Is sharing this link a good idea? I'm strongly in favor of helping Givewell get better data, but I'm worried about introducing selection effects into their survey.
EDIT: GW is okay with this link. "We’re looking for replies from GiveWell followers ... We’re fine for you to pass it around if you want."
I wondered the same thing, but notice that they ask how much you donated to them / through them. So even if we flood in and produce hundreds of responses, as long as we're honest, the analysis will show that the people who are actually donating money look like X and everyone else looks like Y, and Givewell can decide appropriately. (And if LWers really are donating a ton to Givewell or its top charities, well, then the flooding is not a problem at all.)
Weighting GiveWell donors' responses more heavily will mean giving the opinions of people who are happy its current work disproportionate influence, won't it? It may be worthwhile to distinguish between people who are interested in effective altruism and people who follow GiveWell's recommendations.
Hrm. Somewhat fair, but I think the current (as far as I can tell) selection effect of "people who regularly read the Givewell blog" is about as bad. I think people who don't currently read Givewell regularly, in part because it usually doesn't emphasize the sorts of things they care about, should have the opportunity to tell Givewell that they'd pay more attention if Givewell emphasized the sorts of things they care about.
I asked about this on the blog itself, will take this down if they ask me to.
They also sent a newsletter linking to the survey to their mailing list, so even if I don't read their blog that often I still took the survey because I had found out about it in the mail.
Givewell's blog has recently begun a series of 5 self-evaluation posts (they are on the 4th right now) which discuss where the organization is at and where they're going. They're all worth a read. In particular, they build up to a survey for Givewell followers about how you'd like the organization to direct their research in the future, with options to emphasize existential risk and research even if the evidence is lower quality.