Chapter 8: Rationality

This has been a great symposium. Thanks to everyone for participating! And of course special thanks go to Michael Ridge for writing such a rich and thought-provoking book, as well as for being so attentive and responsive to his critics.

To everyone: Apologies that this post is relatively long. Please feel free to ignore any points you find confused or think miss the mark! In what follows I'll first present an overview of the chapter. Early on I'll confront an interpretive question which will give me reason to address aspects of Sections 1 and 2 together, but after that I will proceed linearly. I then conclude with some more critical points. Editorial comments appear in square brackets “[]”.

The main conclusions of the chapter seem to be the following: (1) Talk of rationality is talk of internal coherence among one's attitudes and actions. There are no intrinsically irrational ends apart from logically contradictory ones. (2) Internal coherence is not a normative notion, so judgments of rationality are not normative judgments. However, in discussion among rational agents assertions of rationality and irrationality will tend to exert a kind of motivational pressure. Thus these judgments admit of Ecumenical Cognitivist treatment. (3) What I'll call the LINK thesis, since it links rationality as a success notion to rationality as a capacity (what I will call “rational agency”). It works to explain why rationality should be thought of as a matter of internal coherence: “To say that someone is rational in the success sense just is to say that he successfully adheres to those norms to which anyone who counts as rational in the capacity sense must, for the most part, adhere” (226).

Let's begin with this last thesis. Ridge follows it immediately with a list of ten capacities which, he offers, it is platitudinous that rational agents possess. They can be roughly grouped into three categories. Items 1-4 deal with the agent's capability to genuinely set genuine ends for itself, in the sense of committing itself to an end (or a principle, if that is distinct?); 5 and 10 with the ability to act enkratically in a broad sense, on the basis of its normative judgments or its commitments to ends; 6-9 with the capacity to revise ends when they conflict or need to be specified, and to take the means to their ends because they believe them to be the means to their ends.

It does seem to me that the items on the list are platitudes, at least in the sense that every view of agency ought to find some way of vindicating them, and every sensible view I can think of does. (However, I make no claims that my summary of them preserves this property!) And in reading them one might get the impression that LINK merely asserts that these capacities suffice for the agent to be scored by the norms which form the success standards of rationality. That is, one might read the second occurrence of “adhere” as elliptical for “successfully adhere”, the “must” as normative or epistemic, and the hedge “for the most part” as indicating that we clearly do not think the standards of rationality are necessarily overriding or always meet with conformity. On this reading, calling an agent “rational” is not genuinely normative in the way that calling a toaster “a good one” is not: it is to assert that something ranks highly by a (conventionally?) associated standard without endorsing that standard at all. Or at least, without giving it much endorsement: the normative reading of “must” would indeed give assertions of rationality something of a normative presupposition, but they needn't have any immediate implications for settling the thing to do because of the hedge. On the epistemic reading LINK would presuppose only that rational agents are mostly pretty rational.

However, when we get to Section 2 (specifically 234-5) it becomes clear this is not the way to interpret LINK. The “must” is epistemic, “adhere” isn't elliptical, and “for the most part” modifies the way in which rational agents follow the norms of rationality. The idea is that in order for us to attribute those capacities to a rational agent we need to be able to interpret her as being committed to the norms. In other words, this is a thesis which attributes constitutive norms on a rather large scale. By way of illustration: What's the difference between belief that P and faith that P? It's not how responsive you are to evidence, for some people are not very good at updating on their evidence. Instead, one might think, it's whether or not you are following the rules of evidence at all. Ridge's thought is that what it is to be a rational agent is to follow (most of) the norms of rationality.

What sorts of norms are constitutive of agency, in Ridge's view? He gives nothing like a list, but he name wide-scope means-end coherence (236) and coherence of different ends with each other (237). He points out that this preserves a unity with the theoretical case, for which the paradigm instance of irrationality is inconsistency in belief.

