I know I'm bringing Drescher up a lot recently, but this exchange reminds me of some of his points, and how, after reading Good and Real, I see Haidt's work (among other people's) in a different light.

Drescher's theory of ethics and decision making is, "You should do what you [self-interestedly] wish all similarly situated beings would do" on the basis that "if you would regard it as the optimal thing to do, then-counterfactually they would too".

He claims it implies you should cast a wide net in terms of which beings you grant moral status, but not too wide: you draw the line at beings that don't make choices (in the sense of evaluating alternatives and picking one for the sake of a goal), as that breaks a critical symmetry between you and them.

Taking your premise that fish don't reflect on their actions, this account would claim that they likewise do not have the moral status of humans. But it would also agree with you that it's insufficient to point to how they eat each other, because "I would not want some superbeing to eat me simply on the basis that I eat less intelligent beings."

Also, Drescher accounts for our moral intuitions by saying that they are a case of us being choice machines which recognize acausal means-end links (i.e. relationships between our choices and the achievement of goals that do not require the choice to [futurewardly] cause the goal). This doesn't necessarily contradict Haidt's argument that we judge things as right because of e.g. ingroup/outgroup distinctions (he says that functional equivalence to acausal means-ends links is all that matters, even if the agent simply feels that they "care" about others), but it does tend to obviate that kind of supposition. [/show-off]