Would you still want them to flip the switch, even though it would result in your death.

Try not to Kobayashi Maru this question, at least not yet. I know you can criticize the scenario and find it unrealistic. For instance, you may say you won't push because the man might fight back, and you'd both fall but not till after the trolley had passed so everyone dies. So imagine the fat man in a wheelchair, so he can be lightly rolled off the bridge. And if you're too socially constrained to consider hurting a handicapped person, maybe the five people tied to the tracks are also in wheelchairs. If you think that being pushed off a bridge is more terrifying than being hit by a train, suppose the fat man is thoroughly anesthetized. Yes, this is an unrealistic thought experiment; but please play along for now.

Just so you know, the common term for this around here is don't fight the hypothetical.

Alternatively, identify external factors that can be statistically shown to increase defection, and then lower the influence of those external factors rather than expect people to magically overcome them. If you can statistically demonstrate that hungry people are more likely to defect, and you don't want people to defect, what will suit you better: bitching that anyone who defects because they're hungry is a morally bad person, or actually handing them a meal?

I'm not sure that's the entirety of what he's getting at. I think he's saying "don't make it acceptable for people to make excuses for defecting, because people will then use that as an excuse in cases where they would otherwise cooperate."

That said, your idea is still a good solution to the way you interpreted that statement.

Optimal performance may be maximized, but the output isn't.

I would be surprised if there were less overall errors in the final product if it started at 2 per page, rather than say 1/4 per page.

This is also valid against the suggestion in the OP. Although humans will catch more errors if there are more to begin with, that doesn't mean there will be less failures overall.

Other people use my computer fairly often, and I just set it to shift back and forth with command+shift+1. I can change it before I give it to them, so others won't even know if I don't want them to.

A possible answer:

There are many different kinds of pain and pleasure, and trying to categorize all of them together loses information.

For starters, the difference between physical and mental pain and pleasure.

To get more nuanced, the difference between the stingy pain of a slap, the thudy pain of a punch, the searing pain of fire, and the pain from electricity are all very distinct feelings, which could have very different circuitry.

I'm not as sure on the last paragraph, I would place that at 60% probability.

Have you actually asked if they can tell a difference, or have they just not said anything? Because it's considered socially rude to tell someone they need to take a shower.

A bit too much, yeah. Over half the post is defensiveness and reasons why you might object, without refuting those objections.

If we don't say bacteria need ethical reflections, then it is very unlikely that plants will either.

Without going into the details, you could hypothesize a simple mind than automatically rejects any argument. This would by itself prove the No Universally Compelling Arguments theory.