Suppose we are considering whether being A is more morally valuable than being B. If we don't require evidence when making that decision, then lots of ridiculous conclusions are possible, including racism and sexism.

The issue isn't whether you require evidence. The issue is solely which moral yardstick are you using.

The "evidence" is the application of that particular moral metric to beings A and B, but it seems to me you're should be more concerned with the metric itself.

To give a crude and trivial example, if the metric is "Long noses are better than short noses" then the evidence is length of noses of A and B and on the basis of this evidence we declare the long-nose being A to be more valuable (conditional on this metric, of course) than the short-nose being B. I don't think you'll be happy with this outcome :-)

Oh, and you are still starting with the predefined conclusion and then looking for ways to support it.