Labor 2.0 – The Digital Sweatshops, pt 1

Lately, I’ve tended to find myself well behind the curve in most things – a far cry from my former life on the bleeding edge – so it’s not really surprising that I only just discovered Amazon’s Mechanical Turk and the wonderful word of crowdsourcing. The term “crowdsourcing” derives from “outsourcing” and it refers to the process of farming out work, or outsourcing, to a crowd of people, most commonly over the internet. (See Wikipedia for a more detailed discussion.)

The term “crowdsourcing” is relatively new, having been coined in 2006 by Jeff Howe in a Wired magazine article, but the concept – particularly as it works on the internet – has been around a lot longer. For example, CDDB (CD database) started life as what we might now recognize as a crowdsouce application, with users entering data on each new CD into the database. In the beginning, CDDB was free to access and use, and many people were angered when it went commercial – essentially taken a lot of volunteer labor and turning it into a private commodity. (Again, see Wikipedia for a more detailed discussion.)

It was  a sort of digital enclosure act, by which a formerly free and common property, worked in common by a group of people as a general good, was expropriated and sealed off – enclosed – as private property. A similar process occurred with IMDb (the Internet Movie Database), which began life as the communal effort of the people on the rec.arts.movies Usenet newsgroup (more here) before ultimately ending up in the clutches of

Mechanical Turk takes the kind of micro-entry work process that built up the databases that fuel CDDB (aka Gracenote) and IMDb and turns it into a business model. “Requesters” take tasks that are needed for their computer application but can’t be done by computer and turn them into bite size chunks which are posted on Mechanical Turk as HITs (Human Intelligence Tasks). Mechanical Turk workers log on to the system, accept one of these HITs, complete the task and submit the result. The pay for a completed HIT ranges from 1 cent on up – currently, the largest amount paid for any single HIT is $5.09.  The vast majority of them are far towards the lower end of that scale. Of the 1647 HITs available now as I write this, 1387 of them (84%) pay 50 cents or less.

There are a large variety of tasks, but a few trends are clear. Tagging is a common task at the very lowest end of the scale – looking at an image, reading a phrase or paragraph, watching a video or listening to some audio and then quickly assigning a few tags or categories to it. Transcribing is one of the more highly paid HITs, and one company, CastingWords, has built its entire business on the work of “Mechanical Turkers” – taking in transcribing tasks from a range of companies and then farming then out to the army of turkers, first for transcription and then again for double-checking and editing.

A large number of HITs in the mid-range of pay involve content production for websites, generally in the form of small chunks of text. For example, one asks you to write a 250 word review of an indie video game, and then provide categorizing info on the game (platform, price, URL, etc.), for a payment of $1.00. Another asks you “share your room painting project” by submitting a photo of the room and writing 300 words on how you painted it, what brand and color paint you used, etc. – for $1.00. Yet another hit paying $1 asks for 450-550 words on a home improvement topic – with one of the topics being “home improvement tax credit 2009 stimulus.” A travel website asks you to create an account on its site, log in, and post a travel guide for a destination, not in the United States, that includes entries for at least 5 specific places or things in that destination – for the relatively princely sum (by Mechanical Turk standards) of $1.50.

Obviously, this work is horrendously underpaid by the standards of any developed country. And while there is the suggestion that people in less developed countries might be logging on and doing HITs, the vast majority of users of Mechancial Turk seem to be from the United States, and indeed many of the HITs require you to be in the United States. And Mechanical Turk will only pay you through an Amazon gift certificate or payment into a US bank account, and so is of limited usefulness for people overseas actually trying to support themselves.

And for anyone, anywhere it is a pretty severe form of alienated labor and a digital version of the sweatshop (sadly not old for  You don’t really know who you are working for. The work is mostly mindless little chunks that require not real human intelligence, but what Amazon likes to call artificial artificial intelligence. Computers still can’t reliably tell that a picture of a shoe is a picture of a shoe, so a human is needed to add a tag for the image database. Exploitation is easy since the Requester can take the work but then “reject” it in the system and so avoid having to pay.

A noticeable percentage of the HITs are essential scams – requiring you, for example, to go to a website, fill out a questionnaire, give your mobile number and then respond to an SMS message. All for an ostensible pay of $4. The HIT says the task is to “verify that this system is functional,” but in fact it seems to be just a way of scamming SMS fees.

Most damningly, the rate of pay is so absurdly low. It would be impossible to earn minimum wage (for anywhere in the US) working on Mechanical Turk. But I suspect, with unemployment in California now over 12%, there are people (like me) willing to do anything just to keep a little money trickling in, and I wouldn’t be surprised to find that 2009 and 2010 end up being boom years on Mechanical Turk. It would be a sensible business man who invested some capital into exploitation of this desperate pool of workers now, and used the content generated to take off in a big way when the economy picks up and people will pay for whatever he has built up during these lean times.

I have more to say on this… and I suspect I will be bringing class into it, and trying to demonstrate the violence inherent in the system

For More…

And on related topics…


4 responses to “Labor 2.0 – The Digital Sweatshops, pt 1

  1. Hi,

    I really like your article on Labor 2.0 – The Digital Sweatshops.

    Your writing on Amazon’s Mechanical Turk HITs (Human Intelligence Tasks) becoming a digital sweatshop is truly insightful.

    Looking forward to more great postings from you.


  2. Amazon’s Mechanical Turk is a SWEATSHOP OPERATION. A class action lawsuit needs to be filed agaist them for violating U.S. Fair Labor/Minimum Wage laws. They take advantage of people desperate for a job.


  3. Contrary to popular belief, the majority of people making pennies an hour on Amazon Mechanical Turk live in the USA. I eagerly kook forward to Amazon LOSING multi-millions for violations of USA Labor Law. They aren’t the first corporation to get away with gross violations of our labor laws for a long period of time before being caught and penalized.


  4. Visit “Turker Nation” forum before you work at Amazon Mechanical Turk.

    People work for hours on end at Amazon Turk for way less than minimum wage; often they aren’t even paid for work they do because Amazon permits requestors to refuse to pay.

    Amazon should be smacked with a class action for permitting the mis-classification of workers, as should Dolores Labs, (Crowdflower) Retail Data and others.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s