Jason Bridges

University of Chicago


Phil 31410, A03—lecture 14 notes

Scanlon on reasons and rationality

(What We Owe to Each Other, chapter 1)


We won’t finish Scanlon today.  What I want to do now is talk about the positive view of agential reasons [again, this our term of art for reasons for action] one can glean from his discussion.  Interwoven with this are criticisms of the use to which the notion of desire is put in philosophy of action.  We may understand these worries as objections to weak Humeanism [the view that a reason for an action must include as a component a psychological state that is not a belief—e.g., desire, passion] as we have understood it.  We’ll begin to talk about those criticisms today, but it helps to have the positive view in place.


I. Reasons and rationality

Scanlon’s picture of agential reasons places what Smith calls the normative concept of a reason at its center.  The upshot is an account of the reasons for which people act, hence for the explanation of action, that is radically at odds with the Humean views we have been considering.

A very interesting feature of Scanlon’s discussion is that the radically anti-Humean implications sneak up on one.  They are portrayed as inevitable upshots of various considerations about the normative concept of a reason.  And a unified treatment of reasons for belief and action just emerges as well.


The moral may be that in arriving at a philosophical account of something, the starting point is all.  It will determine how one tries to accommodate the rest of the phenomena, and what one notices or fails to.


Scanlon acknowledges that the word “reason” as other uses; he mentions the one I mentioned at the beginning of the course, according to which anything that explains anything is said to be the reason for what it explains.

Scanlon’s own choice of terminology to avoid this ambiguity is “reason in the standard, normative sense”:


A reason in the standard normative sense is a consideration that counts in favor of something.


A number of initial points of clarification ought to be registered right off the bat:

1.     Considerations might be understood as propositions, as things that have a truth value.  Or they might be understood as circumstances or states of affairs—that is, things that obtain (rather than are true).  We won’t choose between these options.  Some philosophers think there is a substantial difference between them.  I doubt it, but even if there were, it isn’t one that will matter to our discussion.
The important point is that circumstances are expressed by declarative sentences.  If a business document at one point says, “The following considerations seem relevant to the decision…”, then what will likely follow is a list of sentences: “Company A is likely to go bankrupt”, “The market is especially volatile this quarter,”  etc.
It may be that certain considerations are complicated enough that they cannot be expressed in a single sentence, but only in a bunch of them, or if one insists on their being expressible in a single sentence, then in one long conjunction.  Thus: “Over the last three years, a number of companies have attempted to sell such a product.  None have met with success.  Indeed, all but one ended up being investigated by the FDA…”  Be that as it may.  The point is that they are propositions, states of affairs, circumstances, things expressed by declarative sentences.
They aren’t, by contrast physical objects.  This chair isn’t a consideration.  But that this chair is wobbly might well be a consideration relevant to something or other.

2.     Moreover, it would seem that considerations, at least those that can be reasons for things, must be true propositions.  That this chair is wobbly can’t be a reason for anything if the chair isn’t in fact wobbly.
Of course, one might think that p is a reason for something even if it’s not the case that p, because one falsely believes that p.  One might think, say, that a reason for not putting one’s drink on this chair is that it’s wobbly.  But then one is wrong.  There is no such reason for not putting one’s drink on this chair.
This point is not special to the standard, normative sense of talk of reasons.  It holds as well for the notion of reason according to which anything that explains something is a reason for that which it explains.  The reason the tree fell down can’t be that there was a heavy wind last night if it’s not true that there was heavy wind last night.
(There is a complication that arises here, which leads Dancy to disagree with this factive conception of reason, but I think that’s a mistake.  We’ll talk about the complication later on, and when we get to Dancy will consider whether his response to it is warranted.)

3.     Being a reason is a relational property.  A consideration is never a reason, period; it’s always a reason for something.  Thus the same consideration may be a reason for one thing and not for another.  This chair’s being wobbly may be a reason to throw it away and not a reason to, say, eat it.
This means that there isn’t a class of reasons, no more than there is a class of things that are, say, on the left.  You can’t just be on the left, period, you always have to be on the left of something or other.  Similarly, a reason is always a reason for.
In speaking of the class of reasons, we might mean the class of all things that are a reason for something or other.  As the class of all fathers might be understood as the class of all people that are the father of someone or other.  But such a class might potentially include all considerations.  One can imagine, for any state of affairs, a situation in which that state of affairs is a reason for something.  And so if that situation obtains it will be.

