You're right that I didn't mean this necessarily to be about a specifically moral sense of 'ought'. As for the suggested inference about baby-punching, I would push that onto the 'other things being equal' clause, which covers a multitude of sins. No acceptable theory can entail that one ought to be punching babies, I agree.

The picture I want to suggest should be taken seriously is that on one side of the fence are naturalistic properties, on the other are properties such as rationality, having wants/beliefs, being bound by norms as to right and wrong (where this can be as meagre a sense of right and wrong as "it is right to predicate 'cat' of this animal and wrong to predicate 'dog' of it" -or, "it is right to eat this round, red thing if you desire an apple, it is wrong to eat it if you desire a pear"), oughts, goals, values, and so on. And that the fence isn't as trivial to break down as one might think.

I'm understanding a utility function is something like a mapping of states of affairs (possible worlds?) onto, say, real numbers. In this context, the question would be giving naturalistic sense to the notion of value -that is, of something's being better or worse for the agent in question- which the numbers here are meant to correlate to. It's the notion of some states of affairs being more or less optimal for the agent -which I think is part of the concept of utility function- which I want to argue is outside the scope of natural science. Please correct me on utility functions if I've got the wrong end of the stick.

To be clear - the intent isn't to attack the idea that there can be interesting and fruitful theories involving utility functions, rationality and related notions. It's just that these aren't consistent with a certain view of science and facthood.