Propositional Attitudes as Commitments: Unleashing Some Constraints

Abstract

In a series of articles, Asbjørn Steglich-Petersen and Nick Zangwill argue that, since propositional attitude (PA) ascription judgements do not behave like normative judgements in being subject to a priori normative supervenience and the Because Constraint, PAs cannot be constitutively normative.¹ I argue that, for a specific version of normativism, according to which PAs are normative commitments, these arguments fail. To this end, I argue that commitments and obligations should be distinguished. Then, I show that the intuitions allegedly governing all normative judgements do not even purport to hold for commitment-attributing judgements.

Résumé

Dans une série d'articles, Asbjørn Steglich-Petersen et Nick Zangwill font valoir que, puisque les jugements d'attribution d'attitude propositionnelle (AP) ne se comportent pas comme des jugements normatifs en étant soumis à la survenance normative a priori et à la contrainte du Parce que, les AP ne peuvent être constitutivement normatives. Je soutiens que, pour une version spécifique du normativisme, selon laquelle les AP sont des engagements normatifs, ces arguments échouent. À cette fin, je soutiens d'abord que les engagements et les obligations devraient être séparés. Ensuite, je démontre que les intuitions qui régiraient prétendument tous les jugements normatifs ne prétendent même pas s'appliquer aux jugements attributifs d'un engagement.

1. Introduction

Several philosophers have claimed that propositional attitudes (PAs), like beliefs and intentions, are essentially or constitutively normative.² By this, they generally mean that some norms, specifically norms of rationality, should be essentially involved in characterizing the nature of these mental states. For example, it is in the nature of belief that believing P is correct iff P is true, or that somebody believing P should believe its implication if she happens to give any verdict on this, and it is in the nature of intention that intending to P essentially requires intending the necessary means to P. Let's call this view ‘PA normativism.’ Irrespective of how this idea may be cashed out, if PAs are essentially or constitutively normative, and it is reflected in their concepts that they are so, then PA-ascription judgements (PAAJs), like ‘he believes that P,’ or ‘I intend to P’ would be normative judgements. As Allan Gibbard puts it: “when I attribute mental content — when I say, for instance, that Ebenezer is thinking that he has lost his keys — I'm somehow speaking oughts.”³ Starting from this observation, Asbjørn Steglich-Petersen and Nick Zangwill show that PAAJs are not subject to some robust intuitions governing normative thought in general.⁴ Based on this, they conclude that PAAJs cannot be normative and, therefore, PAs are not constitutively normative. This is, as far as it goes, a good strategy to determine whether a group of judgements, not obviously normative, are ultimately normative or not. Suppose that something new is discovered. In order to determine whether it is an instance of an entity already known or not, the best way is to check whether the characteristic features of the entity in question apply to this new thing or not. One of the advantages of this argument is that it does not presuppose any specific conception of norms in force for PAs. Whether such norms are conceived of as deontic or evaluative, first-personal or third-personal, as long as PAAJs are normative judgements, this argument works well.⁵

The widely held intuitions to which Steglich-Petersen and Zangwill appeal are a priori normative supervenience and what Zangwill has called “the Because Constraint.”⁶ These are well-respected norms governing normative thought, which relate the normative judgements to the non-normative ones. From the observation that PAAJs seem not to be governed by these intuitions, it is concluded that PAs cannot be normative states; they are either non-normative states with no necessary relation to normative properties,⁷ or non-normative states that essentially consequentially relate to normative requirements.⁸ To face this line of thought, PA normativists can either argue that PAAJs are, after all, subject to such intuitions, properly understood,⁹ or they may show that these intuitions do not hold for all normative judgements.¹⁰ In this article, I grant that these principles do not apply to PAAJs, so the former proposal will not be considered here.¹¹ For the latter strategy to be successful, of course, we cannot directly claim that PAAJs are exceptions or counterexamples to these intuitions, because this would clearly be question-begging. A more promising approach is to show that some uncontroversially normative judgements are not subject to these intuitions and then to show that there is a plausible account of PA normativism that precisely construes PAAJs in terms of this kind of normative judgement. It is this strategy which is pursued here. In particular, it will be argued that thinking of PAs as a special kind of commitment, though clearly a version of normativism, has the advantage of matching our intuitions regarding the PAAJs, as well as their divergences from other normative judgements regarding these constraints.¹² The argument of this article, therefore, also lends conditional support to the attitude-as-commitment model over its rivals; given the considerations that Zangwill and Steglich-Petersen put forward, if PAs are constitutively normative, their normativity should best be cashed out in terms of the normative category of commitment.

