from mind to world, whereas our desires and conative attitudes are aimed at satisfaction and have a direction of fit that goes from world to mind. Or should we say that the difference between good and bad reasons comes from the type of target aimed at in each domain, and the respective values and norms—goodness and truth—that orient each one? If we adopt the neo-Brentanian or buck-passing conception, our sensitivity to values depends on the possibility of formulating correct judgments in each field, where a correct judgment is one in which the attitudes are well adapted to their object, the true or the good. An attitude of belief is well adapted or correct if it is true and rationally grounded, and an intention or desire is correct if it conforms to what is desirable. On this conception, normative weight shifts from values to reasons: an attitude is correct if it is based on good reasons. In the most general sense, a reason is a “consideration” that “favors” an attitude, which has to do both with a fact about the world and our relationship with this fact. For example, the fact that the building is on fire is a (normative) reason to jump out of the window. But the fact by itself cannot motivate an agent if they do not know it. When they come to realize the fact, the normative reason becomes their motivation to jump out the window. But the agent could have had all sorts of other reasons to jump out the window (for example, he might have bet a lot of money that he was capable of it). The problem is knowing what the appropriate normative reasons are.

5There is an objection to this position: [12] how can we account for cases in which we seem to have good reason to do something, even though these reasons do not correspond to the appropriate attitude one should have towards this thing? Here is the canonical example. Suppose a powerful demon threatens to torture you to death unless you believe that 2 + 2 = 5, or your boss threatens to fire you unless you believe their neon orange and green tie is extremely chic. The threat of torture means you have excellent reasons to believe what the demon or your boss tells you to believe. But the propositions the demon and your boss demand you believe are not based on appropriate reasons: one is false, the other is completely absurd. It is not correct to think that 2 + 2 = 5 because it is false, and it is not correct to admire your boss’s hideous tie because there is no reason to admire anything hideous. There is no doubt we may come to believe what the demon or the boss tells us to believe—belief is not voluntary, but powerful incentives can lead us to believe, and indirect means like rewards or torture give us various means to do so—but the reasons in these cases do not seem to be the right ones. It is not just that they are not right in the sense of being useful: they are not right in the sense that they are not of the right kind. If a reason is what counts in favor of an attitude, there seem to be considerations “in favor” of believing (or admiring, or any other attitude) what the demon pushes us to believe by threatening us. But these are not relevant considerations for a belief in general. The good reasons to believe are the evidence we have. When our reasons are dictated by our interests or by reasons unrelated to the evidence, they are reasons of the wrong kind, not the right kind. Classic examples where instrumental reasons or utility dominate the formation of beliefs or intentions include Pascal’s wager, where instrumental reasons for believing seem to outweigh those for believing on the basis of evidence, and Kavka’s toxin puzzle, [13] where the reward associated with the formation of an intention seems to outweigh the formation of the intention itself. The distinction between good and bad reasons applies to every attitude. If the demon asks me to want to drink a saucer of mud, or if my boss threatens to fire me unless I admire his tie, the problem is the same. It is the same with propositional attitudes like hope, for instance: if the demon threatens me with dreadful torture unless I hope that Johnny Depp gets back together with Vanessa Paradis, I will have just as much reasons as in the other cases to have this inane hope. The distinction applies to emotions and passions like love and hate: you may love or hate someone or something, but this is not the same as wanting to love or hate someone in view of somehow profiting from it. The good reasons for loving are love for the object, and not the desire to love the object for some other purpose. The general problem of the right kind of reasons is that there seem to be reasons for having an attitude which are inappropriate to the object of these attitudes.

