There are a few ways you might expect to be able to do better:
There’s effectively a size floor on things GiveWell can afford to look at because of the amount of money they want to move and limited staff time.
GiveWell recommendations are tailored for a set of preferences that may not be the same as yours, e.g. a preference for high levels of confidence and easy to explain evidence such as RCTs, even at the expense of EV.
Some pieces of information are easier for you to learn than to communicate or for others to verify. For instance, it might make a lot of sense for you to trust a friend you’ve known since childhood a lot, another friend to trust them a little based on your say-so, but strangers on the internet to trust them not at all based on your say-so.
Aceso Under Glass’s post about Tostan is a good example of novel research that has some overlap with the second and third consideration.
I agree that in general you should contribute your research to the commons.
There’s effectively a size floor on things GiveWell can afford to look at because of the amount of money they want to move and limited staff time.
My criticism is narrow and specifically about this point.
Yes, there is a size floor. New researchers will pursue smaller projects. But that is an example of diminishing returns to research. The next researcher should pursue projects only slightly smaller than the previous researcher, for the same reason that the previous research pursued large projects.
Thanks for clarifying. I should have been able to figure this out from your original comment. Oops!
If there is opportunity for more research, shouldn’t the answer be for GiveWell to expand or be duplicated?
GiveWell has limited ability to scale even with more money—my sense is that they don’t have a bunch of potential analysts lined up whom they’d be delighted to hire (or try out) if only there were the money for it. This is a feasibility objection to expansion on current margins. However, GiveWell might reallocate resources to this from other things (e.g. outreach) if they had enough money, so if you have other strong reasons to think GiveWell can do better than you can, it might be good to give them money anyway.
On duplication, doing one’s own research is a sort of duplication, at a smaller scale. I suspect that GiveWell is way past the point at which returns to scale diminish.
(Disclosure: I worked at GiveWell in the past. I have no current affiliation with GiveWell. My opinions are my own, an not a representation of their official opinion.)
There are a few ways you might expect to be able to do better:
There’s effectively a size floor on things GiveWell can afford to look at because of the amount of money they want to move and limited staff time.
GiveWell recommendations are tailored for a set of preferences that may not be the same as yours, e.g. a preference for high levels of confidence and easy to explain evidence such as RCTs, even at the expense of EV.
Some pieces of information are easier for you to learn than to communicate or for others to verify. For instance, it might make a lot of sense for you to trust a friend you’ve known since childhood a lot, another friend to trust them a little based on your say-so, but strangers on the internet to trust them not at all based on your say-so.
Aceso Under Glass’s post about Tostan is a good example of novel research that has some overlap with the second and third consideration.
I agree that in general you should contribute your research to the commons.
My criticism is narrow and specifically about this point.
Yes, there is a size floor. New researchers will pursue smaller projects. But that is an example of diminishing returns to research. The next researcher should pursue projects only slightly smaller than the previous researcher, for the same reason that the previous research pursued large projects.
Thanks for clarifying. I should have been able to figure this out from your original comment. Oops!
GiveWell has limited ability to scale even with more money—my sense is that they don’t have a bunch of potential analysts lined up whom they’d be delighted to hire (or try out) if only there were the money for it. This is a feasibility objection to expansion on current margins. However, GiveWell might reallocate resources to this from other things (e.g. outreach) if they had enough money, so if you have other strong reasons to think GiveWell can do better than you can, it might be good to give them money anyway.
On duplication, doing one’s own research is a sort of duplication, at a smaller scale. I suspect that GiveWell is way past the point at which returns to scale diminish.
(Disclosure: I worked at GiveWell in the past. I have no current affiliation with GiveWell. My opinions are my own, an not a representation of their official opinion.)