There is an archived version here.

See also this comment by Gwern.

What are, in you assessment, some of the most cost-effective ways of throwing money at the problem of reducing existential risk?

This was an interesting read, which I only discovered because the post was highlighted on the front page. I took some quick notes, which I share below in case they are useful to anyone.

• Rescuing Jews during the Holocaust wasn’t an especially effective “intervention”, compared to the internal politics of each Nazi-controlled country where Jews lived.
• Moreover, rescuers didn’t appear to have many common traits, so creating would-be rescuers to prevent future genocides is not very tractable.
• One trait that appears to have been shared by rescuers is that they “were taught to appreciate a tolerance for people who were different from themselves”. But the author objects that this may just be a fake post hoc “explanation”.
• Most rescuers appear to have been, in Eva Fogelman’s typology (see below), “moral rescuers”. They also appear to have been reactive rather than proactive: they helped only after being asked to help, though sometimes this caused them to eventually become proactive helpers.
• “People were generally willing to let the Holocaust proceed without intervening. It almost always took a personal plea from a persecuted person for altruism to kick in. Once they weren't just an anonymous member of indifferent crowd, once they were left with no escape but to do a personal moral choice, they often found out that they are not able to refuse help.”
   

Fogelman’s typology of rescuers

• Moral rescuers: The people whose main motivation was: "How could I have acted differently?" or "How would I be able to live with myself if I haven't helped?" Interesting tidbit: It seems that they rarely express those feelings in religious terms.
• Judeophiles: As far as I understand, these were mostly people who had loved someone Jewish, suspected that they may be of Jewish descent themselves (e.g. born out of wedlock) or who were admirers of Jewish culture, the latter mostly on religious grounds.
• Concerned professionals: This is an interesting group. Professionals, such as doctors or diplomats whose job is to help people in need. They just went on and continued what they perceived as their work. It must have required particular understanding of what "work" means though. For example, diplomats often defied orders of their governments to help the Jews.
• Network rescuers: Rescue organizations. Or, often, just anti-Nazi organizations which also saved Jews on the side. The author claims that the main motivation for this group of rescuers was hate of the Nazi regime. Saving Jews was more of a side effect.
• Child rescuers: Oh my, I totally forgot that kids were also playing part in this shit. In any case, they rarely made any conscious decision. They were just dragged into it by their parents.

Jew rescuing and cognitive dissonance

Another psychological effect I see in play here (although with much less confidence than with the bystander effect) is cognitive dissonance and, specifically, the effect it has on one's morality, as explained by Carol Tavris in her Mistakes Were Made (but Not by Me) book.

The book asks you to imagine two students who are very much the same. On the test one of them decides to cheat, the other one decides not to. This may be because of completely external reasons. For example, one of the students have prepared for the topic A, the other one prepared for the topic B. By accident, the test focuses on topic B. The second student doesn't have to cheat because she's prepared. The first student doesn't know much about B and so she decides to cheat.

After the test, both students try to minimize their cognitive dissonance. The non-cheating one is likely to endorse statements such as "all cheating is bad" or "only bad people cheat" and "all cheaters should be expelled". The cheating student, on the other hand, is more likely to identify with statements such as "the tests are only a farce" or "cheating is not a big deal". (See Carol Tavris explain the mechanism in more detail in this video.)

Now try to apply that to a person being asked to help by a Jew in distress.

They may decide not to help because the stakes are too high. If the Nazis found out, they would execute the entire family. But the understanding that you've basically sentenced a person to death is not an easy one to live with. To ease the cognitive dissonance between what the subject believes about himself and what he had done he's likely to start believing things like "Jews are not human" or "Jews are intrinsically evil and should be eliminated for the benefit of all". In the end he may turn in his neighbor, who's hiding Jews, to the Gestapo.

If the strategy failed in predictable ways, shouldn't we expect to find "pre-registered" predictions that it would fail?

This, and see also Gwern's comment here.

“How many Oxford dons does it take to change a lightbulb?”

CHANGE‽‽‽

Greg Lewis discusses this at length here.

Metaculus generates lots of valuable metrics besides the "meaningless internet points" about which Zvi and others complained. If Yudkowsky had predicted regularly, he would have been able to know e.g. how well-calibrated he is, how his Brier score evolved over time, how it compares to the community's, etc.