I don't understand.

If you're trying to dam a river, and you only have 100,000 bricks, then there is a normative solution, i.e., the solution that has the greatest chance of successfully damming the river. Talking about solutions that require one million bricks is talking about a different problem that is only relevant to people with millions of bricks. So when you say, "identity shouldn't act as a normative consideration", that sounds to me like, "you should already have one million bricks, there is no normative solution if you only have 100,000 bricks". Using 100,000 bricks to dam a river isn't using an approximation of the solution you would use if you had a million bricks. That's why I say "normative" is a misleading word here. It implies that you should try to approximate the million-brick solution even when you know you don't have enough bricks to do that: a tenth of a great million-brick dam is one millionth as useful as a complete 100,000-brick dam. Why not just renormalize such that your constraints are part of your environment and thus part of the problem, and find a normative solution given your constraints? Otherwise the normative solution is always to have already solved the problem. "What would Jesus do? Jesus would have had the foresight not to get into this situation in the first place." "Normative" is always relative to some set of constraints, so I don't see why normative-given-boundedness isn't a useful concept. I'm reminded of Nick Tarleton's intuition that decision theory needs to at some point start taking boundedness into account.