6But the difference between good and bad reasons is still only intuitive. What is it based on? We might say that the right kind of reasons for belief are epistemic reasons—proof, rationality, and justification—while the wrong kind of reasons are pragmatic, or instrumental. But that explains nothing: why are our epistemic reasons by definition the good ones? Is it because they are epistemic, and belief is an epistemic state? This is tautological. And why should fear of punishment or interest not be reasons to believe that are just as good as proof or justification? Perhaps we could say that epistemic reasons are intrinsic or essential to belief, whereas pragmatic reasons are extrinsic or accidental. But this also seems tautological: what basis do we have for saying that some reasons are essential or constitutive, and others are extrinsic? At first sight, a more satisfying answer is simply that pragmatic reasons for believing simply are not reasons to believe, but are instead motives to believe. The demon’s threat is not a reason to believe that 2 + 2 = 5, but a very strong motive or incentive. The same is true of the fear that my boss will not promote me unless I think their tie is chic. The distinction here would be similar to the one between having a reason and having an excuse or a pretext. An excuse can be a good reason but, even when it is, it is not the right kind of reason. But this answer, which has been called “skepticism about the problem of good and bad reasons,” [14] also seems to beg the question: why could the threat not be my reason, and why is it not an appropriate reason in the circumstances? There seems to be something ad hoc about the distinction. We have to explain why it is that reasons of a certain sort are good and others are not.

7In fact, this is exactly the problem Hume raised in saying that reason is not practical by itself. The issue is knowing which attitude is consistent with reason—which one is rational. Hume says that it is not against reason to prefer the destruction of the world to a scratch on my finger. In other words, all our desires—including the most frivolous, like wanting Vanessa Paradis to get back together with Johnny Depp—are equally “rational”: reason is the slave of desire. From the Humean point of view, the answer to the problem is extremely simple: every reason is a good reason, as long as it consists in or is motivated by a desire. From the Humean point of view, asking why a reason is good is simply asking why a reason is a reason. And the answer is therefore trivial, or circular. The Humean answer to the problem of the nature of reasons is that there is no answer to the question of when a reason is rational, and when an attitude is correct or not. But there is an exception to this. While Hume—on the classic reading, at least—says that one can rightly desire anything at all, and that it is not the case that some desires are rational and others are not, he nonetheless admits that belief has a direction of fit which goes from the mind to the world, and that a belief is correct if and only if it is true, and incorrect if it is false. He even gives the name “reason” exclusively to our mind’s cognitive faculty. This is the sense in which he says that “reason is the slave of the passions”: our beliefs, as far as they are true or false, are at the service of our desires.

Parfit on state-given and object-given reasons

8An answer to the Humean challenge must be able to tell us what the criterion is that distinguishes good from bad reasons. A simple distinction can set us on the right track. When we believe something for a non-epistemic reason—for a pragmatic reason, in pursuit of a certain good or in order to avoid something bad—our reason is not a reason to believe this thing, but a reason to want to believe this thing. In the first case, our attitude bears on a particular content of a belief (a proposal or an intentional content), while in the second our attitude bears on our attitude, or more precisely on the attitude’s desirability. Piller talks about “content-related reasons” as opposed to “attitude-related reasons.” [15] Parfit talks about object-given and state-given reasons. “Bad” reasons all seem to be of the second type, and “good” reasons are all of the first. In the case of desires, the distinction is between wanting something for itself, because that thing is intrinsically good, and wanting to have the attitude because doing so would have good consequences. [16] If the demon threatens me with torture unless I want to drink a saucer of mud, my reason for obeying him will be my desire to want to drink the saucer of mud, and not my desire to drink this saucer of mud. The first attitude is a first-order one, bearing directly on the content, and the second is second-order: it is an attitude about an attitude.

9Consider again the case of the demon who threatens us with torture unless we believe that 2 + 2 = 5. He has a powerful brain scanner which tells him what it is I believe or want. I have no object-given reason to believe that 2 + 2 = 5. I have a reason to want to believe it, because I do not want to be tortured. But I have no object-given reason to believe it, even if I want to, because under normal conditions we cannot just believe what we want. I can acquire such a reason through indirect causal means—by brainwashing, for instance, or taking some drug. But even if I thereby come to believe that 2 + 2 = 5, it will never be anything other than an object-given reason produced or acquired on the basis of a state-given one. And this indirect method is risky, because even if I have the belief in question, I could very well acquire evidence that 2 + 2 does not equal 5, and so abandon the indirectly acquired belief. I can also use other techniques, like forgetting the counter-evidence I have acquired. [17] We could also consider making ourselves irrational, and then forgetting that we are: the beliefs one would thereby acquire would be object-given. [18] But even if we successfully acquired these beliefs by these indirect methods, our reasons would always be reasons to want to have these beliefs, or to cause oneself to have them, and not reasons to have them. Typically, object-given reasons involve a relationship between a state and a content, while state-given reasons involve a relationship between one state and another, that is, a second-degree attitude. These second-degree attitudes can themselves be caused by agents who can manipulate their attitudes voluntarily, whereas object-given reasons seem essentially non-voluntary and outside the agents’ control. On this criterion, good reasons are those that cannot be manipulated. But the cases of belief manipulation just described assume that, beginning from a state-given reason, our beliefs can become object-given reasons once manipulated. Can these “ill-gotten” objects, so to speak, count as acceptable reasons in the eyes of those who defend object-given reasons as the only good kind of reason?