At this point let's return to where we left off in Section 1. Ridge points out that his view can explain why a clause about normative normative judgments appears on the list of platitudes about rational agency: to judge that one ought to X is to have (Ridge says “occupy” on p. 228, and I am not sure whether this makes a difference) a normative perspective which is in part constituted by an intention (which Ridge also glosses as “commitment of the will”) to do whatever certain standards would require, as well as to believe that any such standard would require X-ing. “To fail to [X] would be to fail to take a constitutive means to one's end, and hence to be instrumentally irrational” (ibid.). Ridge later asserts that because the commitment expressed by a normative judgment is a “high-level planning state” as opposed to an immediate intention, this also suffices for an explanation of the possibility and irrationality of akrasia. When we judge we ought to X and consciously give voice to this high-level commitment, it usually generates a present-directed intention. But it doesn't always, and when this happens we exhibit the same kind of means-end incoherence.

There seem to be two other important lines of thought in this section. One is an argument for instrumentalist conceptions of rationality over more substantive, Kant-inspired ones from the possibility of rational amoralists. The twist is that Ridge argues ordinary speakers would without qualification attribute rational agency to such amoralists, so that given LINK Kantians would need to either (i) attribute persistent failure in abiding by the norms they are actually following, which seems implausible in part because we have good evidence that we shouldn't even interpret them as following those norms, or (ii) deny that these amoralists are rational in the capacity sense, biting the bullet on folk intuitions. [Here I've presented the options slightly differently than the way Ridge does. I'm a little worried that the presentation in the text relies on the premise that the amoralist is successfully exercising the capacities outlined in the ten given platitudes and therefore is rational in the capacity sense. But a Kantian may demur and look for additional platitudes.] The other emphasizes that because judgments of rationality are not normative, calling someone “rational” is compatible with the sort of substantive normative commitments which the latter seem most concerned to derive. Just as well, on Ridge's view it is perfectly compatible to be a nihilist [about the normative, presumably – someone who does not rule out any norms for acting or deliberating] while still making judgments of rationality and irrationality.

(Jumping to Section 2, p. 237...) Ridge does seem a little more moved by intuition that fully-coherent anorexics might be irrational, and to this he has two responses: (i) Insofar as it's robust it might be “an acceptable casualty of systematic theorizing”, and (ii) there is an alternative sense of “rational” on which it is genuinely prescriptive, meaning something like “self-destructive without good reason”. Ridge explains that coherence-rationality can explain why we would be driven to attribute prescription-rationality: we quite generally attribute motivations to people in virtue of which unreasonable actions like starving themselves to death would be coherence-irrational. Ridge goes on to give some reasons to think that coherence-rationality is unified and useful for making sense of agents in a way that gives it a kind of priority over prescription-rationality, and that judgments of coherence-rationality can in a sense settle the immediate thing to do in spite of not being normative.

[It might be good at this point to do some theoretical score-keeping. In brief, so far Kantians are biting the bullet on the rational knave intuition and Ridge is biting the bullet on the irrational anorexic intuition. The norms of instrumental rationality do indeed display a kind of unity that would be broken by adding in more substantive norms. But does this really support the contention that our pre-theoretic notion of rationality on balance favors the instrumental conception? Kantians might agree that the ten platitudes hang together but insist that they, as well as LINK, only say something about internally coherent agency, not rational agency. According to them rational agency is a matter of being minimally responsive to reasons or the like. After all, they might say, there do seem to be strong conceptual connections between “reason” and “rationality”, and it used to be orthodoxy that if anything was a normative notion, “rationality” was. These intuitions they can explain while Ridge can only explain them away. They'd also presumably say the same thing about theoretical rationality: Wilberforce the dinosaur-denier may be fully internally coherent, but he's hardly rational. So I'm worried that we're heading towards a stalemate – perhaps even of the sort that is characteristic of verbal disputes.]

