Jonathan Zittrain speaks at the 2010 European Zeitgeist conference about the future of innovation as observed by trends in crowdsourcing and cloud labor. Entertaining but ultimately unsettling. It's 21 minutes long but is thought-provoking enough to deserve a view.
I decided to try Amazon's Mechanical Turk for the first time Tuesday after reading about it for years in places like Wired. It's a crowdsourcing site for micro-jobs: a place where anyone in the world can go to earn small amounts of money for accepting HITs (Human Intelligence Tasks) of microwork posted by requesters, who pay 10% over the amount they agree to pay. Payments go directly into the Turker's Amazon Payments account and can be rolled into one's own bank account. Payments can be anywhere from $0.00 (yes, believe it or not, people will work for nothing) to nearly $40 for long-form transcription, which requires a high level of skill. The vast majority of HITs pay in the single-penny range.
Right off, I decided the lowest-paying jobs were not worth my time. I sorted by Highest Reward first and found several jobs requiring enough skill that they would actually pay some money. Many of the higher-paying jobs have a qualification that must be completed first (writing ability, accuracy in transcription, etc.) I earned about $19.00 yesterday in 2.5 hours' worth of work, which works out to about minimum wage, doing things like rewriting short articles, completing short audio transcription jobs, and (just for fun) completing a side-by-side image survey on things like whether Miley Cyrus's face is more chubby/shiny/etc. than A-Rod's. That last one only earned me 3 cents but I wanted to see what the crazy-low-paying jobs were like.
The most fun job I did took four minutes and paid 50 cents: a requester wrote in asking for advice on what color to paint the area in their dining room below the chair rail. They'd painted the upper part of the wall "Fennel Seed" by Benjamin Moore. I looked up the color and gave them my opinion as well as three different colors I thought would work well with it. They replied "Great Response!!!" and even gave me a 50-cent bonus. That was a dollar happily earned in 4 minutes, which would work out to $15 an hour if I could do that all day.
It seemed innocent enough, but then I started to think. I did one HIT where I was asked to find a product on Amazon, write a review, and submit it. I knew nothing about the product so had to go by what was on the Amazon page. There were several high-paying HITs that looked quite dubious: filling out a profile on a dating site (in effect, putting yourself--complete with your picture--on a dating site for $8.00), completing credit profiles, coming up with descriptive terms for pornographic images, and so forth. Mechanical Turk is entirely anonymous: anyone can request work, and the workers remain unknown to the requesters. I started to wonder if all this work was entirely ethical. It was starting to look like some of it was completely fine, but other HITs definitely looked shady.
The site has drawn hundreds of thousands of workers. I would like to do more research on the typical "turker" - I'm coming late to things other people have been doing for years and want to catch up. I have heard that a large number of the 200,000 using MT at any given time are from India, and that they not only produce good work, but are willing to take the lower-paying HITs since the standard of living there is so different. I found this astonishing lecture by Harvard law professor Jonathan Zittrain, where he explores the potential and actual downsides, ethical, legal, and otherwise, of crowdsourcing work in this way.
It is absolutely fascinating what technology permits us to do, but just because we can do things, doesn't mean we always should. The video is making me think hard about Mechanical Turk and other types of cloud labor - particularly the way they're being exploited - and about their implications for the future of work. It's also made me think very carefully about the MT HITs I accept in the future. I think I'll stick to transcription.
No comments:
Post a Comment