The article proceeds as follows: Section 2 discusses the intuitions allegedly governing all normative judgements and shows how PAAJs diverge from them. Section 3 deals with the attitude-as-commitment model. The distinctness of this view from other versions of normativism is not sufficiently appreciated in the literature: this is because commitment is usually treated as just a different way of speaking about obligations and, as such, its distinguishing features are levelled away. Section 4 addresses this issue and shows that the logic of commitment is interestingly different from the logic of obligations. Having done so, Section 5 shows how and why the intuitions discussed in Section 2 do not govern commitment-attributing judgements. This article finishes with a brief conclusion.

2. Divide and Conquer

What are the a priori intuitions governing normative judgements? The first one is a priori normative supervenience. According to this, necessarily two objects/properties/persons that are identical in their non-normative respects, are also identical in their normative respects. A priori supervenience is, therefore, a coherence constraint. If, for example, this action is damn good, and there is another action that is exactly alike it in all non-normative respects, it would be incoherent to judge that the second action is not also damn good.¹³ Note that this is different from any a posteriori supervenience that one might hold about normativity or other subject matters. It is a special feature of normative thought that there is a robust intuition that the supervenience of the normative on the non-normative is conceded a priori. Having this in mind, in the rest of the article, I will use ‘supervenience’ and ‘a priori normative supervenience’ interchangeably.

I follow Zangwill in calling the next constraint “the Because Constraint.” According to this, if a subject makes a normative judgement like ‘this car is damn good,’ or ‘he ought to do P,’ she should be ready to back up this judgement by a non-normative one, at least if challenged to do so.¹⁴ The non-normative judgement invoked should show the non-normative property or states of affairs in virtue or because of which the normative judgement holds true: thus, the Because Constraint. This dependence of normative judgements on non-normative ones seems to hold a priori. Namely, somebody who makes a normative judgement but is not ready to ground this judgement by a non-normative one seems to be confused about the meaning of the normative concept she is applying.¹⁵ Suppose somebody says ‘this is terrible,’ but when asked why she thinks so, she replies ‘because this is just terrible.’ In this case, our reaction would be one of surprise or confusion. There ought to be something non-normative making this thing terrible, and the subject making this judgement ought to appreciate this point by being ready to mention this ‘something’ through a non-normative judgement.

This is unlike empirical judgements, which are not usually susceptible to this constraint a priori. In other words, even though we usually ground our empirical judgements, it does not hold a priori that upon making an empirical judgement we should be ready to ground it through another judgement on pain of being confused about the meaning of the non-normative concepts we are applying.¹⁶ Moreover, the kind of grounding relation we have in the Because Constraint is different from other grounding relations. As Steglich-Petersen stresses, a trace in the sand can ground the judgement that somebody has walked here.¹⁷ But it doesn't make any sense to hold that ‘somebody has walked here’ holds in virtue of the fact that there is a trace in the sand. Therefore, the metaphysical grounding relation that the Because Constraint requires should not be conflated with the evidential support and epistemic grounding that all judgements, normative or otherwise, must have.¹⁸ So, the Because Constraint is special both in its a priority and in the specific grounding relation that it claims between the normative and non-normative facts.¹⁹

Now suppose that PAs are constitutively normative and that this is reflected in their concepts, namely competency in applying PA-concepts presupposes respecting their normativity. Therefore, PAAJs are normative judgements and they should be subject to the above-mentioned constraints. Unfortunately, none of these is correct for PAAJs. Regarding the Because Constraint, as Steglich-Petersen writes:

It is not an a priori requirement that whenever S attributes some attitude, S must be prepared to provide a non-normative judgement in support of it, which satisfies the conditions for non-normative grounding (evidential grounding which does not satisfy those conditions is sufficient).²⁰

