Normative Reasons are not Good Bases: a Reply to Gregory

Abstract

In a recent paper, Gregory defends the claim that a normative reason is a good basis for Φ-ing. He claims that a “basis” is what is commonly known as a motivating reason. By “good” Gregory means good in its attributive sense, as something which is good as a kind. In this paper I argue that it is not plausible that normative reasons are motivating reasons that are good as an instance of their kind. I argue that in order to assess this claim, pace Gregory, we need to know what it is for a motivating reason to be good as a kind. I canvas some potential answers to this question provided by Gregory, first that motivating reasons are things that play a role in a causal structure, and second that motivating reasons are what is believed to be a normative reason. I argue that since neither of these is plausibly good as a kind, we should reject RGB.

In a recent paper, Gregory defends the following analysis of a normative reason¹:

• RGB A normative reason to Φ is something that is a good basis for Φ-ing.

Where “Φ” stands for an action, a belief, or some other mental state. A “basis” is what is commonly known as a motivating reason, that I take here to mean the consideration(s) that (a) explain why A Φ-ed and (b) which explanation is one that the agent recognised at the time of responding as their reason for so responding.² By “good” Gregory means good in its attributive sense, as something which is good as a kind.

In this paper I argue that it is not plausible that normative reasons are motivating reasons that are good as an instance of their kind. I argue that in order to assess this claim, pace Gregory, we need to know what it is for a motivating reason to be good as a kind. I canvas some potential answers to this question provided by Gregory, that motivating reasons are things that play a role in a causal structure, and that motivating reasons are what is believed to be a normative reason. I argue that since neither of these answers is plausible, we should reject RGB.

The clearest prima facie problem with this view is that there is a clear gap between motivating reasons that are good qua motivating reasons and normative reasons understood in a common-sense way. This gap generates an extensional adequacy problem. Some good motivating reasons are not normative reasons (intuitively). Consider a good basis for belief. A “good basis” for a belief in the sense of motivating reason that we defined above is one which is good of its kind. To be a good motivating reason is to be a non-defective explanation of why A believes that p that he recognises. Suppose that A believes that the cat is on the mat because the moon is made of cheese. By hypothesis A’s motivating reason explains his belief that the cat is on the mat, but it clearly does not justify A’s believing that the cat is on the mat.

However, Gregory makes clear that this is not what he intends by “good motivating reason”. So what is a “good motivating reason”? Gregory claims that he can be silent about what it is to be a good motivating reason:

My claim here is not that RGB employs the phrase “good basis” to refer to some as yet unspecified property. Rather, the point is that RGB can be silent about which other properties make a difference to how good something is as a basis for doing something. Here is a comparison: if I claim that what you ought to do is just whatever you have most reason to do, that claim is not robbed of content by my allowing that it is an open question as to which things you have most reason to do. Similarly, when I claim that reasons are good bases, that claim is not robbed of content by my allowing that it is an open question as to which bases are good.³

However the comparison is not apt because, in fact, we know what it means for you to have overall to reason to do something (though we do not necessarily know the grounds), but we don’t, as yet, know what it means (for Gregory) for something to be a good motivating reason. What it means to have overall reason to do something is for there to be greater reason to do that thing, than to do otherwise. What is a reason? It is a consideration that favours a response. That definition allows us to intelligibly use the concept of an “overall reason”.

Can we use the term “good motivating reason” in the same way? As mentioned above, Gregory claims that “good” here is attributive goodness, or goodness as a thing of its kind. What is it for something to be good as a kind? According to Thomson there are features which pick out what she calls goodness-fixing kinds.⁴ First, they must be capable of being defective as a kind. Toasters are normative kinds but pebbles are not normative kinds. Second, this kind is picked out by a function or capacity of the kind. Toasters have the function of toasting toastables, while pebbles have no such function.

So, in order to know what it is for some motivating reason to be good as a motivating reason, we need to know whether it is capable of being defective, and what function or capacity is picked out by the kind “motivating reason”. In order to answer this question we need to know what a motivating reason actually is. So what does Gregory take a motivating reason to be? Gregory is open to different interpretations of a motivating reason: either (a) something that the agent believes to be a normative reason, or (b) causation of the right kind by other kinds of mental states.⁵ Let us consider these claims in turn.

Gregory claims that one possible candidate for what a motivating reason is, is something that is believed to be a normative reason.⁶ In order to assess whether some F is good as a kind it is helpful to know what, ontologically, F is. In this case, F is what is believed, so it is plausible that F is a proposition. However, it is not plausible at all that propositions are the sort of thing that can be attributively good. Thomson tells us that

Nobody, I am sure, thinks that ascribing ‘being a true proposition’ to (as it might be) the proposition that snow is white is making a favourable evaluative judgment about that proposition. The kind ‘proposition’ is plainly not a goodness-fixing kind: no proposition is good qua proposition – no proposition is a better specimen of the kind ‘proposition’ than any other.⁷

The point should be clear. Assigning attributive goodness to some F is applicable only where F is capable of being defective as an F. But there is no such thing as a “defective” proposition. So this interpretation of a motivating reason will not play the role in RGB that Gregory wants it to play.

