(i) the belief that you would desire to phi if you were more rational
(ii) the desire not to phi [or, more weakly, the absence of any desire to phi]
The agent who holds both these attitudes is irrational by their own lights. So this opens a path from theoretical to practical rationality. If you're rationally required to have belief (i), then it would seem you're also rationally required to have the corresponding desire (or at least not to have the opposite desire).
I owe the above ideas to Michael Smith. Now, Clayton discusses a similar principle:
(1) If a subject judges that she should Φ and it’s not the case that she should refrain from judging that she should Φ, it’s not the case that the subject shouldn’t Φ.
Here's are a couple of variants:
(1a) If S ought to believe that she should Φ, then it's not the case that S shouldn't Φ.
(1b) If S ought to believe that she should Φ, then S should Φ.
The guiding thought here is that a practically rational agent does what she judges she ought to do. So it would place inconsistent requirements on her to prohibit Φ-ing even as we require her to judge she ought to Φ (which will lead to her Φ-ing on pain of irrationality). That would effectively be to require her to be irrational.
What do you think of these principles?