Unlike normative judgements, like ‘this action is despicable,’ PAAJs do not require a non-normative grounding and are usually self-standing in this respect. One feels no conceptual pressure to ground the judgement ‘I believe that P’ through a non-normative judgement showing the properties in virtue of which this judgement holds true. PAAJs either require no ground or, when they do, they can be sufficiently grounded through other mental ascriptions. For example, ‘I believe that P’ can be sufficiently grounded through my belief that P^Q. But if PAAJs were normative, they could not play the role of a sufficient ground for other PAs. But since they, in fact, do play such a role, they are not normative judgements.²¹ As already pointed out, this does not mean that there can be no evidential support for these judgements; things like behavioural clues and assertions (directly or indirectly expressing these attitudes) can certainly be evidence and justification for these ascriptions. But we should distinguish between these evidential supports, present for all judgements, normative or otherwise, and the Because Constraint.²² It is only on the most implausible behaviourist account of the mind that these behavioural clues are considered as the grounds of PAAJs, in the sense required by the Because Constraint.²³

The unruliness of PAAJs with respect to this constraint can be discovered from the opposite direction as well. If PAAJs were normative, then

they could not act as non-normative support for normative rationality judgements in a way that satisfies the Because Constraint. But propositional attitude ascriptions can act as such support. For example, when judging it irrational for S to believe that p, it clearly suffices as non-normative support to judge that S believes some other propositions inconsistent with p.²⁴

Some clarifications are needed so as to make this objection non-question-begging.²⁵ There is no denying that normative judgements can ground other normative judgements. As Daniel Laurier points out, the judgement that ‘I ought to open the window’ can be grounded through the judgement that ‘I ought to let some fresh air in.’²⁶ But the grounding normative judgement cannot act as the sufficient ground. For one can again apply the Because Constraint and challenge this new judgement: why should you let some fresh air in? Because, say, I can hardly breathe in here. The chain of the grounding judgements for normative judgements can include other normative judgements. However, ultimately, a non-normative ground is what ends this chain. And the Because Constraint does not require anything more than that.²⁷ What Steglich-Petersen points out is that, at least sometimes, PAAJs clearly provide a sufficient and self-standing ground for some obviously normative judgements, which are judgements of rationality.²⁸ Therefore, given the Because Constraint, the PAAJs cannot be normative.

Analogously, Zangwill points to this in terms of what he calls “the problem of explanatory priority.”²⁹ According to this problem, we tend to explain the rational norms in force for a subject, i.e., what she ought to infer and do, in terms of her PAs, not the other way around. In other words, PAs explain the rational requirements (permissions) governing our thinking and action and it does not make any sense to claim that we have a specific PA because we ought to think and act in a specific way. Rather, we ought to think and act in a specific way because we have a specific PA. Therefore, PAs should not be regarded as constituted by these norms since these norms are themselves explained by PAs. Added to the remark that these explanations are sufficient and given the Because Constraint, this shows that PAs should be conceived of as non-normative states that ground the rational requirements.³⁰

The next divergence between normative judgements and PAAJs concerns a priori normative supervenience. Although the intuition that the normative supervenes on the non-normative is pretty universal, it is not even close to that in the case of PAAJs.³¹ As Steglich-Petersen puts it:

[W]hereas the supervenience of the mental on the non-mental is known only a posteriori, the supervenience of the normative on the non-normative is analytic and known a priori. It is a conceptual truth about normative concepts that they are applied on the basis of non-normative properties. But it is not a conceptual truth that the mental supervenes on the non-mental, and one is not conceptually required to apply mental concepts only on the basis of non-mental properties …. In other words, even if propositional attitudes in fact supervene on non-mental, non-normative properties, this is not an a priori truth that one would display conceptual incompetence in denying.³²

The fact that PAs supervene on natural facts (particularly brain states) is almost an orthodoxy in the contemporary philosophy of mind. But few, if any, would claim that there is an a priori intuition behind that. No incoherence follows if I do not have a specific belief or intention which I do in another situation identical in all non-normative respects. Millions of people, Zangwill says, believe that mind can survive the death of body or that mind supervenes on nothing other than itself. It is implausible to hold that all of these people are denying an a priori intuition about the mind, as they would be in rejecting the normative supervenience.³³ This asymmetry, according to this line of thought, betrays the non-normativity of PAAJs.