4.     The relationship at stake here is a normative one, in the sense that it is a relationship that it is one of something’s supporting, justifying, favoring, counting for something
That p is a reason for q, if it is a fact, is a normative fact.  It’s a fact that one must express using normative terminology (which might just be the word “reason” understood in its standard normative sense).
But this does not entail that the proposition that is the reason itself has normative content.  That might be so.  One might think that the fact that punching people is mean is in some sense a normative fact, and that moreover it counts in favor of refraining from punching people.  But many considerations we take to be reasons for things don’t have such a normative content.  That this chair is wobbly, if that is indeed so, is a purely physical fact about the chair.  That the chair’s being wobbly is a reason for me to throw it out is a normative state of affairs, but the normativity lies not in the chair’s being wobbly but in its standing in the reason relationship to the act in question.


The relational character of reason-hood (in the standard normative sense) raises the question what kinds of things there can be reasons for.  Clearly there can be reasons in this sense for actions: that this chair is wobbly might be a reason for throwing it out.  And there can be reasons for beliefs: that this chair is wobbly might be reason for thinking that the university’s furniture budget is under-funded.  And there can be reasons for other things: that someone is violent may be a reason for fearing contact with him.  That the world is going to pot might be a reason for being depressed.  Etc.

Scanlon proposes to unify this disparate class with the notion of a judgment-sensitive attitude.


A judgment-sensitive attitude is an attitude whose possession is sensitive to judgments about reasons for or against it.


Thus a belief is a judgment-sensitive attitude: what beliefs one has is sensitive to one’s judgments about what reasons there are for or against believing that.

Hunger, says Scanlon, is not a judgment-sensitive attitude.  One’s being hungry is not sensitive to considerations that one takes to count for or against being hungry.

[This might be wrong. It’s true that it doesn’t make sense to think of hunger (or the lack of it) as an apt or warranted response to the circumstances, in the way it does for anger and the like.  Nor does it make sense to think of it as supported or not supported by the facts.  But insofar as one could extinguish it if one chose to, and some yogis are apparently able to, we may ask what reason a person has for not choosing to do so.  We can even imagine someone who has complete control in both directions over whether he feels hunger, and then it is always a possible question what reason he has for being hungry or not.  But the point can be rescued by supposing that what there is a reason for is the intentional feat of extinguishing.]


Scanlon holds that judgment-sensitive attitudes are the only attitudes for which there can be reasons for and against.

This might seem truistic, given how the term “judgment-sensitive attitude” was defined.  But it isn’t.  One might think it possible that there could be reasons for and against attitudes our holding of which isn’t sensitive to those reasons.

Scanlon clearly thinks this doesn’t make sense.  But he doesn’t pursue the matter, and it seems a substantive and potentially interesting issue why it should be so.


Scanlon actually makes an even stronger claim:

Judgment-sensitive attitudes are the only things for which there can be reasons (in the standard, normative sense).


One worry one might have about the claim that judgment-sensitive attitudes are the only things one can have reason for is that reasons for action are left out of the picture.  Actions, after all aren’t attitudes; they are (depending on which the views we earlier discussed one adopts) things one does or the events of one’s doing those things

Scanlon’s response to this concern is to note that any reason, in the standard normative sense, for doing something is a reason for doing it intentionally.

(This reflects a point we earlier dwelled upon: although an event of doing one thing will likely be an event of doing many other things as well, only some of these will be done intentionally, and moreover it will only be some selection of those that are done for reasons.)

This point, says Scanlon, shows that actions are things for which there can be reasons only in virtue of the connections between these actions and judgment-sensitive attitudes, which shows in turn, he says, that acknowledging that there are reasons for actions does not require departure from his basic claim about what things there can be reasons for.

It’s not easy to see precisely what Scanlon has in mind here.  One possibility is that he subscribes to the view that to do something intentionally is always to do it with a certain intention.  Then the reason is for the action in virtue of being for the intention with which it is done.

