The problem is that those are post-hoc rationalizations. "Morality" in day-to-day life, it seems clear to me, is a set of mental heuristics (a la Kahneman's Sys

The problem is that those are post-hoc rationalizations. "Morality" in day-to-day life, it seems clear to me, is a set of mental heuristics (a la Kahneman's System 1 thinking). It produces good enough answers for individual-scale decision making like the "real world" equivalent of the trolley problem, because the heuristics have evolved to implicitly take into account the uncertainty you've listed. But it doesn't make a given decision because of those well-considered caveats, it's just calibrated to hit the mark more times than not. Then, after it's made the decision, it's possible for one to find reasons to justify what it says. The problem is that said moral heuristics are also used to apply morality to non-individual-scaled issues. The answer produced for, say, "Should self-driving cars be designed swerve and crash into one person over going straight into five?" doesn't have those caveats, but the heuristics recoil all the same. So the point of the thought exercise is to reduce the number of "moral variables" to a minimum set, to a) examine what our gut heuristics say and b) see if that actually maps reasonably between various situations. Inserting the "real world" into it is the exact opposite of the point.