Now consider the second interpretation of a motivating reason. On this causal interpretation, a consideration gets to be a motivating reason by playing a certain causal role. What sort of thing plays this causal role? According to the standard view in the philosophy action what should play this causal role is a belief-desire pair.⁸ Can a belief-desire pair be good as a kind?

First, let us ask whether a belief can be good as a kind and suppose that this is a particular sort of belief, namely a belief about a normative reason.⁹ Are beliefs about normative reasons a kind in the sense that tables, chairs, knifes, are kinds? I have claimed that a way to delineate kinds is to ask what their functions or capacities distinguish them from other things. As knife is for cutting, a chair is for sitting, and so on. However, again, as Thomson points out, believing does not plausibly look like it is good as a kind because believing is not the sort of thing that one does well or badly, unlike, for example, playing a Mozart sonata.¹⁰ Playing the sonata can be considered a kind because the piece can be played well or badly. But one cannot, by the same token, believe well or badly.

In reply we might suppose that it is the function of belief to aim at the truth, or to represent the world veridically. Then a purported “good” normative belief is one that has as its content the true proposition that ‘A has a normative reason to Φ’. Suppose that I drink the gin because I believe that its being delicious is a reason to drink it. What it is to be a good normative belief is for my belief that the gin is delicious is a reason to drink it to be true.

We might think that this shows how wishful thinking is a bad basis for action. Wishful thinking does not lead to the formation of true beliefs, and thus true normative beliefs. So wishful thinking undermines the function of belief, which is to aim at the truth. Since wishful thinking is bad qua normative belief, it is by RGB a bad basis for action.

However what this seems to describe is not a belief being better or worse, but rather a belief being correct or incorrect. A belief that is correct is true and a belief that is false is incorrect. However, a belief is not better as a belief if it is true, and a belief is not worse as a belief if it is false. Correctness does not entail goodness and incorrectness does not entail badness. Again, since a belief cannot be better or worse, it is not the sort of thing that can be good as a kind. I try to substantiate this point further, drawing on work by Thomson, below.

In addition, the above analysis seems exactly backwards. We usually think that a true belief is made true by something like a fact or state of affairs. We would thus expect that a true belief about a normative reason, R, is made true by R. But if on this interpretation we claim that motivating reasons are beliefs, RGB asks us to accept the reverse of this claim. Being a normative reason, R, is just being a true belief about that normative reason, R. This reversal seems highly counterintuitive.

The second component of the belief-desire pair is the desire. Are desires the sort of thing that can be good as a kind? To answer this question we can again ask what, if any, functions or capacities are applicable to desires which delineate them from other things. Does a desire have a function or capacity, in virtue of which its performance or instantiation is better or worse? It what follows I consider two potentials functions of a desire: (i) that the function of a desire is to be satisfied, and (ii) that the function of a desire is to aim at the good.

The first interpretation of desires as bases claims that the function of a desire is to produce action that satisfies that desire. The idea here would then be that a desire, D₁, is good to the extent that it produces action which satisfies D₁; desire D₂ is bad to the extent that it fails to produce action which satisfies D₂.¹¹

First, this doesn’t fit particularly well with other examples of goodness-fixing kinds because failure of this function does not obviously demonstrate a “defective” property of a desire. Suppose that you desire to give a portion of your income to charity, but that you are unable to actually act on this desire because you are unemployed. The fact that your desire fails to produce a satisfying action does not obviously impugn that desire. Your desire is not made “worse” because it cannot now be satisfied. Or suppose that you desire to jump out of the ground-floor window because the building is on fire and this is the only exit not blocked by the fire. The fact that the window is sealed shut, meaning that you are unable to jump out does not make your desire “worse” in any tangible sense. It is still perfectly sensible to want to jump out of the building, despite the fact that you cannot.

Second, tying the evaluation of desire to desire-satisfaction, which in turn determines the existence of a normative reason via RGB forces us to accept a Humean theory of normative reasons. A Humean theory of reasons (HTR) says roughly that if R is a reason for an agent, then that agent must have a desire that is satisfied by her complying with that reason.¹² Clearly, if what it is for R to be a normative reason to Φ is that R be a desire to Φ that results in its satisfaction, it cannot be true both that A has a normative reason to Φ and that A does not have a desire which would be satisfied by his Φ-ing.