10The difference between first-order (object) and second-order (state) reasons enables us to distinguish the two types of reasons, but can it provide a criterion for what kind of reason is the right one and what kind of reason is the wrong one? For the defender of state-given reasons, this difference amounts to begging the question: they will insist that it does not tell us why a causally obtained reason is not a reason. One answer might be that, when a belief is caused intentionally, what we are actually dealing with is a cause rather than a reason. After all, reasons for wanting to believe are practical reasons, just like reasons for acting—for instance, reasons for wanting to go for a walk. And no one denies that reasons for acting are bona fide reasons. Defenders of object-given reasons will respond that this just begs the question again, because practical reasons to believe are not epistemic reasons based on evidence and data. To which the defender of state-given reasons could respond: but why should practical reasons to believe not be reasons? After all, the practical reasons that Pascal gives the libertine certainly seem to be reasons. [19] We are going around in a circle. We get no further if we say that the difference between good and bad reasons is down to the difference between normative and motivating ones. [20] A normative reason to believe that P is that P is supported by the evidence, which is our reason to believe that P is true, whereas a motivating reason is the cause of our attitude of belief. When the demon threatens me, I have no normative reason to believe that P, but I have a motivating reason to. Good reasons are those that can be normative, not just motivating. The problem is that this criterion reproduces the idea that the good reasons for belief are epistemic or associated with evidence, whereas bad reasons are pragmatic. We still have no decisive criterion for making the distinction.

11Parfit proposes another criterion for distinguishing the two types of reason: [21] looking at when reasons of each type compete (when we cannot satisfy both types of reason at the same time) and when they conflict (when each kind of reason demands a different answer). But in the demon case, the reasons do not compete: I have excellent (object-given) reasons to believe that 2 + 2 do not equal 5, even while I have excellent (state-given) reasons to want to believe it (I do not want to be tortured). And if I assume I have state-given reasons to believe that 2 + 2 = 5 (because I would be tortured if I did not), these reasons cannot conflict with my object-given (epistemic) reason for believing that 2 + 2 = 4. The defender of state-given reasons may grant this conclusion but refuse to deduce from it that there are no state-given reasons: instead, they may think that the latter are practical reasons, whereas object-given reasons are epistemic. And who would deny we have these two types of reason? But this does not solve the problem of whether we can compare or imagine a conflict between these two types of reason. To do so, we must be able to weigh or evaluate the force of the respective reasons for the same belief content. [22] But no comparison of this sort is possible: how could we weigh our evidence that 2 + 2 = 5 (which is non-existent) against the benefits (which are vast) of believing that 2 + 2 = 5?

12Parfit argues that the defender of state-given reasons may claim that there can be object-given reasons for our desires. [23] This is basically what the scholastic maxim Nihil appetimus says when given an objective reading: our reasons for wanting something are due to the fact that the thing in question is good. And if we want some particular object of our desire, the very fact we want this good will be a state-given reason. Imagine a variation on the demon scenario: the demon will torture me if I do not want him to kill me; so I ask him to kill me and, because he can detect all my desires, he frees me. In this case, the defender of state-given reasons would say that we have an object-given reason to have a desire, which gives us a state-given reason to push ourselves to have this desire. But the premise is false: I have object-given reasons for the demon not to kill me, which are also reasons not to want him to kill me.