Given all these, should we conclude, along with Zangwill and Steglich-Petersen, that PAAJs are non-normative? In the remaining sections, it is argued that it is too soon to succumb to these arguments.

3. Attitudes as Commitments

Conceiving propositional attitudes, particularly beliefs and intentions, as special kinds of commitments is a pretty recent idea in the literature put forward by a group of normativists. Although Wilfrid Sellars suggested this conception,³⁴ it was Robert Brandom who first worked out this idea in great detail in his magnum opus, Making It Explicit.³⁵ Quite independently and with some differences, this idea has been defended by Richard Moran,³⁶ Isaac Levi,³⁷ Alan Millar,³⁸ Akeel Bilgrami,³⁹ Thomas Baldwin,⁴⁰ Annalisa Coliva,⁴¹ and Nicholas Tebben.⁴² Fortunately, the argument of this article is not affected by the differences between these characterizations and works for all of them. However, for the sake of brevity, I focus on Brandom's characterization of this thesis and just point to others when it helps to clarify the issues.⁴³

Commitment is in no way an unfamiliar category; in our ordinary lives, we are all familiar with such a notion. Promising is the paradigmatic case of undertaking a commitment towards somebody to fulfill something.⁴⁴ Signing a contract is another example of undertaking a commitment. By undertaking a commitment, a subject incurs a responsibility to live up to the commitment and authorizes others to count on this. Therefore, commitment is clearly a normative concept, in the sense that there is no way to specify commitments non-normatively.⁴⁵ Brandom suggests that we replace beliefs and intentions with doxastic and practical commitments respectively, which he collectively calls “discursive commitments.”⁴⁶ For him, commitments can do all the explanatory work expected from intentional states.⁴⁷

It should be noted that different kinds of commitments may vary in what they make the subject committed to. In the case of promising, it is doing what one has promised that is the object of the commitment. What is it that the subject is committed to in having PAs? To start with, PAs are “commitments to think various things and to do various things.”⁴⁸ Discursive commitments, for Brandom, are inferentially articulated commitments, i.e., commitments to have other commitments determined by the inferential status of the object of the commitment. Entertaining a belief that P is undertaking a commitment to the truth of P. By undertaking this commitment, a subject incurs the responsibility to defend P, when challenged, by providing reasons, which are other commitments,⁴⁹ and to be committed to the implications of this commitment. For example, by being committed to ‘it is red,’ a subject becomes committed to ‘it is coloured’ (this is what Brandom calls a “commitment-preserving inferential relation”⁵⁰), and the subject should be ready to defend this commitment, for example, by invoking her reliability in discerning the colours. Similarly, intentions are modelled as practical commitments to make something happen. Undertaking practical commitments is also inferentially articulated in terms of means-end relations (‘I intend to go to the shop, and getting on the bus is the only way available, so I have to get on the bus’), and incompatibility relations (‘I should not be committed to read this book now, if I don't know how to read’).⁵¹ However, compared with promises, intentions are looser commitments in the sense that, unlike the case of promising, the subject can easily abandon them usually without any criticism. But as long as she holds the intention, she ought to respect these obligations. As Millar puts it, “I can discharge this commitment in one of two ways: by doing the necessary, or by giving up the intention.”⁵²

In parallel with the normative category of commitment, Brandom puts forward the normative category of entitlement, by which he means “permission.”⁵³ These are two primitive deontic statuses that are irreducible yet complementary: “Doing what one is committed to do is appropriate in one sense, while doing what one is entitled to do is appropriate in another.”⁵⁴ Therefore, entitlement to one's commitments is another normative dimension that opens up when we construe PAs as commitments. As mentioned above, vindicating one's entitlement to beliefs (doxastic commitments), when properly challenged, is one of the responsibilities that a subject incurs when undertaking a doxastic commitment. Further, by undertaking a commitment, a subject becomes entitled to many other commitments that may be unavailable to her beforehand. For example, by undertaking a commitment to ‘this is a dry match,’ I become entitled to the commitment ‘this will light if struck,’ even though I am not compelled to accept that (this is what Brandom calls an “entitlement-preserving inferential relation”⁵⁵). Given all these, Brandom defines two incompatible commitments as those in which commitment to one precludes entitlement to the other.⁵⁶ ‘It is red’ and ‘it is colourless’ are incompatible, since by undertaking the former a subject loses her entitlement to the latter. Of course, one can have commitments to which one is not entitled: the dimension of entitlement disallows this, but does not make it impossible.⁵⁷

Commitments for Brandom cannot be separated from the practical attitudes of attributing and undertaking them. In fact, Brandom thinks that commitments are the “creatures” of the social practice of attributing/undertaking them.⁵⁸ Even if we deny this strong claim, it is still plausible that commitments are always accompanied by these practical attitudes. There are no free-floating commitments; they should be either undertaken or attributed to someone.