Though HTR is accepted by many philosophers, it remains a controversial thesis and is its truth is therefore a matter of substantive debate. Especially problematic for RGB, then, is that it would appear to preclude this debate on the interpretation we are considering. Since RGB is an analysis of what a normative reason is, if RGB is true, it would imply that Humeans and non-Humeans are simply talking past one another. Non-Humeans are apparently conceptually confused when they claim (impossibly) that an agent can have a reason to act even if no desire of theirs is satisfied thereby. This is a very strong claim, one that requires substantial argument from a defender of RGB.

In addition, a further problem with RGB entailing HTR is that it, again, restricts the analysis of a normative reason to only reasons for action. For instance, it not plausible to claim that a reason for belief be understood in terms of the function of a desire. Since the analysis is intended to cover reasons generally¹³ and, as we have seen, beliefs are plausibly not goodness-fixing kinds, appealing to desires undermines this part of the view.

Let me now turn to the second interpretation of desires as bases. The second interpretation of the function of desire claims that desires are good or bad depending on whether the object of that desire is good or bad. By analogy with the claim that the function of a belief is to aim at the truth, the function of a desire, it is claimed, is to aim at the good. Note that, on this view, we have introduced a further, distinct, sense of “goodness”. Desires are good in the attributive sense we have been considering when they fulfil their function or capacity. Objects of desire are more plausibly understood as good (or bad) in the predicative sense, and are thus not to be explicated with reference to a function or capacity.¹⁴

On this interpretation of RGB, for A to have a normative reason to Φ is for A to have a desire to Φ and for A’s Φ-ing to be good. I leave it an open question how exactly to understand the object of a desire’s being good, and what makes it good. I present a number of problems with this view as an interpretation of RGB, in order of seriousness.

First, this interpretation of RGB entails commitment to at least one version of what is often referred to as the “guise of the good” theory (GG).¹⁵ GG understood here is the claim that desires aim at the good.¹⁶ It is, however, controversial that an analysis of normative reasons should entail commitment to GG (even if GG is fairly plausible). There are well-known counterexamples to GG, for example in cases of perversity, accidie, depression, and so on.¹⁷ It may be that a weaker version of GG avoids some of these difficulties, but it is nevertheless a potential liability for the analysis.

Second, another controversial implication of this interpretation is that it entails Psychologism about normative reasons. Psychologism is the view that normative reasons are mental states: in this case, desires. Although Psychologism is accepted by most philosophers as a claim about motivating reasons, it is a controversial claim about normative reasons for action.¹⁸ There is a (admittedly non-universal) consensus that normative reasons for action must be facts, even if motivating reasons are mental states. Again, the fact that RGB entails this disputed claim represents a cost for the analysis.

Further, Gregory’s official position is that normative reasons are analysed in terms of attributive goodness and facts about the explanations of what we do.¹⁹ The sorts of things that he takes as bases are facts: ‘… that it is hot and dry today is not a good basis for taking my coat with me on my walk.’²⁰ Note that the view that bases (motivating reasons) are facts about belief-desire pairs is not the view that bases are belief-desire pairs, which is the view under consideration. So this adherence to Psychologism seems to be at odds with Gregory’s own position.

Third, let us consider again the difficulties that beset our trying to claim that belief is an evaluative kind. Thomson claims that if beliefs or desires are kinds of any sort, they are correctness-fixing, not goodness-fixing kinds.²¹

Similar to goodness fixing kinds, a correctness-fixing kind (C) itself sets the standards that a C has to meet if it is to be correct as a C. How does correctness differ from goodness? Thomson claims that mental states with propositional contents have correctness conditions. Some mental state is correct if, and only if, there is a condition under which that mental state is deserved.²² Those conditions will be identified by reference to the propositional content of the mental state in question. For example, as discussed above, A’s believing that p is correct if, and only if, p is true, and thus A’s belief that p is deserved. Similarly, A’s desiring that q is correct if, and only if, q is a desirable (good) object, and thus A’s desiring that q is deserved.

Thomson notes that while all goodness-fixing kinds are favourable evaluative properties, not all correctness-fixing kinds are.²³ For instance, as claimed above, while beliefs can be understood as correctness-fixing kinds, they are not favourable evaluative properties.

To see why, we need to take note of another distinction among correctness properties. Let us examine Thomson’s example of asserting as a correctness-fixing kind. Notice that the truth of the proposition ‘A asserted that p correctly’ can turn on how the agent asserted that p. For example, A may assert that Bert is Alice’s brother correctly through saying the words ‘Bert is Alice’s brother’ while B may assert that Bert is Alice’s brother incorrectly through saying the words ‘Bert are Alice’s brother’. This kind of correctness is what Thomson calls “internal” correctness. Thus, for C to be internally correct is that C be an enterprise or activity for which there is a way of carrying out that activity correctly or incorrectly.²⁴

The other way in which asserting can be correct or incorrect is due to its having propositional content. Whether the propositional content of A’s assertion is true or false is a further way in which A’s asserting can be correct or incorrect. This kind of correctness Thomson calls “external” correctness. Given this distinction we can ask whether internal or external correctness are favourable evaluative properties of some kind.

