This section begins with a question that Gauthier poses, one that Gaus thinks we ought to take very seriously: why pay attention to morality’s demands when those demands can be so restrictive? Where does morality get the authority to constrain the pursuit of our objectives? The Hobbesian answer to this question that Gauthier favours, one that Gaus concedes is extremely appealing, is this: accepting morality’s constraints is instrumentally rational. Morality doesn’t constrain the pursuit of our ends, it is rather the most efficient means of pursuing them, and so the tension between morality’s authority and our own aims dissolves entirely. The aim of chapter 4 is to show why this attempt to derive the authority of social morality from instrumental reasoning is doomed to fail, and the first section of the chapter is devoted to a close examination of what instrumental rationality is.
Gaus begins by alerting us to two dangers in discussing instrumental rationality. First, it’s important not to confuse instrumentally rational behaviour with self-interested or selfish behaviour. Instrumental rationality makes no assumptions about what aims or objectives people might pursue, and so super-altruists can be every bit as instrumentally rational as super-egoists. Second, Gaus warns against confusing Gauthier’s question with a similar-looking question, the famous “why be moral?” question. To ask this question, however, is to mistakenly suppose that even if morality’s normative basis and content are securely understood, we face a further question about whether we ought to be moral that must somehow be answered from outside of morality. Gauthier’s question is different and more sensible; he wants to know whether there is a good basis to support morality’s alleged authority over us. Maybe morality’s claims to authority are just a sham, and we would be better off not making moral demands, in the way we now think there’s no good basis for talking about angels and demons.
With those caveats out of the way, what is instrumental rationality? You might think it must be to a conception of Rationality as Effectiveness: Alf’s action X is instrumentally rational if and only if X-ing is the effective way to achieve his desired goals. The problem with this view (to borrow some helpful terminology from Derek Parfit’s forthcoming book, On What Matters) is that it relies on a fact-relative view of effectiveness. That is, Alf’s X-ing is only instrumentally rational if it turns out that, as a matter of fact, X-ing achieves the desired objective. But surely if X-ing looked like the right decision on the basis of all the evidence available to Alf, X-ing was instrumentally rational even if things turn out badly? If the weather service tells Alf there’s no chance of rain, then surely Alf does not fail to be rational when he doesn’t carry an umbrella, even if it does rain later that day. And surely Betty is not instrumentally rational if, for all the wrong reasons, she carries an umbrella and then ‘gets lucky’ when it starts to rain.
These objections might lead us to endorse Subjective Rationality, whereby Alf’s action X is rational if and only if his choice to X is based on his beliefs, and if his beliefs were true, X-ing would be the effective means to achieving his desired goals. But this account fails because (again to borrow from Parfit) it relies on a belief-relative standard, and this doesn’t seem like the right standard for assessing a person’s rationality. Alf might hold obviously false and silly beliefs, and it seems wrong to declare that acting on the basis of absurd or obviously false beliefs can vindicate one’s actions as instrumentally rational. Gaus concludes that what we need is a conception of instrumental rationality that is similar to the subjective view, but places some restrictions on what can count as a plausible belief. As Gaus puts it, Alf’s action X is instrumentally rational only if Alf soundly chooses X because he believes it is the best prospect for achieving his goals. Here Gaus relies on the rough idea that Alf’s beliefs must not be grossly defective from his own epistemic perspective, and the deliberation leading to action must be similarly not grossly defective, but it seems clear Gaus is wary of setting a more stringent standard.
Having defined the basic idea, Gaus then highlights three fundamental features of instrumental rationality:
- More is Better than Less. This is pretty self-explanatory. If you can get more of what you want without sacrificing anything else that you want, then it is instrumentally irrational to prefer less of what you want.
- Modular Rationality. Actions are routes to achieving our goals, and so instrumental rationality tends to be inherently forward-looking: the reason to perform any action must be explicable in terms of what that action will bring about. But things are actually more complex because, as Gaus explains, there are some direct consumption activities that cannot be explained in this way. Alf eats ice cream now not for some future purpose, but rather because he likes eating ice cream. He’s not doing it now because this will bring future benefits, he’s eating now because eating now is the benefit. This raises a worry about circular explanations, since we could then “explain” any individual’s act X by saying “he’s X-ing because his aim is to X”. Gaus thinks we can overcome this problem by relying on the type/token distinction in cases of direct consumption. So we explain Alf’s token act of X-ing by saying that his aim is to engage in acts of type X, and we thereby distinguish the explanandum (the token) from the explanans (the type). The more general point about instrumental rationality being modular is that it can be formalized as a decision tree. At each point in the tree Alf faces a decision about how best to promote his goals from this point forward, and each decision he makes brings him to a new node in the tree, where he faces a similar decision. Previous decisions have no necessary relevance to the current decision, in the sense that something which may have been instrumentally rational at t0 (threatening one’s enemy with nuclear annihilation should he attack) does not, at t1, make it instrumentally rational to follow through on that threat if one’s enemy has now attacked and there is no benefit to launching one’s nuclear arsenal. This point about modular rationality is significant, and will play a major role in the chapter’s critique of Gauthier.
- Consistency. Instrumentally rational agents have transitive preferences. There is some dispute about this, but I don’t think we need to get into the details.
- Gaus rejects the fact-relative standard of Rationality as Effectiveness, and he rejects the belief-relative standard of Subjective Rationality. Instead he favours a loosely defined view where Alf’s act must be based on beliefs and deliberation that are not grossly defective in order for Alf’s act to qualify as rational. Some people might object to Gaus’s unwillingness to make this standard more precise, but this is not the worry that I want to press (sometimes standards just have to be imprecise). Instead I want to point out that there is a third, very plausible, conception of rationality that Gaus ignores. Let’s call it (again borrowing from Parfit) Evidence-Relative Rationality. On this view, Alf’s act X is rational if and only if X-ing appears to be the effective means of advancing Alf’s goals on the basis of all the evidence accessible to Alf. Does the evidence available to Alf provide him with apparent reasons to believe that X-ing is the most effective route to his goals? This seems like a very plausible conception of instrumental rationality, and it avoids the counter-intuitive results Gaus presses against the first two views. Gaus’s view seems to me to sit somewhere in between the belief-relative and evidence-relative standards: one’s beliefs might not be grossly defective but that doesn’t show that they are the beliefs that are warranted by the available evidence. It’s not clear to me why we should prefer Gaus’s intermediate position over the more demanding Evidence-Relative standard.
- Second, going back to some of the discussion from last week: I wasn’t entirely confident about Gaus’s distinction at the outset between the ‘why be moral?’ question as opposed to Gauthier’s question about the basis of morality. If someone asks, “what is the basis for morality’s alleged authority over me?” and he believes any successful answer to that question must appeal to some non-moral basis, like instrumental rationality, then I think the person asking the question is really asking something like the “why be moral?” question, only dressed up in different language. This is all a bit speculative, but to suppose, as I think philosophers like Gauthier do, that morality must have a point or deliver some benefits in order to have real authority over me, seems pretty indistinguishable from the person who says, “yes I see that morality’s requirements are justified because they reflect what it means to treat others as free and equal agents, but why should I follow morality’s requirements?”.