What makes the commitment-model of beliefs unappealing to many is that it seems to contradict the widely held view known as ‘doxastic involuntarism’: according to this view, beliefs are involuntary in the sense that the subject does not decide to have them. In this sense, beliefs are obviously not like promises, intentions, and even assertions whose forming is clearly voluntary. And since beliefs are involuntary, the objector continues, they cannot incur responsibility; one cannot be deemed responsible for that over which one has no control. Therefore, it does not make any sense to construe beliefs as commitments. This objection, however, misfires. True, doxastic commitments can be involuntary. I open my eyes and come to believe many things (to undertake doxastic commitments). Nevertheless, my responsibility now does not concern the formation of belief. Granted, I should not be held responsible for something over which I have no control, but my responsibilities here do not concern this; rather, by coming to believe something, I incur a responsibility to follow the consequences of my beliefs and to defend the belief if challenged, namely, if shown to be incompatible with my other commitments. If I cannot successfully defend the commitment, probably by abandoning the commitments incompatible with it or by showing that the challenge is not well placed, I lose my entitlement to it and I should no longer undertake this commitment. Therefore, even though — unlike the cases of promising and signing contracts — doxastic commitments are not usually voluntary, this does not mean that the responsibility dimension goes away. The involuntariness and the responsibility each concern different dimensions and, as such, do not undermine each other. In the case of promising, I am both responsible for making the promise and living up to it. But, in the case of belief, it is holding the belief (not its formation) that my responsibility consists in.⁵⁹

It may seem that this view is nothing but another way to explicate the thesis that PAs are constituted by rational requirements and obligations. If this is so, then we will end up just where we started; PAAJs sufficiently explain the rational obligations and as such they should be non-normative judgements. Moreover, unlike evaluative and deontic judgements, PAAJs do not conceptually cry out for a metaphysical ground. Unfortunately, the defenders of this view have not done enough to dispel this thought. But the relationship between commitments and obligations is more subtle. Identifying these two normative categories makes us lose sight of the specific logic of the commitments, along with their undertaking/attribution. The next section tries to bring this important point home.

4. Commitments and Obligations

What is the relationship between commitment and obligation? To begin with, being committed to something is not identical to being obliged to do something. Rather, undertaking commitments explains the obligations that a subject becomes subject to. Unfortunately, even Brandom sometimes writes in a way that urges the contrary. For him, commitments are just another way of speaking about obligations — a way that does not imply the hierarchical picture of obliging others implicit in the concept of obligation.⁶⁰ But this cannot be true; I should see my friend because I have undertaken a commitment to do so, but it is not the case that I have undertaken this commitment because I should see her. It is the same when we attribute a commitment to somebody else; that she is committed to P is not identical to the claim that she ought to P, even though this obligation conceptually stems from the commitment in question. Therefore, even though undertaking commitments essentially imposes some obligations on the subject, commitments cannot be identified with such obligations. There is a unidirectional path from commitments to the course of action that should be done as a result of undertaking them. Accordingly, if belief or intention is a specific kind of commitment, as suggested in the previous section, this does not mean that beliefs or intentions are a cluster of obligations or (rational) requirements imposed on the subject. Rather, they are commitments whose undertaking/attribution essentially enjoins some responsibility to comply with these obligations. Therefore, the rational norms essentially in force for PAs are responsibilities that a subject incurs as a result of undertaking these commitments. This point is analogous to Zangwill's explanatory priority objection to conceiving PAs as normative states, as discussed in the previous section. Commitments explain the obligations, not the other way around. Therefore, the two cannot be identical.

