What is the haiku author? Adventures with Mechanical Turk

I’ve been making some excursions into the land of Amazon’s Mechanical Turk, where many thousands of anonymous workers with IDs like A30K1ZD4E07JX do cognitive piecework (“Human Intelligence Tasks”) for several cents a “HIT,” making what amounts to a dollar or two an hour. Mechanical Turk once instance of crowdsourcing, getting things done by employing the often low-paid enthusiasms of large crowds.

I’ll write more about the labor politics of Mechanical Turk later. I’m still exploring that. But one of my explorations in qualitative engagement with people I can’t meet has been documented on a blog I created Haiku Turk. I put up calls for a penny, ten cents, and fifty cents and asked people for a haiku. I wanted a parsimonious, fun way to make a human connection through the anonymizing web interface. In the last batch, I decided to intervene in the anonymity by offering people the opportunity to list their IDs and pick a nom d’plume.

Some of the strangely delightful haikus submitted include:

McCain picked Palin
Palin is not good for us
McCain, do not die [permalink]

graveyard shift provides
rare opportunities for
paid masturbation [permalink]

To learn more about the conditions of working in Mechanical Turk, you can check out my other engagement, Turk Work: Bills of Rights. It’s not very synthesized, but it’s there to wander through to hear different workers’ ambivalences. Unscrupulous work requesters who won’t pay up are a huge complaint, as is the very low pay. But some also value the flexibility of at-home, variable time labor they can do as a form of play or extra cash whenever they have the time.

I am plotting an intervention. More news on that later.

Category: Uncategorized  Tags: , , ,
You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.
7 Responses
  1. zelda says:

    Though I’ve been aware of MTurk, I haven’t been paying close attention, so I appreciate the ways you’re thinking about it, Lilly. I only skimmed the Bill of Rights blog, but as so often happens with net-based argumentation, the main thing that struck me was how several hundred years of ferment gets regurgitated in a compressed time span (on the blog, one day). The result is a kind of discourse mashup– on the Bill of Rights blog, for instance, one finds all these hot points in the history of labor politics interwoven in endless variations:
    –piecework is great, don’t rock the boat
    –fair pay for work
    –worker’s rights
    –minimum wage
    –unionization
    –ownership of intellectual labor

    What I don’t see, interestingly, is any real luddist spirit–the anger that is expressed by the workers over their own exploitation is strangely muted. I wonder what’s mainly at work– fatalism? Low expectations? Low stakes (hardly worth taking to the streets for 1 penny an hour… yet)? Belief that things will somehow work out? Something else?

    What, in any case, would be the equivalent here of breaking the machines to get the attention of the bosses?

  2. it’s a really weird labor market — the fact the requester can arbitrarily choose to deny payment seems a bit off.

    for a lighter perspective: http://waxy.org/2008/11/the_faces_of_mechanical_turk/

  3. Lilly says:

    Thanks, Brendan! I really liked the photo project you linked to.

  4. six says:

    It’s also interesting that we have an AI researcher who seemed reassured by the “Faces of” project:

    “As a researcher I’ve been wanting to use Mechanical Turk for some of my machine learning tasks. This clears away the fog about who are the real turkers and reduces my anxiety about using it.”

    I wonder why?

  5. zelda says:

    After checking out the waxy.org website, I’m convinced that MTurk is going to unloose a flood of pseudo-research on anything you can easily study by posting it as an MTurk job. With typical MTurk job results numbering in the dozens to hundreds (depending on pay scale), it’s more ‘scientific’ than the usual polled-my-friends-and-family pontificating that media pundits love to indulge in but falls way short of the kind of numbers out of which significant patterns might emerge. (Not even to mention the problem of well designed studies, controls, etc.)

  6. Lilly says:

    Yeah, totally. I would love to see someone actually do research on how the amount of pay and the time of day you post affects the kinds of responses you get. And then, of course, there’s the already pretty extensively discussed limits of even well executed surveys that many pundits probably don’t know about. Waxy.org’s study, for example, doesn’t describe or critically assess their methods.

    There’s also this weird sense I have that where journals and publishing are supposed to stand for some sort of generic societal witnessing as a the standard of what it counts to make knowledge, there’s this sense of measuring increasingly mediated bodies as a way of building your claims on pseudo-experiments. Then again, maybe it’s just one step away from using census data someone else collected. But at least those methods are interrogatable.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>