Recently on Twitter, in response to seeing a contest announcement asking for criticism of EA, I offered some criticism of that contest’s announcement. That sparked a bunch of discussion about central concepts in Effective Altruism. Those discussions ended up including Dustin Moskovitz, who showed an excellent willingness to engage and make clear how his models worked. The whole thing seems valuable enough to preserve in a form that one can navigate, hence this post.
Quick note of gratitude: thanks for taking the time to highlight the Twitter discussion in a permanent and discoverable way! I generally stay off Twitter to avoid unbounded time sinks, but it comes at the cost of missing cool exchanges like these.
Forget it Zvi, its <strike>Chinatown</strike> Man-of-System town. Man of system will never understand local knowledge and just think it measn "can't observe as much" because local is smaller geographically than global is. You can beat them over the head with the fact it means, say, a hot dog vendor has an intuitive grasp on what traffic on their street looks like and are better at spotting bombers than than any rules the man of system can devise (look up Duane Jackson's story, for reference). Deviant Olam's first, second, and third bit of security advice at home is "get to know your neighbors".
I'm 100% on board with the give based on your local knowledge, with a little bit of seeking out donations that aren't tax incentivized. One of my hobby horses is giving a car to folks who would lose their jobs without transportation - not wildly expensive cars, just cars that ran well but would only trade in for <$3,000. This is basically the white trash version of giving a laptop to a grad student.
Holding an open application process (what some people might call an All-Pay Auction...) to give money to a grad student is the Man of System version, and, uh, yeah do the math. Its a negative EV proposition.
Thanks for writing this up. I notice that Dustin is coming across very, very badly to me. It is reminiscent of some people I have worked with in collaborative project implementations who, upon being shown some errors or shortcomings in e.g. a data load file, dig in their heels and say effectively "Nuh uh! NO, YOU!" instead of trying to understand what is the issue and finding the source of the problem. While I have some sympathy there, having had people try to shove their mistakes on me and having to push back, there is a reflexive "never admit error" flavor that gets really old after a while. I am possibly over sensitive to that, or just the nature of Twitter in general, but Dustin does not sound like someone who is really interested in understanding your critique, or really any suggestion that isn't "Give Dustin's organization money." Probably a trait selected for when climbing an organization, I suppose.
Then again, the writers of Talent came across badly to me too, so maybe I am just grumpy these past days and don't realize it :)