There may be worries with this move.  Note, though, that holding that to do something intentionally involves doing it with an intention does not require the implausibly over-intellectualized view that every action requires the prior presence of an intention.  To take a vivid example, I might in a sudden rage punch you with the intention of doing that, even though before I punched you I had no such intention.


Another implication of the restriction is that the only things for which there can be reasons for and against belong to thinking creatures.  For only creatures that can judge can have judgment-sensitive attitudes.  This implication Scanlon thinks is wholly unproblematic.


Indeed, it is line with what Scanlon takes to be an attractive notion of what it is to be a rational being:


A rational creature is one that can have judgment-sensitive attitudes, hence the capacity to recognize, assess, and be moved by reasons.


A cat, says Scanlon, isn’t a rational creature.

This doesn’t imply we can’t ascribe beliefs and goals to cats.  The implication of Scanlon’s position is rather that cat-beliefs and cat-desires aren’t judgment-sensitive attitudes.

Thus if we observe what we take to be a change of belief in a cat—say it’s meowing at box of food and then when one opens it and shows the cat that it’s empty and it slinks ways—we can’t see that as a matter of the cat’s realizing that there’s no reason for it to want to get into the box.

(This seems especially plausible in light of the difficulty of getting a cat, at least any cat I’ve ever had, to stop believing something like that.  They keep at it and keep at it and eventually just seem to get bored or frustrated and leave.  No light ever dawns.)


Here “rationality” is a term that applies to all creatures with a certain capacity.  Creatures that lack this capacity are not in this sense rational.  But nor are they appropriately said to be irrational.  My cat isn’t irrational.  It’s non-rational.

Why?  Because irrationality is a term of disapprobation that can be applied only to creatures that are rational in these sense of having the identified capacity.  Being irrational will be some failure in the exercise of this capacity.

Now, if one is not irrational, then one may be said to be rational.  So this shows that the word “rational” does double duty.  It applies to all creatures with a certain capacity, and then again it applies to that creature, or not, on particular occasions, depending on whether the capacity is being appropriately exercised.

It would be nice to have at least in outline an account of this use of the term as well.

And this too Scanlon thinks is immediately suggested by reflecting on the notion of a reason in the standard normative sense.


A rational creature is rational (in the sense opposed to “irrational”) insofar as his or her judgment-sensitive attitudes are appropriately sensitive to her judgments about reasons for or against them.


Thus, for example, a person is irrational if she judges she has decisive reason to stop believing that p but continues to believe, or judges that she has decisive reason to wash the car but fails to form such an intention.

As Scanlon notes, the conception of rationality encapsulated by this principle is thin.  It does not require anything substantive about what one should judge one’s reasons to be.  Thus one might have highly implausible views on that score, but so long as one’s attitudes are in line with one’s judgments, there’s no failure of rationality.  There may well be a failure, but it is of a different character.

So if I judge that I have no reason not to walk rather than drive to work today, and thus form an intention to do so, I’m not in irrational by the lights of this principle—even if it’s in fact raining and I should have known at this time of year (the rainy season) to check the weather and see if it was.



II. Scanlon on reasons explanations of actions

Note what follows from the account of rationality (in the sense opposed to irrationality):


When a rational person takes it that there she has compelling reason to Ψ that p (for some judgment-sensitive attitude Ψ), she will Ψ that p, and moreover will do so precisely because she takes herself to have compelling reason to do so.


To say that one has compelling reason for something is to say that in light of all the relevant considerations one has sufficient reason to do it: there aren’t, for example, competing reasons speaking against doing it that swamp the reasons for.  One might also speak of a particular reason as compelling, by which one means that the overall circumstances are such that it amounts to a sufficient or decisive reason.

Given the connection between reasons for actions and reasons for intentions with which they are done, this suggests the follow format for reasons explanations of actions.


Scanlon’s account of reasons explanations: Reasons explanation are of the following form:
a) When S acts rationally: S φ’ed because she took consideration C to be a compelling reason for φ-ing.

