@gm_silverman @dbgreenwald There is no demand for it, and a lot of opposition to a globalized participant pool. Those demographics are readily available where they have value (AI/ML) but are a poison well to most US research. Even US samples are hardly real nowadays from similar selection pressures IMO.
@gm_silverman @dbgreenwald 2021 vs 2019 % of our users (comparing to the international payments pre-release):
US: -15% (was >80% of our userbase in 2019, down to 70% now)
India: +28%
Brazil: +557%
UK: +7%
Italy: +141%
@gm_silverman @dbgreenwald The invite system is the same for everyone. Even US workers go through MTurk's internal vetting system, but it requires no extra effort on the workers part. It is opaque, but as trends go they've definitely increased the presence of non-US workers vs 2 years ago.
@gm_silverman @dbgreenwald Not sure where you got that info, but the non-US population on MTurk is substantially larger & more diverse than it's ever been. India/Brazil makes up a bulk of the non-US populace, but Canada/Aus/EU populations are also larger than 2 years ago.
@BigDealMandy @MomTips4Success You can also follow this page: https://t.co/liYm7cy1Sk
It works a bit like finder/similar tools but specifically for qualifications.
@nerdpro That's probably not a bot. We provide that service to workers & it's safe to click the link (assuming it's to our site), but if you do find someone misusing it or want more info feel free to reach out!
You can read about the actual functionality here: https://t.co/9WeBUHCxWq
@todd_gureckis@arossotto@DG_Rand@AaArechar@byrd_nick Can't have 30 people walk into a shelter to kick puppies then act surprised when a dog bites. I know from our side the instances and breadth of abuse from academic sources has grown immensely.
I think it definitely takes a toll, though maybe contributory towards optimizing $?
@todd_gureckis@arossotto@DG_Rand@AaArechar@byrd_nick I read a lot of these (and your piece is one of the better ones), but it's strange that none of seem to include a mirror.
Some researchers just aren't very good at research. Some aren't great at being people.
Not to be crass but perhaps people should stop peeing in the well?
@brendan642@yoavgo@nlpmattg But yes, as you point out that swing can be wild and reducing to averages tends to hurt the most "at risk" populations as well (ie, those using assistive tech, etc)
@brendan642@yoavgo@nlpmattg Most workers use their time in the task to calculate their hourly per task (which doesn't account for time searching for work). Some extrapolate that for their entire work day but that's less common metric to share and usually workers just use flat daily/weekly/etc goals.
@nlpmattg@yoavgo@asayeed Definitely not as simple as "workers wouldn't work" - there are endless examples of large projects with tons of ethical shortcomings on the platform. Esp if the conversation is larger than simple $/hr conversions.|
Many of the (b) pts are to some degree standard conditions
@solda_eco By all means do what works for your study but as a person who works on MTurk regularly this is the kind of thing that trips me up and has nothing to do with my quality of work or my reaction time inside of a HIT itself.
@solda_eco I understand but the execution is helping trip people here. It's like these kind of things
https://t.co/7tb7wpBrza
Honestly after reading it multiple times now, I promise I'd have a solid shot of pressing the wrong button no matter what with how it's worded.
@solda_eco Also a TON of participants report having issues with your pilot functionality (500 errors) on our system, which could have contributed to your original drop-out rate perhaps?
https://t.co/9J8Q11oW6e
https://t.co/Tq0SpbtgKk
@solda_eco A lot of those workers see instructions similar to this multiple times a day/week, it becomes pedantic to re-read every line over and over.
I understand what you're going for, but strongly disagree with the approach. And especially drawing much in the way of conclusions from it.
@marybhargis In order to contact workers through MTurk you'll need to be comfortable using the API. No simple way of doing it through the web UI.
Recommend using @cloudresearch for longitudinal studies, it's just easier and worth the minor cost increase.
https://t.co/estaBBZ1fY