There is also another related reason; obligations can be defined in terms of permissions. Brandom is right that doing so presupposes the formal notion of negation, but still this move is successful.⁶¹ ‘I am obliged to do P’ can be defined as ‘I am not permitted not to do P.’ However, even with such a presupposition, commitments cannot be defined in terms of entitlements (rights, permissions). Brandom says that “It does make sense to think of being committed to do something as not being entitled not to do it.”⁶² But this cannot be right. It is true that I am not entitled not to x as a result of being committed to x; but it is clearly different from defining the commitment in terms of the entitlement. Intuitively, I may not be permitted not to x but still bear no commitment to x. But if I have undertaken a commitment (or am legitimately attributed one), I would not be entitled not to do it. This is not, however, a case of defining commitments in terms of entitlements. Rather, the consequent of this conditional holds because, as a result of undertaking a commitment, I come under an obligation to fulfill x, which now means that I would not be entitled not to do it.⁶³

Notice that undertaking a commitment is not acknowledging an obligation either. I may not accept that I ought to stop smoking. But when I accept that, this still does not amount to undertaking a commitment to do so. There is an extra ingredient to undertaking a commitment. Obligations ensuing from commitments are, so to speak, self-imposed obligations⁶⁴; to be committed to stop smoking, I should have bounded myself to do so. This is not only acknowledging that a norm is in force for me, but also creating this norm ab initio.⁶⁵ In this sense, Bilgrami seems right that there is an “agential” or “performative” aspect to commitments.⁶⁶ Commitments are specific to rational beings, who can autonomously bound themselves and incur responsibilities. Finally, it is clear that undertaking a commitment does not mean that the subject will discharge it. The only link between undertaking a commitment and discharging it is that the subject should discharge the commitment that she has undertaken on pain of being criticized for her failure. The link, therefore, is itself normative through and through.⁶⁷

I conclude that, despite their close connection, commitments should not be identified with obligations.⁶⁸ By identifying commitments with obligations, we lose sight of the distinguishing features of commitment-talk. The importance of these features will become salient in the next section when we see how commitment-attributing judgements differ from other normative judgements, including many obligation-attributing judgements, with respect to the intuitions discussed in Section 2.

5. Commitments vs. the Intuitions

Finally, let's see how this normative account of PAs confronts the intuitions discussed in Section 2. This section aims to show that, when it comes to commitments, these intuitions do not even purport to hold. Let's start with a priori normative supervenience. Interestingly, in a rather different context, Bilgrami has noticed that commitments, including PAs construed as such, are not subject to the a priori supervenience relation. As he puts it:

[T]wo worlds may be exactly alike in descriptive facts but not in performative facts …. Supervenience is not a thesis that is made for, made to fit, the performative aspect of promising. Whatever it fits, it does not fit the performative aspect. That is exactly analogous to the point about the supervenience of values or of intentionality (thought of as values or commitments) on non-evaluative facts.⁶⁹

Let's reconsider the idealized case of promising. Suppose you have undertaken a commitment to somebody. Now, there seems to be no a priori pressure to concede that necessarily in a situation exactly alike this in all non-normative respects, you would have undertaken the same commitment or any commitment at all. Similarly, suppose that you are committed to the truth of a proposition or to making something happen. Does it imply that in a non-normatively indiscernible world you necessarily would have the same commitment? It doesn't seem so. Those who do not have a clear intuition here should be careful not to let some normative properties of the first scenario enter into the supervenience base of the second one. Promising and signing documents, for example, are undertaking commitments, not non-normatively specifiable actions that enjoin commitments. ‘Signing,’ ‘contracting,’ ‘agreement,’ and ‘promising’ are all normatively charged words presupposing the practice of undertaking and attributing commitments.⁷⁰ When these actions are detached from this practice — for example, once they are considered simply as depositing some ink on paper or making some noises — this illusion disappears. If we are careful not to let these normative features enter into the supervenience base, I think one feels no a priori pressure to concede that necessarily, in a non-normatively indiscernible world, one would have the same commitment or any commitment at all, similar to what we have in the case of PAs. No incoherence follows if I have a different set of commitments in a situation exactly alike in all non-normative respects. At least when considered a priori, commitments are independent with respect to non-normative facts, and the changes thereof. True, at least when it comes to doxastic commitments, having incompatible commitments (about a non-normative state of affairs) in non-normatively indiscernible situations can challenge the entitlement of the subject to those commitments. But this is different from supervenience and also holds for beliefs; having inconsistent beliefs (about a non-normative state of affairs) in two non-normatively indiscernible situations challenges the rationality of the beliefs. But it does not mean that beliefs supervene a priori on natural facts.