13Can we have object-given reasons to have desires which are also state-given reasons? For example, I want to have a good night’s sleep so that I am ready for an important interview tomorrow. But I have a sort of insomnia which means that every time I particularly want to sleep, this provokes my insomnia. Here, the object-given reason and the state-given reason come into conflict. Such cases also exist for beliefs. Some students take an exam which requires self-confidence to do well in, but they discover that the examiners want to teach them humility, and will give everyone with self-confidence a bad grade (they have a brain scanner). [24] In this case, it seems that we can compare and weigh our object-given reasons and our state-given reasons. But once again, the description of the cases is misleading. I have no object-given reason not to want to sleep. I have object-given reasons to want to not have this desire, and to cause myself to lose this desire. Unlike state-given reasons for not having this desire, these object-given reasons do not conflict with my object-given reason for having this desire. [25]

14Parfit concludes that there is no sense in which object-given and state-given reasons can be compared or put in conflict, and therefore that calling the latter “reasons” is simply a misuse of language. Only object-given reasons are reasons. He therefore settles on skepticism about the problem of the right kind of reason. But he does not give any criterion for distinguishing between them. The only one he offers is that we respond to object-given reasons, whereas we do not respond to state-given ones. [26]

A pragmatist pseudo-solution

15The skeptic about the problem of good reasons apparently relies on the idea of a difference of nature between epistemic and practical reasons. But how valuable is this distinction? A pragmatist will see a far more porous border between the two—as will a Humean. Schroeder offers a set of intuitive criteria to differentiate epistemic reasons from pragmatic or prudential ones, and argues that they are not strictly discriminating at all. [27]

1. Directness: it seems easier to believe on the basis of data or evidence than for prudential reasons. The former are direct and immediate, whereas the latter are indirect and mediate.
2. Rationality: the rationality of epistemic reasons does not seem to be the same as the rationality of practical reasons. The fact that we have a practical reason to believe does not seem to make us more rational.
3. Correctness: epistemic reasons seem to define the correctness of a belief, while pragmatic reasons seem oblique.
4. Phenomenology: Pragmatic reasons for believing are closer to reasons for conative attitudes, such as intention or hope, than epistemic reasons.

16According to Schroeder, these criteria do not allow us to distinguish between attitudes about content and attitudes about attitudes. He argues that there may be reasons related to attitudes that are nevertheless reasons of the right sort. He gives two examples, that of intention and that of belief. When we consider the choice we have between two possible decisions, when we do not yet have all the elements to choose between them—for example, in deciding whether to take a trip that depends on our knowledge of what the weather will be like—we pragmatically evaluate the options, not by evaluating the object, but the attitude or state, that is, the intention or the decision. Our reasons for deciding to form one intention or another are of the right kind—they are related to the good we expect to result, but they are nonetheless about the attitude. Similarly there may be reasons not to believe a proposition if there is not enough evidence that it is true. We can postpone a decision to believe we have a serious illness as long as we do not yet have some particular test results. In these cases of suspension of judgment, our attitude is directed towards an attitude—the judgment—and it is perfectly “good,” that is, evaluated according to the evidence or data available.

17Schroeder defends far more than the distinction between state-given and object-given reasons. He argues in favor of a radical pragmatist position: ultimately, all reasons are related to attitudes or states, and are directed to the cost or benefit of having a particular attitude. Reasons related to contents are in fact a subclass of the reasons related to attitudes, if we think the purpose of our reasons is always to obtain the attitude which fits our goals. Schroeder argues that the aim of practical deliberation involving desires, beliefs, and intentions is to coordinate and control our actions by determining what decisions to make. By definition, then, such deliberation is about attitudes or states, and the reasons for our intentions and decisions are in this sense always of the right kind. In the same way, on Peirce’s account, the role of belief is to put an end to uncertainty and so to fix our opinions. The fact that our reasoning is beneficial when we rely on the truth explains why we consider evidence as “good reason” to believe. But it also explains why we suspend judgment when we do not have enough evidence—an attitude related, not to the content, but to the attitude.