This, at any rate is the form of reasons explanations when agents are rational.  When they’re rational, their judgment-sensitive attitudes, include those that issue in and (and render intentional) actions, are in line with all of their judgments about what counts for and against. 

Suppose I’m deciding whether to watch TV all day or wash the car, and decide that even though there’s reasons in favor of both, on balance I ought to wash the car: the consideration that I promised my wife I’d do it outweighs, perhaps even renders irrelevant (more on the different possibilities here later), the considerations speaking in favor of watching TV all day.


But note that there’s nothing to prevent us from gleaning from Scanlon’s discussion an account of irrational action as well.  Many of you thought that Velleman’s account yields too rarefied a notion of what a full-blooded action is.  This will be a worry if the account has the implication that one performs an action only if one does what one judges oneself to have most reason to do.  Velleman says that we have an action only when the state apt for playing the functional role of the agent plays that role in the production of the behavior in question.  That state, he suggests, is a desire to do what one has reason to do.  The question is whether, in order for that state to play its role, it has to actually win out in the battle of competing desires, such that one ends up doing the thing that would satisfy it: what one takes oneself to have most reason to do.  I agree that this is at any rate a live question in the interpretation of Velleman.

Note that the issue arises precisely because of Velleman’s reductive strategy.  He wants to make plausible the idea that a part of an agent, a state or structure within him, can play the agential role.  To make this plausible, it behooves him to not conceive of the role as just: the item that does the thing in question for a reason.  He needs to put a little more meat on the bones here.  He needs to conceive the role in such a way that it is plausible that an internal mechanism could play the part.  To put it another way, he needs to describe the role in way amenable to functionalist treatment, in terms of the performing of operations on inputs thus yielding outputs.  So he’s led to a picture of the agent’s role as akin to that played by a central processor in a computer: taking in all the relevant data and issuing in a decision.

But one way worry that this yields an over-intellectualized conception of the agent’s role, such that in many cases of doings that we’re pre-philosophically inclined to regard as exercises of agency, there’s nothing plausibly seen as playing that role, hence on Velleman’s view nothing that counts as an action.

One version of this worry is that one may think it’s setting the bar too high to require that every case of one’s acting on one’s desires be filtered through the aim of doing what one has most reason to do.


But Scanlon does not have any comparable commitments.  Thus nothing stands in the way of him regarding as actions cases in which one doesn’t do what one takes oneself to have compelling reason to do, indeed in which one does something one takes oneself to have compelling reason not do.


What Scanlon is committed to is only that even in such cases the explanation of one’s action is that one takes certain considerations to count in favor of one’s action.

Thus we may add a further clause to his account:


Scanlon’s account of reasons explanations: Reasons explanation are of the following form:
a) When S acts rationally: S φ’ed because she took consideration C to be a compelling reason for φ-ing.
b) When S acts irrationally: S φ’ed because she took consideration C to be a reason for φ-ing.


So even if I think that the weight of reasons comes down in favor of washing the car, if I nonetheless opt to sit around watching TV in light of the consideration that I’ll find it pleasant, that’s an action on my part.


Are there the materials here for a constitutive account of agency, of the sort we saw Smith and Velleman and Bratman to seek?  Perhaps.  One might take the account to be that what it is for a doing of something to be an action is for there to be a consideration C such that the agent did the thing in question because he took C to be a reason for doing that.

But this will not be a satisfying account for a naturalist like Velleman.  Naturalists want an account of action which portrays action-hood as consisting wholly in a doing’s being caused by certain psychological states.  Whether this account of reasons explanations meets this requirement depends on whether the “because” here can be understood as simply asserting a causal relationship and no more.

It would seem not, for so construed, the account would be vulnerable to just the kind of worry that prompted Velleman’s revision of the ur-Humean and Bratman-style accounts of agency: namely, that taking-something-to-be-a-reason yields an action only if it causes the behavior ‘in the right way’.  For it seems just as possible with these takings as with the states appealed to in the accounts we previously considered that they could cause behavior in some non-action-constituting way: say (to modify Davidson’s example) I take myself to have decisive reason for dropping the rope attached to the other rock climber, and this so unnerves me that my hand twitches and I do just that.


