I worry that some rationalists, while rejecting wooly dualist ideas about ghosts in the machine, have tacitly accepted the dualists’ baseless assumptions about the gloomy consequences of materialism.

There actually is a way in which they're right.

My first thought was, "You've got it backwards - it isn't that materialism isn't gloomy; it's that spiritualism is even gloomier." Because spiritual beliefs - I'm usually thinking of Christianity when I say that - don't really give you oughtness for free; they take the arbitrary moral judgements of the big guy in the sky and declare them correct. And so you're not only forced to obey this guy; you're forced to enjoy obeying him, and have to feel guilty if you have any independent moral ideas. (This is why Christianity, Islam, communism, and other similar religions often make their followers morally-deficient.)

But what do I mean by gloomier? I must have some baseline expectation which both materialism and spirituality fall short of, to feel that way.

And I do. It's memories of how I felt when I was a Christian. Like I was a part of a difficult but Good battle between right and wrong.

Now, hold off for a moment on asking whether that view is rational or coherent, and consider a dog. A dog wants to make its master happy. Dogs have been bred for thousands of years specifically not to want to challenge their master, or to pursue their own goals, as wolves do. When a dog can be with its master, and do what its master tells it to, and see that its master is pleased, the dog is genuinely, tail-waggingly happy. Probably happier than you or I are even capable of being.

A Christian just wants to be a good dog. They've found a way to reach that same blissful state themselves.

The materialistic worldview really is gloomy compared to being a dog.

And we don't have any way to say that we're right and they're wrong.

Factually, of course, they're wrong. But when you're a dog, being factually wrong isn't important. Obeying your master is important. Judged by our standards of factual correctness, we're right and they're wrong. Judged by their standards of being (or maybe feeling like) a good dog, they're right and we're wrong.

One of the problems with CEV, perhaps related to wireheading, is that it would probably fall into a doglike attractor. Possibly you can avoid it by writing into the rules that factual correctness trumps all other values. I don't think you can avoid it that easily. But even if you could, by doing so, you've already decided whose values you're going to implement, before your FAI has even booted up; and the whole framework of CEV is just a rationalization to excuse the fact that the world is going to end up looking the way you want it to look.