An argument against merely substantive views of rationality
On a very abstract level, I think we can distinguish between at least three different types of theories of rationality.
First, there are merely structural accounts (MSA). The easiest way to define MSAs is in terms of wide-scope oughts. MSAs hold that one is rational just in case one’s attitudes are formally coherent. The following are some example principles that an MSA account might endorse:
Modus Ponens Principle: One must: (If one believes ‘a’ and believes ‘if a, then w’, then one believes ‘w’)
Non-Contradiction Principle: One must: (If one believes ‘a’, then not believe ‘not-a’)
Means to an End Principle: One must:(If one’s end is E, and one believes M is a means to E, then intend to M)
An important characteristic of MSAs is that they hold that agent A is rational just in case her attitudes are in any coherent pattern. So, if A believes ‘a’ and ‘not-a’, she can become rational by ceasing to believe either ‘a’ or ‘not-a’, no matter which (if either) is decisively supported by A’s evidence. The most prominent (former) supporter of an MSA view is John Broome. Derek Parfit might hold an MSA view (who knows), and Scanlon in What We Owe to Each Other flirted with the idea (although it’s not clear if he did hold an MSA view then, and it is clear he doesn’t now).
I think that MSA views are clearly false. Consider Good and Bad. Both Good and Bad are in exactly the same epistemic situations (i.e. they both have the same evidence base). Good is a very good reasoner. She is very good at responding to her reasons. Let’s say at time T all of Good’s attitudes are justified (i.e. all of her attitudes are are decisively supported by her evidence). Bad, on the other hand, is a tremendously bad reasoner. She does not respond to her reasons well at all. At time T, all of her attitudes are the negations of Good’s. So, if Good believes that p, Bad believes not-p. If Good intends that q, Bad intends that not-q. Obviously enough, both Good and Bad’s attitudes are formally coherent. If the MSA is true, then they are both equally rational. But that means that neither is open to criticism. I have the incredibly strong intuition that Bad is open to criticism, whereas Good is not. Not all formally coherent sets of attitudes are equal.
The second type of view is a merely substantive account of rationality (MSuA). MSuAs hold that formal coherence matters only insofar as one’s attitudes are formally coherent when one correctly responds to all of one’s reasons. In other words, there are no wide scope oughts and there are only narrow scope oughts. Thus, when one believes ‘a’ and ‘not-a’, the only coherent set of attitudes one can rationally enter into is the one that is most supported by one’s reasons. If ‘a’ is most supported, then that is one’s only rational option. Niko Kolodny has recently argued for a MSuA.
The third kind of view is a hybrid view (HV). I have a hybrid view. HVs hold that both structural and substantive rationality are important. Here’s an argument against MSuAs. Remember Good and Bad, and now consider Bad Bad. Bad Bad thinks that Bad is an excellent reasoner. She wants to have all of Bad’s attitudes. Unfortunately, Bad Bad is bad at copying Bad (she is bad at being bad). Thus, at T, she has most of Bad’s attitudes, but some of Good’s. She strongly believes that Good is an idiot, and staunchly wants to emulate Bad. I think that Bad Bad is open to a type of criticism that Bad isn’t, viz., one can say to Bad Bad “Look, you’re not even consistent. Your attitudes don’t even match up with each other.” One cannot say this to Bad. All of her attitudes match up with each other. That’s not to say that Bad isn’t open to a whole lot of criticism. But that type of criticism is of a different sort. That is the type of criticism one is open to when she fails to comply with the narrow-scope requirements. In this way, Bad is more criticizable than Bad Bad (because some of Bad Bad’s attitudes are the same as Good’s). Nonetheless, Bad Bad violates some wide-scope requirements, and is thus open to a type of criticism that Bad is not. Thus, I think the hybrid view is the only view that can honor our intuitions.
post is interesting. i didn’t read it in detail, but i wanted to point out there is a lacuna in your treatment. niko kolodny’s view in “why be rational?” brings this out. i won’t sketch the view here, but i’ll just say how its arrive at. he wonders whether the normative force of structural requirements can be explained in terms of reasons (i am assuming you think reasons explain the normativity of substantive requirements is explained in terms of reasons). he concludes that they cannot at least directly. he also argues that we should not accept that there is another primitive normative property which explains the normativity of structural requirements. he, then, gives his view which he calls the transparency account. it claims, roughly, that structural requirements are only apparantly normative and what explains their apparent normativity are reasons.
i’m not saying this as an objection to what you say. it is just a suggestion for something to take a look at because it fits uncomfortably with your taxonomy.
another view which you fits a bit awkwardly with your taxonomy is a view which claims requirements of coherence explain our what we have reason to do as well as explain the structural requirements. Michael Smith has this view he thinks the psychology of an agent who has full information and is in wide reflective equilibrium can explain all of it.
again not an objection, but more of a comment that the taxonomy you suggestion fits a bit awkwardly with these proposals.
i guess you to talk about kolodny sorry i didn’t see that. i’m not sure i followed your response but i only looked at it quickly. i’ll have to look at it again before i say something about that.
Giving a good story about the normativity of structural requirements is tough, and I certainly don’t claim to have done that here (or anywhere). Kolodny (in “Why Be Rational” and “Why Be Disposed to be Coherent”
argues for an error theory against the structural requirements. His argument is that we take the structural requirements to be important because we think that the coherence of our attitudes when we respond appropriately to our reasons matters. Formal coherence as such carries no normative weight at all. I think that the case of Good, Bad, and Bad Bad shows that he’s wrong about that. Formal coherence is better than incoherence, even when the coherent set isn’t the appropriate substantive set. Now I don’t claim to have a good story explaining why formal coherence as such is normative. I’ll get back to you.
I don’t understand why you think that Kolodny doesn’t fit nicely in with my taxonomy. I think that he is the paradigm case of someone with a merely substantive account.
Smith’s view is one that fits a bit awkwardly with my divisions. But since these are supposed to be divisions amongst theories of rationality, and not reasons, I think that we can ignore his account of reasons, and instead focus on his account of rationality. I take it his account of rationality is a merely structural one, but it’s not clear. You know more about Smith than I. Insight would be helpful here.
As for the Kolodny point, my second post was meant to back off the claim that Kolodny is awkward for your taxonomy.
The problem with your treatment of Smith is purely structural is that he thinks the structural account can generate the sorts of reasons which you think are intuitive in your counter example. He will say a fully informed agent who is in reflective equilibrium will desire that there actually selves do such and such where such and such are whatever the reasons are in the case you sketched.