ren

mturk twitter alt, privileged crowdworker

[ren @sarinmt]

good morning

[ren @sarinmt]

friendly reminder to all surveyors that tiktok cannot ruin your survey pool if you use @cloudresearch instead of prolific🥰

[ren @sarinmt]

.@LiveTokenCo lol wtf is this

this wouldn't happen if you added more weight to FMV /price difference for deals

[ren @sarinmt]

if you're thinking of using prolific, just give this thread a read and kindly reconsider https://t.co/5IDlJ8Wq0r

[Leib Litman @leib_litman]

@ekadamer @peterhurford @Prolific Yes I read it. That's why I replied.

[ren @sarinmt]

@ekadamer @leib_litman @peterhurford @Prolific "We have no way of knowing whether your software filters workers coming from MTurk"

when the settings theirselves say "Block Low Quality Participants" and "CloudResearch Approved Participants"?
then have the audacity to call any criticism of the study 'bad faith'?

[ren @sarinmt]

@ekadamer @leib_litman @peterhurford @Prolific why nitpick the "wrong setting" part of this 5 tweet long thread when that doesn't even hold water? you disabled the 'wrong' setting(s) that would make sure you only got data from CR vetted workers. every setting highlighted in the original tweet is required for a pure CR dataset

[ren @sarinmt]

@ekadamer @peterhurford @Prolific "extremely polished version of CR"

except that these highlighted settings are turned on by *default*, which is how CR gets its own pool of workers from the mturk worker pool. turning off these settings is shady at best, as it means you manipulated your competition for comparison

[ren @sarinmt]

while i support the sentiment behind this, i will never fund turkopticon as long as it supports the trashing of the reputation of legitimate, well-paying requesters by the multitude of workers who will lie to help theirselves at any cost https://t.co/fE3DsDFgNl

[turkopticon @turkopticon]

NEW: We are launching a #fundraising campaign for Turkopticon, to help Turk workers share info about exploitative practices and fight for stable, dignified working conditions.

Learn more here: https://t.co/tUveZxgDbk
Donate here: https://t.co/LNZwh1suJW

[ren @sarinmt]

cw // gore, open image at your own risk
PSA for NETS213 University of Pennsylvania students

if you're going to post the template batch to mturk, be aware that your image dataset includes pictures of explicit gore, which your professor has failed to filter out

[ren @sarinmt]

Can you also explain why you're paying workers on the platform $3/hr for your work, while maintaining a high rejection rate and potentially ruining people's livelihoods on MTurk? https://t.co/kynkvwnZ3C

[ren @sarinmt]

Hey @DoorDash, can you explain why one of your engineers is exploiting @amazonmturk by rejecting all the work submitted to him so that he gets data for free on the platform?

[ren @sarinmt]

@OScharenborg @laurent_besacie @amazonmturk @aleboz from an mturk worker, i wouldn't let a 7-8 year old paper about mturk completely steer you away from using the platform for your academic studies, as it's still a legitimate method to gather study responses if you learn how to make the most out of it

[ren @sarinmt]

i just wanted to help :(

[ren @sarinmt]

@LaurenGrewal @nirajana_phdmom @JennyGin18 2) reporting the workers to turkprime
if i remember correctly, turkprime will block faulty survey takers from all of their future studies under the platform, so once you start blocking/reporting enough of these 'YES, GOOD, VARY GOOD SURVEY' workers you'll get higher quality data

[ren @sarinmt]

@LaurenGrewal @nirajana_phdmom @JennyGin18 there are a few ways to go about this:

1) using stricter qualifications
you only had 90%+ approval rate and 100 HITs requirements on your last survey, almost everyone will fall under that category

you'll get better results with at least 98% approval rate and 1000+ HITs reqs

[ren @sarinmt]

@sshepherd_phd @LaurenGrewal on mturk, did you consider paying above US federal minimum wage for your studies? would probably help with getting higher quality data on the platform, js https://t.co/DTx7XoINPn

[ren @sarinmt]

@gleemie @amazonmturk not a clear answer, but palani isn't as innocent as you want to believe they are

[ren @sarinmt]

might wanna fix this @CloudResearch_1

[ren @sarinmt]

i also worked on this

it's pretty cool to see that the work i do on mturk can also further academic research https://t.co/b9lz3z0DUw

[Noa Garcia @noagarciad]

Our paper "KnowIT VQA: Answering Knowledge-Based Questions about Videos" has been accepted at AAAI2020! Check out a preprint here https://t.co/PSAFcvlt0H

[ren @sarinmt]

oh hey i was one of the mturk workers that did the work described this paper, glad to see that i could help with furthering NLP research https://t.co/Vivs7hfPAG

[Ben Zhou @BenZhou96]

Check out our new #emnlp2019 paper where we studied temporal commonsense: https://t.co/Z9EESDEvuI. We collected a QA dataset MC-TACO🌮(leaderboard coming soon) and showed that it's a new challenge to existing systems. Co-author with @DanielKhashabi, Qiang Ning and Dan Roth.