18Schroeder’s objection is weak. He tells us that when we put off a decision in the absence of data, or suspend a belief in the absence of evidence, our reasons are state-given and not object-given, and are oriented in both cases towards the benefit we expect from possessing the attitude. But that is wrong. The distinction is not between reasons not to have an intention or not to have a belief, but between reasons for having an intention or reasons to suspend a belief. But the latter are perfectly epistemic—they involve the evidence or data at our disposal and the appropriateness of the intention to take the decision, and are directed towards the object of the attitude. To suspend our judgment is indeed to hesitate between judging that P and judging that not-P, but this is hesitation for an epistemic reason. Even when one suspends one’s judgment pending further information, the suspension is motivated by epistemic reasons. The fact that there may be state-given reasons not to believe something in no way invalidates the distinction between epistemic and pragmatic reasons to believe. A patient who waits to learn more about her test results before judging whether she has an illness suspends her judgment because she does not have enough evidence, not for reasons which are pragmatic or related to the attitude she wants to have.

19This allows us to answer the question of the commensurability of the two types of reasons. There are often practical elements in our theoretical reasoning (for example, we suspend our investigation to verify a certain hypothesis, which is an action; we accept propositions for the sake of our reasoning without believing them to be true) and theoretical elements in our practical reasoning (we need to know and evaluate probabilities in order to act). But this combination of practical and theoretical reasons does not mean they are commensurable. By adopting the pragmatist conception of inquiry as the fixation of belief (to use Peirce’s famous phrase), Schroeder thinks he has shown that inquiry is always guided by state-given reasons. But he is mistaken: it is the truth that is the subject of investigation. Unless the pragmatist means to assimilate truth and utility—something no serious pragmatist can support—they must recognize that good reasons to believe remain epistemic.

Conclusion: Proper attitude required

20To a large extent, skepticism about the problem of good reasons has to do with the possibility of giving a satisfactory analysis of the idea of value in terms of valuation attitudes and reasons we may have for adopting a (cognitive, practical, emotional) attitude towards an object—in other words, for formulating a judgment to this effect. It breaks with neo-Brentanian analyses of value in terms of reasons. The difficulties that Humeans raise are based on a generalized scepticism about the notion of a practical reason. A response to this skepticism cannot rely on affirming a Platonic conception of value. Instead, it must be based on a justification of the notion of good reasons for an attitude, connected to primitive facts about the nature of the attitude in question. If we can determine what makes the attitude correct, we will have justified the analysis in terms of reasons.

21The only way out of this circle is to return to the neo-Brentanian conception: the nature of value depends on the correctness of attitudes, and the correctness of attitudes determines the nature of the reasons we have for these attitudes and their fit with their object. It follows from this way of thinking that reasons are the unanalyzable first terms on the basis of which all normative notions are to be analyzed. This is why the associated program is known as the reasons first program. The problem of the wrong kind of reason is a problem for this program only if one assumes that the notion of reason still has to be analyzed. The objection can be rejected if we accept that bad reasons simply are not reasons at all. They are no more reasons than toy soldiers are soldiers. This is why I fundamentally agree with Parfit and Skorupski’s skepticism about the problem of kinds of reasons.

22We can defend a further thesis: the reasons for an attitude are grounded in the conditions for attitude correctness. In the case of belief, the correctness condition is simply that a belief is correct if it is true. This is the norm of belief, and this norm is primitive. [28] It grounds all epistemic norms and all associated normative notions: reason, reasoning, rationality, and epistemic value. To have a good reason to believe is to have a reason to comply with the correctness condition for belief. To reason correctly in the theoretical domain is to reason in compliance with this correctness condition. To be rational is to conform to this condition, just as it is to be sensitive to epistemic value. It is an attitude’s correctness that provides the constitutive norm for each type of associated judgment and reason for action: in the practical domain this is desire and intention, in the aesthetic domain it is the judgment of taste, and so on for a range of different emotions and feelings like regret, joy, and sadness. [29]

23We thereby come back—by a rather circuitous route—to the old Kantian idea that each of the mind’s faculties (of knowing, desiring, and judging) has a proper use that determines the reasons appropriate to each. This proper use is what I have here called correctness conditions, and what Kant calls the higher use of each faculty can serve as the ideal that each attitude conforms to.