These worries about genuinely normative uses in the penumbra of “rational” lead Ridge in Section 3 to defend an Ecumenical Cognitivism about “rationality” on which many uses of the term carry a generalized conversational normative implicature – one for which no special stage-setting is required but which can be cancelled. This is, of course, aided by the fact that we may safely assume the people we talk with are rational agents and, if LINK is right, will thereby be reliably motivated by what's rational. That pointing out such a fact implicates advice can by explained by Gricean injunctions to be helpful and cooperative. [Quick note on cancellability: “Murdering Sandeep would be rational, but I strongly advise against it” sounds worse to me than “Murdering Sandeep would be rational from your perspective, but I strongly advise against it.” If others share this intuition then it does seem awfully similar to the intuition used to argue against Ecumenical Cognitivsm about moral talk, p. 95.]

In Section 4 Ridge considers how his own objections to cognitivism about practical normativity apply here and finds them lacking. The normative implications of “rational” are cancellable, and judgments of rationality are only motivating through some independent desire to fulfill one's various ends. [Though how is this latter the case? If I'm rational in the capacity sense, won't I thereby have a basic disposition to, say, drop one of my ends once I judge that I cannot possibly fulfill all of them? And doesn't that provide the needed motivation? Indeed, it seems that on Ridge's view such a disposition cannot exist independently of judgments of rationality insofar as the ability to make such judgments (perhaps) requires rational agency.] Ridge thinks that analytic reductionism has some plausibility in this case because in contrast to morality, which surely has substantive implications for how to live one's life, “rationality” is a purely formal notion. Hence formal platitudes may capture it. [Though one hopes they're not so formal as to be analytic since, as the “schmagency” debate shows, it is notoriously difficult to see how analytic truths can set genuine norms!] And the Twin Earth objections that felled synthetic reductionism about practical normativity don't apply here: if we came across a community of people who applied a term T to actions which manifest weakness of will and often recommended them on account of their being T, it seems we wouldn't translate T by “rational”.

And that's where the chapter concludes. What follows below are some longer reflections of my own.

CONSTITUTIVE NORMS AND LINK. Surely everyone thinks that there are deep connections between agential capacities and norms of internal coherence, but there's a question as to whether LINK, or at least Ridge's understanding of it, captures them. I see three ways to raise problems for this view. (1) Advocate a general skepticism about constitutive standards. This would make LINK fail on a presupposition. (2) Accept that some mental states or attitudes have constitutive standards, but hold that the matter is always normative: to say that someone has the belief-capacity or is following the norms for belief is to say that they ought to respond to evidence in certain ways, or at least that they would ought to if other matters were not pressing, or some such. This makes attributions of rational agency normative, and it appears to give attributions of rational success a normative presupposition at the very least. It doesn't contradict LINK, however. (3) Accept that attributing rational capacities is not normative, but argue that possession of the capacities does not settle whether one is following the norms of rationality.

I'm rather friendly to the first strategy, but it's clearly too large an issue to broach here. In the next section I press an aspect of the second strategy. Here I'll try to bring out the force of the third.

Most of the capacities we unquestionably have, like my capacity to bend my right leg, do not of themselves set standards, much less constitutive standards which we are thereby interpreted as following. So I can bend my leg; does that mean I ought to bend it as much as possible? How far should it be bent, and for how long? Some people favor teleo-functional accounts of how the selection of capacities might set standards, but these all have problems too. There's a ganglion cell in the frog's retina which selectively fires for small black disks moving very rapidly in their receptive field. Does that mean it ought to respond to flies – or instead to black disks generally? Could it be that the receptive field ought to be a little bit larger? The disposition itself (much less the associated capacity) doesn't seem to settle it. (Obviously I'm taking after Wittgenstein here.)

So there's a worry that the ten platitudinous capacities on p. 227 admit of the same, non-standard-setting reading, one that doesn't lead us to interpret them as following any particular set of rules (contra p. 235). Indeed, it was that reading which led me to misinterpret LINK. So rational agents can set and abandon ends; does that mean they ought to set and abandon as many ends as possible? When should I reconsider intentions and ask, “What shall I do?” The capacities don't seem to say. So how are we supposed to get from the capacities to the norms?

