I’m not sure how useful this data is given that there are major distribution effects. ie. If I distribute the survey through Less Wrong, I’ll find a lot of people who first heard of the movement through Less Wrong, ect.
Yes − 80,000 Hours only briefly mentioned the survey at the end of one newsletter. If we had promoted it more heavily, we could have probably got more than twice as many submissions, and they would be more tilted towards people who first found out about EA from 80,000 Hours. As a comparison, our annual impact survey gets around 1000 responses each year, which would make it about the same size as this survey.
Yep, that is an issue. One idea might be to look at the data for each referral source (e.g., how everyone who heard about the survey through Facebook heard about EA, then how everyone who heard about the survey though SlateStarCodex heard about EA, etc.).
I agree, this is something we acknowledge multiple times in the post, and many times throughout the series. The level of rigor it would take to bypass this issue is difficult to reach.
This is also why the section where we see some overlap with Julia’s survey is helpful.
I’m not sure how useful this data is given that there are major distribution effects. ie. If I distribute the survey through Less Wrong, I’ll find a lot of people who first heard of the movement through Less Wrong, ect.
Yes − 80,000 Hours only briefly mentioned the survey at the end of one newsletter. If we had promoted it more heavily, we could have probably got more than twice as many submissions, and they would be more tilted towards people who first found out about EA from 80,000 Hours. As a comparison, our annual impact survey gets around 1000 responses each year, which would make it about the same size as this survey.
Yep, that is an issue. One idea might be to look at the data for each referral source (e.g., how everyone who heard about the survey through Facebook heard about EA, then how everyone who heard about the survey though SlateStarCodex heard about EA, etc.).
I agree, this is something we acknowledge multiple times in the post, and many times throughout the series. The level of rigor it would take to bypass this issue is difficult to reach.
This is also why the section where we see some overlap with Julia’s survey is helpful.