{"items": [{"author": "Gordon", "source_link": "https://www.facebook.com/jefftk/posts/626437222662?comment_id=626455680672", "anchor": "fb-626455680672", "service": "fb", "text": "Wow, great job! How long did it take you?", "timestamp": "1378025564"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/626437222662?comment_id=626463969062", "anchor": "fb-626463969062", "service": "fb", "text": "@Gordon: maybe 4hr total? But a lot of it was bus-time.", "timestamp": "1378042829"}, {"author": "Nick", "source_link": "https://www.facebook.com/jefftk/posts/626437222662?comment_id=626478045852", "anchor": "fb-626478045852", "service": "fb", "text": "Thanks, Jeff!", "timestamp": "1378051025"}, {"author": "Pablo", "source_link": "https://www.facebook.com/jefftk/posts/626437222662?comment_id=626491947992", "anchor": "fb-626491947992", "service": "fb", "text": "Great work. I think having these transcripts is really valuable.
(You may want to bear in mind that, as Satvik Beri pointed out recently, one can get one hour of audio transcribed for about 60 USD on Amazon's Mechanical Turk. According to Satvik, it typically takes five hours of work to transcribe an hour of audio.)", "timestamp": "1378060930"}, {"author": "Carl", "source_link": "https://www.facebook.com/jefftk/posts/626437222662?comment_id=626545854962", "anchor": "fb-626545854962", "service": "fb", "text": "Thanks, Jeff.", "timestamp": "1378086930"}, {"author": "Satvik", "source_link": "https://www.facebook.com/jefftk/posts/626437222662?comment_id=626564397802", "anchor": "fb-626564397802", "service": "fb", "text": "60 USD from transcription services, ~30 USD from Amazon Mechanical Turk.", "timestamp": "1378096844"}, {"author": "Sasha", "source_link": "https://www.facebook.com/jefftk/posts/626437222662?comment_id=626598334792", "anchor": "fb-626598334792", "service": "fb", "text": "My rough summary of the views, having read it:
All agree that flow-through effects (FTE) dominate near-future effects of almost any intervention, but disagreement on what that implies.
Holden sceptical that the marginal value of researching FTE justifies it \u2013 in particular for him and Givewell employees, less so (but still so) for other EA groups. He reasons that a) the kind of evidence you\u2019d need to come up with for a robust case is scarce/non-existent, and b) non-robust expectation guesstimates are, in his experience, so tainted, that he doubts they would persuade him to change any of his priorities.
He also points to the interconnectedness between fields of research (and advances in general), suggesting that it muddles predictions of whether an advance in a field has higher expectation than one in a similar field.
Others (esp Paul?) more positive. Their case something like: there could be tremendous value from researching FTE, and almost no research has been done, so that a moderate amount of research (something around 8 work-years by Rob\u2019s estimate) would provide a pretty clear view of how effective such work would be. In the worst case scenario, the failure of the concerted effort would be published so that others wouldn\u2019t duplicate it, in the best case it would provide a huge return. If such a thing is ever worth doing, it\u2019s probably worth doing ASAP.", "timestamp": "1378115079"}, {"author": "Nick", "source_link": "https://www.facebook.com/jefftk/posts/626437222662?comment_id=626598758942", "anchor": "fb-626598758942", "service": "fb", "text": "For what it's worth, I'm pretty intermediate and think that it makes sense for GiveWell to continue its current research plan for now. If I have a difference of opinion with Holden it is probably that I would be in favor of looking at FTEs earlier than him and maybe looking harder.", "timestamp": "1378116403"}]}