How about the Because Constraint? Let's start with the explanatory priority objection that Zangwill raised against PA normativism. Recall that Zangwill contended that rational requirements are explained through the PAs, not the other way around, and, as such, PAs should be understood as non-normative properties that ground the normative rational requirements, as the Because Constraint requires.⁷¹ The commitment-model of PAs, however, shows that this argument does not work. Since, on the commitment model, which is clearly a version of normativism, PAs behave exactly similarly in this respect: I ought to infer such and such because I am committed to P and I ought to do such and such because I've promised you to do so. Therefore, the explanatory priority objection has nothing to tell against seeing PAs as commitments, even though it might be effective against other versions of normativism. This is so because, as we saw, commitments create and explain the obligations they impose. Further, commitment-attributing judgements are sufficient for grounding judgements of rationality as well; it is irrational for you to be committed to Q because you are committed to P, which is incompatible with it. Intuitively, these are sufficient explanations in the sense that nothing more is expected. Moreover, this goes beyond saying that commitments can sufficiently explain each other (as PAs can do); I am committed to P, because I am committed to P^Q.

Finally, like PAAJs, attributing/undertaking commitments does not require non-normative grounding in the sense demanded by the Because Constraint (other than innocuous evidential support like expressions and behavioural clues). You are committed to P because you have just said so, or you are behaving in a way that entitles such an attribution; these are epistemic reasons for attributing commitment, but it does not make any sense to claim that I am committed to P in virtue or because of a non-normative fact like behaving in a specific way. It is a priori that there should be something grounding why this car is damn good, like the fact that it is fast and cheap. But there is no such an a priori requirement for ‘Tom believes that P’ or ‘Tom is committed to P.’ This is so because, as we saw in the previous section, commitments are special normative statuses that autonomous beings are capable of undertaking. Undertaking commitments is to bind oneself to a specific course of action, to make oneself responsible to do something. Commitments create obligations and nothing further to the fact that a commitment is undertaken is required to explain the obligations that the subject has incurred as a result of them.

We say that I am committed to this because I have signed this contract. Isn't it a grounding relation? I don't think so. This explanation is more like I am committed to this because I have promised you to do so. But, as I discussed above, promising and signing documents are undertaking commitments, not non-normatively specifiable actions that enjoin commitments.

Even though these constraints do not govern commitments, it should not be concluded that commitments are worryingly ungrounded. Commitments can and should be challenged and reasons can always be asked for them. As already discussed, this is what the orthogonal dimension of entitlement provides for commitments. But commitment-attributing judgements do not hold in virtue or because of their entitlement. The relation between the commitments and their entitlements is not like the relation between the normative judgements and their grounding non-normative ones. I may be committed to something and, happily, I may have entitlement to it. But I also may be committed to something without being entitled to it, in which case I ought to revise my commitment. One might dig in and insist that commitments are non-normative since they are not subject to the a priori supervenience and the Because Constraint. But this move is implausible, since these constraints are themselves based on our intuitions regarding some paradigmatic normative judgements (like ‘this car is damn good’ and ‘I ought to see my dentist’) and commitment-attributing judgements, at least in the sense used here, are, intuitively, as normative as these paradigmatic normative judgements. Therefore, the non-applicability of these constraints to the commitment-attributing judgements cannot be construed as a sign of their non-normativity, but rather shows the limited scope of the constraints.

As should be clear by now, as well as the commitment-attributing judgements, which are not subject to the constraints, some normative judgements stemming from commitment-attributing judgements are not governed by these constraints either. We saw that deontic judgements stemming from commitments (e.g., you ought to see me because you have promised me to do so) can be sufficiently grounded through commitment-attributing judgements, which are normative judgements. Therefore, the Because Constraint does not apply to this group of normative judgements either. If we have axiological (evaluative) judgements that can be grounded through commitment-attributing judgements, they will also be exceptions to these constraints. Suppose that I violate a commitment that I have made to you on which you have counted wholeheartedly, say that I will go to the cinema with you. This action — namely, not going to the cinema with you — is bad because it violates the commitment that I have undertaken toward you. Various non-normative facts can contribute to the degree of badness of this action — for example, counting on my commitment, you prepared many things that will be ruined by my violating my commitment. Despite this, the fact that this action is bad requires no more grounding than that it is a breaking of my commitment. And this sufficient ground is normative through and through.

