I've sort of gone over this before, but here's the basic premise: We should strive to take actions which appear logical,* regardless of to which arbitrary realms of agent interaction they belong. I've devised an extremely simplistic way of looking at this here. Consider the following two statements:
1. Mass murder is acceptable.
As noted previously, there is no such thing as mathematical data uncoupled from empirical observation; there have to be empirically observable objects which can be added or subtracted in order for arithmetic to even exist.
So, then, is there a difference between the above statements? Consider their inverses:
1. It's wrong to commit mass murder.
Do you agree with 1.? Do you agree with 2.? Great, so what's the problem, and why are we placing one in a different category from the other?
And technically, no, I can't "prove" that either statement is true, contrary to what some mathematicians and scientists might allege. With this limitation in mind, if I were to use the logic of most people, I'd have to declare 2+2=4 a "subjective opinion" in the same sense that that term gets applied to things like morality. This is where the pragmatic element comes into play: we have to make decisions. Period.
Note, however, that I bring this up not just to demonstrate that all truth claims should be measured by the same metrics, but also to elucidate my take on what's worth promoting. It's a statement both to those who view suffering as too subjective a phenomenon to care about addressing, and to those who are gung ho about ending suffering, but who are coming at things from a "moral" perspective, which probably isn't fundamental enough. If we want to end suffering, we can't just teach people that it's not a good thing; we have to make them into logic machines for any kind of situation.
* In this context, "logical" refers to any action which appears more logical than all competitors, thus relegating the competitors to being "illogical," given that none of them were or will be opted for.