As Thomson points out, only internal correctness can be a favourable evaluative property for asserting, not external correctness.²⁵ The truth value of A’s asserting that p is clearly irrelevant to how well A asserts that p. By contrast, the truth value of A’s describing that p is relevant to how well A describes p. The difference between asserting and describing is that while both asserting and describing have objects and contents, the content of A’s asserting is its object, while the content of A’s describing is not identical to its object.

Now, as Thomson points out, neither beliefs nor desires can be internally correct since what it is for C to internally correct is that C be an enterprise or activity for which there is a way of carrying out that activity correctly or incorrectly. But beliefs and desires are mental states, not activities, so they cannot a fortiori be an activity that is carried out correctly or incorrectly. However, beliefs and (arguably) desires have propositional contents. Thus they are open only to the possibility of being externally correct.

If this is right, desires and beliefs being externally correct is not a favourable evaluative property, for the same reasons given above for asserting. Just as the object of an assertion is identical to its content, so too, the object of a belief is identical to the content of that belief, and the object of desire is identical to the content of that desire. It is not as if A’s believing that p is “better” qua believing if p is true, or that A’s desiring that q is “better” qua desiring if q is good. Again, by contrast, A’s describing that r is better qua description if it is a true description.

Finally, we can agree that being a correct desiring entails, on the view under consideration, a favourable evaluative judgement, namely that what is desired is good. However, that desiring entails a favourable evaluative judgement does not show that ‘being a correct desiring’ is itself a favourable evaluative judgement. As Thomson points out, A’s performing a Mozart sonata is not a better (or worse) performance if the sonata had not been as a good a piece of music as it in fact is. Similarly, a correct desiring is not made better qua desire if its object is good.²⁶ Further, since Thomson herself is a leading defender of a neo-Aristotelian view about goodness²⁷ this argument against this interpretation of RGB has especial force.

The last objection to this interpretation of RGB is, I think, the most serious. The objection is that if we analyse normative reasons into desires whose objects are good, we will end up with a completely implausible analysis of normative reasons for mental states.

So far I have been discussing normative reasons for action. This fits the present interpretation of RGB well since the object of a desire is typically an action. But, in addition to reasons for action, agents can also have reasons to desire, admire, intend, blame, and so on. Consider reasons for desires. According to RGB, for A to have a normative reason to Φ is for A to have a desire to Φ and that Φ is good. Thus, for A to have a reason to desire Φ is for A to have a desire to desire Φ, and that desiring Φ is good. Therefore, this interpretation of RGB entails that reasons for desire imply having second-order desires. This seems implausible.

Suppose you have a reason to want Fulham to lose against Middlesbrough because you bet against Fulham. According to our analysis, in order for this to be true you must have a desire to desire that Fulham lose. But this is clearly not so. It might in fact be the case that you have a desire not to desire that Fulham lose, because you happen to be watching the game at the Fulham end, and you don’t want your elation to show when Middlesbrough score. That is perfectly consistent with your having your first order desire, and the attendant reason to have that desire (i.e. that you will win your bet).

So this interpretation of RGB, even if it turns out to be a fine analysis of normative reasons for action, fails to accommodate reasons for mental states. Since RGB is intended to be a general analysis of normative reasons, this is a significant cost.

Propositions, beliefs, and desires therefore don’t look like the right kinds. Are there other definitions of a motivating reason that will do the job here? Perhaps motivating reasons are states of affairs that an agent takes to favour acting in some way. Can states of affairs be good as a kind? Here we can suppose that a good motivating reason is a consideration which the agent takes to favour acting in some way, which in fact favours acting in that way. But if that is the definition intended then RGB fails to look like a distinctive thesis.

Finally, suppose we reject the idea that goodness in the proposed analysis is attributive. As Gregory rightly points out, if goodness in the analysis is predicative, the analysis would not look at all plausible. That is because the consideration that is a reason for an agent to act can be a perfectly good reason without itself being a good thing. Your being in pain can be a good reason to remove your hand from the hot stove, but clearly your being in pain is not itself good. So Gregory is right to reject the predicative sense of goodness in the analysis. We should understand goodness in the analysis as attributive. But if we understand it in this way we need an answer to the question of in virtue of what functions motivating reasons are good as a kind. Since the best answers canvassed here are implausible as analyses of normative reasons, we have reason to reject the proposed analysis.