This needn’t be seen as an objection to the account we’ve extracted from Scanlon, which is just that reasons explanations are of the form specified.  It’s rather an objection to construing that account as a naturalistic constitutive account of agency.


Why do I call Scanlon’s account non-Humean?  Well first of all, notice that the only psychological states that are candidates for being one’s agential reasons are beliefs.  Thus the view denies weak Humeanism.  For the only psychological states this account requires us to credit to the agent are beliefs:

1.     Belief that C obtains or is so.
This is required if one is to take C as counting in favor of doing something.  I can’t take my having promised my wife that I’d wash the car as a reason for my doing so unless I believe that that circumstance really obtains, that I did indeed so promise.

2.     Belief that C is a reason for (i.e., counts in favor of, supports) φ-ing.
Scanlon notes that one might regard taking C to be a reason for φ-ing not as a belief, but some other sort of propositional attitude, and he mentions one alternative from Gibbard, where the attitude in question is said to be a matter of accepting norms that say that the one is a reason for the other.  Scanlon acknowledges as well that one may have metaphysical sorts of doubts about the idea that there could be beliefs of this sort, on the ground that what is thus believed to obtain is a normative state of affairs, and that it’s mysterious how there could be any such things in the world.
He spends a great deal of time in sections 11 and 12 trying to undercut these doubts.  I won’t discuss that here, not because it’s irrelevant to the issues we’ve been discussing—indeed, it bears very closely on the issues we discussed last time with respect to the debate between Velleman and Scanlon—but rather because we don’t have time.
In any case, we can say that the attitude will be belief-like.  It will be a matter of accepting a proposition, be it accepting it as so or accepting it in some other sense.
I doubt that Gibbard’s view provides an intelligible alternative, at any rate, so I’ll just assume we’re talking about a belief.

3.     Belief that the reason for φ-ing C provides is decisive, compelling, not overridden or not made irrelevant by other considerations.
This last belief will only be present in the case where the action is rational.


So it’s beliefs all the way down.

On this ground, a Humean will object to it, claiming that we need a desire in the mix as well to motivate an action.  We shall return to this issue next time.


First let me mention quickly mention a different point: that the view of agential reasons suggested here may be non-Humean in an even more fundamental way than denying a role for desires: Scanlon seems to want to deny even psychologism about reasons, that a person’s reasons for acting are psychological states of him.

That is, the view he seems to be moving toward is not that one’s beliefs about what considerations count in favor of doing something, rather than desires or non-beliefs, are the reasons for which one acts, but rather that the reasons for which one acts are the considerations one takes to count in favor of so acting.

On this view, there’s only one type of reason: a consideration that counts in favor of something.  The reasons for which a person acts, what we’ve called agential reasons, the reasons that thus explain a person’s behavior, are also the reasons that count in favor of or support one’s actions.  Scanlon speaks of a person’s “operative reasons” as the considerations that, in virtue of his taking them to count in favor of some judgment-sensitive attitude, moved him to adopt that attitude.  The idea of reason at work here is just the idea of a reason in the standard normative sense.

The unification of the notion of reason this yields may seem in and of itself attractive.  One question that seems to press immediately, though, is how we are to accommodate those cases where a person is wrong that a consideration C counts in favor of doing something—so that what he takes to be a reason for doing something is not in fact such a reason—or still worse, where a person is wrong that circumstance C even obtains—so that what he takes to be a circumstance counting in favor of doing something isn’t actually a genuine circumstance at all.

I don’t want to discuss this issue now: it will be our topic when we read Dancy.

So, again, there are 3 issues that have arisen here, all of which reflect central themes of this course:


1.     What more can be said, if anything, about the force of the “because” in a reasons explanation, given a Scanlon-style construal of reasons for action?
This issue hearkens back, among other things, to Velleman’s search for an account of the “right way” for a motivating reason to cause a doing such that it confers upon it the status of action.

2.     What more can be said about the disagreement between Scanlon’s view and weak Humeanism?
Scanlon’s input on this front will be our next topic.

3.     Does a view like Scanlon’s speak in favor of giving up even psychologism about agential reasons?
This will be the issue when we discuss Dancy, which we’ll do after finishing with Scanlon.