It might be tempting to think that something more is needed here. Why is this action a violation of my commitment? Because it is (or implies) my not going to the cinema. So, perhaps it is this ‘not going to the cinema’ in virtue of which this action is wrong. But I contend that we should resist such a temptation. Going/not going to the cinema specifies the condition of fulfilment/transgression of the commitment undertaken. It is clear that no commitment can be made unless its conditions of fulfilment are, at least implicitly, specified. Moreover, the conditions of fulfilment can certainly be expressed using non-normative judgements. But we should note that these non-normative judgements just show why a specific commitment is fulfilled or violated, not why the commitment in question is undertaken or attributed or why violating the commitment is wrong. Conflating these two can lead us to think that these non-normative judgements are what the Because Constraint requires. Now consider again the judgement that my action in not going to the cinema is bad. As I argued, this judgement can be sufficiently grounded through the fact that this action is against the commitment I have undertaken. I ought to do what I have committed to do and, therefore, it would be wrong not to fulfill my commitments. Granted, we can ask why this violates my commitment, and now you should invoke the non-normative judgement specifying the condition of fulfilment of the commitment undertaken.⁷² The non-transitivity from ‘this action is bad because it is against my commitment’ and ‘this action is against my commitment because it is my not going to the cinema’ to ‘this action is bad because it is my not going to the cinema’ is, I contend, a signal that the second sense of because is not what we have in the Because Constraint. This is bad only because it is a violating of my commitment. But it violates my commitment because it is (or implies) not going to the cinema.

If this is right, then commitment-attributing judgements can sometimes sufficiently back up evaluative judgements regarding actions and states of affairs as well, just as they can back up certain requirements that apply to the subject. This shows that, beside the commitment-attributing judgements, the Because Constraint should further be constrained to exclude those normative judgements (deontic or evaluative) that can be explained through them. Accordingly, these commitment-grounded judgements are not susceptible to supervenience either. For example, given an action that is bad because it violates my commitment and another action exactly alike in all non-normative respects, the second action might not be bad because there might be no commitment in play in the second case. Similarly, I ought to do something because I have undertaken a commitment to do so. Now, since commitments do not supervene a priori on non-normative properties, in a situation that is non-normatively indiscernible, it does not follow a priori that I should have undertaken the same commitment, and therefore it does not follow that I ought to do the thing in question.

I conclude that commitment-attributing judgements are as detached from these a priori constraints as PAAJs seem to be. So, it is not bad news for PA normativists, at least in the version under consideration, that these constraints do not apply to PAAJs, since they also do not apply to commitment-attributing judgements, which are normative through and through.

6. Conclusion

This article argued that, despite their close connection, it is a mistake to identify the familiar normative category of commitment with obligations and requirements. When we respect the special logic of commitments, it becomes salient that commitment-attributing judgements, though indisputably normative, are not subject to certain robust intuitions governing other normative judgements. Therefore, it is just Wittgenstein's one-sided diet that leads us to claim that these constraints hold for all normative judgements. These constraints should be constrained to exclude commitment-attributing judgements and those normative judgements stemming from them. One could yet dig in and insist that commitments are non-normative because they are not subject to these constraints, but given that these constraints are themselves based on common sense intuitions, this move is implausible. Surprisingly, we saw that there is a version of PA normativism in which PAs are construed as specific kinds of commitments. We saw that the commitment-model of PAs is prima facie faithful to our intuitions regarding PAAJs as well as their divergences from other normative judgements. In view of this, the whole project of dividing PAAJ from normative judgement through these intuitions cannot get off the ground. Although this version of PA normativism is developed for quite independent considerations from those discussed here, the fact that this corresponds to the divergences that PAAJs show with respect to other normative judgements, lends it support over its rivals. Therefore, even though what Zangwill and Steglich-Petersen provides might have force against many versions of PA normativism, it has no bite against the attitude-as-commitment model.