INCOMPATIBLE INTENTIONS. Perhaps the problem is that only a few of the capacities are instrumental to establishing the norms which the agent is interpreted as following. Platitude (8), which deals with revising conflicting ends, is an excellent candidate. But – and this is a key claim – not all agents with conflicting intentions are irrational. I intended to be a chemist when I was 8 and now intend to be a philosopher, but that doesn't make me irrational. Similarly, if for reasons of cognitive capacity I couldn't come to realize that two of my present intentions conflict, perhaps I am not irrational. But neither is it the case that the norm only applies when I have two incompatible present intentions which I consciously know to be incompatible. For one thing it's not clear such a case is possible – perhaps I'm best interpreted as not really intending in such a case! And for another it seems we call people irrational who persist with incompatible intentions in virtue of having momentarily forgotten a previous one; they haven't yet “put the two together”. (One may prefer to put the problem as when exactly to attribute an intention to an agent in the sense relevant for the norms, but it comes to the same thing – there's a lot of indeterminacy here.)

What explains the difference between irrational and rational incompatible intentions? What is the rule that I am to be interpreted as following, and why does it take that shape? The problem is not so much the indeterminacy itself as that the agent's capacities (conceived in a naturalistic, non-normative sense, at least!) do not seem to do any work here. I have the capability to conform to lots of different norms for intentions, but which are the ones that I'm following? Neither do my dispositions settle which norms I'm following, as we saw above.

Instead, one might think, two incompatible intentions suffice for irrationality when the agent is responsible for revising them – when the agent merits blame on account of not revising them. That would plausibly make judgments of rationality normative. (Ridge might very well search for a different explanation; perhaps he can appeal to social convention. But it looks like his current position commits him to denying a popular solution to the rule-following problem.)

AKRASIA. On Ridge's view the norms of rationality are wide-scope, including the one that is supposed to explain what is irrational about akrasia – means-end coherence. But most people think there's a difference in rational authority between one's motivations here. It isn't always a rational response merely to revise one's normative judgment, as this view predicts. Also, some think there's an important difference between a commitment and an intention, which are put in apposition on p. 228: it's the difference between being for an action (e.g. being for deciding according to certain standards) and going for that action, which isn't really a normative state of mind. The problem with akrasia is that the action you know you ought to do is one that you've committed yourself to, but you don't plan or intend to do it. The attitudes aren't similar enough to be incoherent in the way that means-end incoherence is, and it requires a different explanation. Perhaps this is a sense in which the account makes for too much unity!

KANTIAN STRATEGY. (p. 231) I wonder if this section gets the dialectic right, particularly when it asserts that Kantians must show there's some incoherence in the idea of instrumental rationality. My understanding of the standard Kantian strategy is that it doesn't so much directly argue straight from the concept of rationality to certain substantive norms such that the very idea of rational action which flaunts them is incoherent. Instead there's an argument from the concept of rational agency (which may involve limited capacities of the sort advocated in the book) to a metaphysical claim about the nature of agents, and from there to a thesis about agents' non-contingent ends. This might lead to that sort of conceptual incoherence if there's nothing more to the nature of agency than being something which follows the norms of rationality, but I don't think Kantians think this. And importantly LINK leaves open that possibility, so that Kantians would not have any special problems accounting for it.

NIHILISTS. I assume nihilists (231) do not rule out any criteria for deciding. But surely they form intentions. This leaves them not ruling out not reconsidering intentions constantly. Lacking such a safeguard they easily could have adopted can lead to their intentions being considerably less successful. Isn't this a kind of practical incoherence which makes them practically irrational? So in what sense is this view of rationality compatible with nihilism if that's not to mean that nihilism is not necessarily irrational? Surely any two views are compatible if all that means is that a sufficiently irrational person could hold them! Indeed, there's a more direct route to the conclusion given that Ridge thinks (233) that following norms of rationality in the capacity sense is a way of accepting certain norms. A total nihilist rational agent would rule out no norms for acting and deciding and yet, in virtue of being a rational agent, rule out some.