Eh, Firenze was taking initiative to dispose of a major problem even if it required actions he considered morally distasteful. Compared to Quirrell, he's pretty dumb, but he hasn't distinguished himself for idiocy the way, say, the Ministry official who took self-destructive joy in obstructing him did. If anything, he probably distinguished himself as cleverer than the norm, if not in any way a peer.

This seems inapt as a generalization about human psychology.

In one psychology experiment which a professor of mine told me about, test subjects were made to play a virtual game of catch with two other players, where every player was represented to each other player only as a nondescript computer avatar, the only input any player could give was which of the other two players to toss the "ball" to, and nobody had any identifying information about anyone else involved. Unbeknownst to the test subjects, the other two players were confederates of the experimenter, and their role was to gradually start excluding the test subject, eventually starting to toss the ball almost exclusively to each other, and almost never to the test subject.

Most test subjects found this highly emotionally taxing, to the point that such experiments will no longer be approved by the Institutional Review Board.

In addition to offering a hint of just how much ethical testing standards can hamstring psych research, it also suggests that our instinctive reactions to ostracization do not really demand identifying information on the perpetrators in order to come into play.

That sounds a lot more like a Rowling type twist than an Eliezer type twist. There are elements that could be interpreted as vague and oblique hints, but it doesn't suggest particularly clever or well-considered behavior on anyone's part.

Well, if they have access to the dorms, they could steal the books and replace them with altered copies.

My regular commute has been impeded by such a set of escalators (currently dismantled for repairs from fire damage) for weeks.

I think that, in a video game sense (which is really the only context where the distinction of "player characters" makes real narrative sense,) "sidekick" type characters probably do tend to be NPCs. But I think this is a major weakness of using a video game framing for the concepts under discussion. Problems are rarely solved in real life the way they're solved in books, but they're pretty much never solved in real life the way they are in video games.

If they're a truly isolated hermit, that distinction would presumably no longer apply, but the world is pretty short on truly isolated hermits.

I think you probably could kill and replace an isolated hermit in a QALY-neutral way (you'd probably need a fairly unhappy person to keep it QALY neutral even,) whereas with social connections in the equation, if you were trying to kill and replace non-hermits in a QALY neutral way, you'd ultimately end up having to do it to everyone.

At the very least because an already-born person will almost always leave survivors aggrieved and/or materially harmed by the act, while aborted fetuses often do not.

Actually I for one gave up Death Note in frustration very early on because I couldn't help focusing on how much of the real inferential work was being done by the authors feeding the correct answers to the characters. Like when L concludes that Kira must know the victim's real name to kill him... there were so many reasons that just didn't work. Kira's apparent modus operandi was to kill criminals, there was no particular reason to suppose he would respond to a challenge to kill anyone else, so the fact that he didn't was already weak evidence regarding whether he could at all, let alone what the restrictions might be. Whether Kira knew his real name or not was just one variable switched between him and Lind L. Taylor. L could just as easily have been immune because he eats too many sweets.

While smart, knowledgeable people can often extract a greater yield of inference from a limited amount of data than others, I find that far too many writers take this idea and run with it while forgetting that intelligence very often means recognizing how much you can't get out of a limited amount of data.

With Fate/Stay Night, one problem is that I was looking at ripped videos on Youtube, while the original material is a "visual novel" with branching paths, so it's possible (but unlikely) that the people who put up the videos missed all the rationality-relevant bits.

I haven't watched the anime, but I have read the visual novel, and the anime does not have a reputation for being a very faithful adaptation. The visual novel at least does share themes that often feature in Eliezer's work, but I wouldn't call them "rationality content" as such. More in the manner of Heroic Responsibility and related concepts.