Note Carl Shulman’s counterargument to the assumption of a normal prior here and the comments traded between Holden and Carl.
“If your prior was that charity cost-effectiveness levels were normally distributed, then no conceivable evidence could convince you that a charity could be 100x as good as the 90th percentile charity. The probability of systematic error or hoax would always be ludicrously larger than the chance of such an effective charity. One could not believe, even in hindsight, that paying for Norman Borlaug’s team to work on the Green Revolution, or administering smallpox vaccines (with all the knowledge of hindsight) actually did much more good than typical. The gains from resources like GiveWell would be small compared to acting like an index fund and distributing charitable dollars widely.”
The problem with this analysis is that it assumes that the prior should be given the same weight both ex ante and ex post. I might well decide to evenly weight my prior (intuitive) distribution showing a normal curve and my posterior (informed) distribution showing a huge peak for the Green Revolution, in which case I’d only think the Green Revolution was one of the best charitable options, and would accordingly give it moderate funding, rather than all available funding for all foreign aid. But, then, ten years later, with the benefit of hindsight, I now factor in a third distribution, showing the same huge peak for the Green Revolution. And, because the third distribution is based not on intuition or abstract predictive analysis but on actual past results—it’s entitled to much more weight. I might calculate a Bayesian update based on observing my intuition once, my analysis once, and the historical track record ten or twenty times. At that point, I would have no trouble believing that a charity was 100x as good as the 90th percentile. That’s an extraordinary claim, but the extraordinary evidence to support it is well at hand. By contrast, no amount of ex ante analysis would persuade me that your proposed favorite charity is 100x better than the current 90th percentile, and I have no problem with that level of cynicism. If your charity’s so damn good, run a pilot study and show me. Then I’ll believe you.
Note Carl Shulman’s counterargument to the assumption of a normal prior here and the comments traded between Holden and Carl.
“If your prior was that charity cost-effectiveness levels were normally distributed, then no conceivable evidence could convince you that a charity could be 100x as good as the 90th percentile charity. The probability of systematic error or hoax would always be ludicrously larger than the chance of such an effective charity. One could not believe, even in hindsight, that paying for Norman Borlaug’s team to work on the Green Revolution, or administering smallpox vaccines (with all the knowledge of hindsight) actually did much more good than typical. The gains from resources like GiveWell would be small compared to acting like an index fund and distributing charitable dollars widely.”
The problem with this analysis is that it assumes that the prior should be given the same weight both ex ante and ex post. I might well decide to evenly weight my prior (intuitive) distribution showing a normal curve and my posterior (informed) distribution showing a huge peak for the Green Revolution, in which case I’d only think the Green Revolution was one of the best charitable options, and would accordingly give it moderate funding, rather than all available funding for all foreign aid. But, then, ten years later, with the benefit of hindsight, I now factor in a third distribution, showing the same huge peak for the Green Revolution. And, because the third distribution is based not on intuition or abstract predictive analysis but on actual past results—it’s entitled to much more weight. I might calculate a Bayesian update based on observing my intuition once, my analysis once, and the historical track record ten or twenty times. At that point, I would have no trouble believing that a charity was 100x as good as the 90th percentile. That’s an extraordinary claim, but the extraordinary evidence to support it is well at hand. By contrast, no amount of ex ante analysis would persuade me that your proposed favorite charity is 100x better than the current 90th percentile, and I have no problem with that level of cynicism. If your charity’s so damn good, run a pilot study and show me. Then I’ll believe you.