Tweaking Technocapitalism: Turkopticon

I’ve posted about Turkopticon here before. Well, it’s up, it has undergone a rev, and it has some users we don’t know who seem to like us. I wanted to talk a little bit about what is at stake in it.

For a long time, I’ve been thining about infrastructure and technology design and, in particular, how certain designs (in certain contexts) end up giving certain people the crap end of the stick. As of late, my friend Six and I have been spending our spare nerd cycles on a particular case of this: Amazon Mechanical Turk, which lets workers do cognitive piecework usually averaging a dollar or two an hour. The low wages, the lack of health protections in a “work environment” (the computer) that has caused my arms and wrists much pain over the years, and the exuberant excitement many have for getting the faceless “crowd” to do work so cheaply were my initial cause for concern. As I started to survey Turk workers about their experiences, workers reported little protection from employers who don’t pay and low wages as big problems. I heard from workers who did Turk after their main jobs to make food and rent when gas prices were high. While I don’t have the power to regulate AMT or radically shift market dynamics at the moment, Six and I put our heads to the first problem of employers who take people’s work and then don’t pay.

So we made Turkopticon, a Firefox extension workers can use to access ratings and commentary of employers/requesters as they browse for HITs (“human intelligence tasks” and an unfortunate acronym). Turkopticon isn’t revolution — it’s not going to fix the fact that jobs are increasingly contingent, that health care costs are insane, and people have fewer good choices about how to make their livelihoods. But it’s a start at drawing attention to an information imbalance that has been letting some requesters abuse people. It’s something that can make us ask why Amazon didn’t design these informational safeguards in to begin with. And lest we think the traditional lines of employer v worker are simply drawn, Dolores Labs provided critical support and feedback. We started off as an empty database asking workers to install our extension, but there wasn’t much for workers to see. Dolores Labs put up a survey for us and got a hundred or so reviews of requesters that formed the seed of the database, motivated in part by their desire to resist Turk being spoiled by crappy employers. (I’ll probably post most about this in future posts.)

Is it just about Mechanical Turk for me? Not really. I see AMT as an dystopian extreme case of a the increasingly contingent, low paid labor I’ve been seeing creeping up for years.

Jobs aren’t a great way to make a living these days. A few trends that disturb me. The practice of hiring temp workers on a mostly permanent basis so that they can be denied health benefits and other perks took Microsoft to court and even got its own neologism: permatemps. The largest employer in 2/3 of US states, Walmart, pays barely enough for a full-timer to make ends meet, claiming to only provide “supplemental income.” About half of a those filing for bankruptcy in a 2005 study cited medical debts as a main cause [pdf source]. Livelihoods are precarious for a lot of hard working people.

People frequently argue that those working for these low wages have a choice. As one person I corresponded with explained, “I realize I have a choice to work or not work on AMT, but that means I would also not need to make the choice to eat or not eat, pay bills or not pay bills, etc.” The thing we need to worry about is not only what choices people make, but what choices people have. Not all jobs are available everywhere. Not all people are equally able to move. Not everyone can afford a solid educational foundation. Not everyone even gets their knowledge and wisdom equally recognized and respected. People do have choices, but some have more choices than others.

Turkopticon is just a little Firefox extension, but for Six and I, it’s also forcing us to think about a lot of issues in labor and politics that we just don’t know enough about, but which have consequences around us every day.

Thanks to Dolores Labs, the 67 turkers who shared their experiences, and those who have been using Turkopticon and reviewing already.

You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.
9 Responses
  1. techlist says:

    Great work! Please contact us about collaboration.

  2. Lilly says:

    85% of Turk workers are in the United States so that is low pay — not even close to a living wage.

    For the workers who live outside of the US, what makes for a good livelihood is much more complex. Even if $2 / hour is higher than average wages, there’s a difference between a job where somebody learns skills that they can use to support their own livelihoods and a job where they do work that is intentionally fragmented to the point that they don’t even know what they’re contributing to. When wages somewhere go up, the turk requesters don’t raise their prices because a bunch of lower paid people are their to do the task cheap. For those who had been making ends meet with AMT or other forms precarious labor, they now have to figure something else out.

    AMT seems like the sort of work that doesn’t develop local resources and infrastructure and can leave any second. Very precarious.

  3. johndburger says:

    Apologies if this have already been bandied about on the site, but:

    Why the implicit assumption that this is indeed low-paying work? “Low-paying” is, of course, a relative term—there are plenty of places in the world where $2 per hour places you above the median wage, yes? It would be interesting to determine what the percentage of Turkers is for whom this is the case. If it is sufficiently high, than AMT might well be RAISING the mean, and thus be a force for (minor) economic good.

    Again, apologies if this is well-trodden ground.

  4. Lilly says:

    Drew, people have a lot of time to play spider solitaire, but people don’t play spider solitaire to make ends meet. But something like 50% of people doing AMT hits are doing it for income, and we’re not talking about pocket change income since the person who did the survey had that as a separate category.

    While ways of building market power as a requester are unclear, workers have a lot less traction because there are many more people having trouble making ends meet than there are employers — especially in this economy. If the labor market is flooded, wages go to zero. In this economy (in the US, where the majority of turk workers are currently from), many might not have jobs to go to in place of AMT.

    Zelda’s insight about being informed about risk gives me great inspiration. I’m feel like Turkopticon is very limited by the fact that it remedies one information asymmetry but doesn’t deal with the risks involved in permatemping — and I bet there were a lot of people not thinking about catastrophic financial events like health problems pushed them into bankruptcy. The indeterminacy and social construction of risk is a big part of it. Relatedly, precarity is another — a word Six turned me onto a few days ago.

    I wonder how risk is being constructed in people’s daily lives. Why is the news pushing false facts about sexual predators attacking 5 of 7 kids on the internet where the lack of health care and precarious wage conditions are concretely endangering a great many communities? Sensationalism seems like the opposite of creating consciousness of common issues. On the pharmaceutical front, Joe Dumit (history of science scholar) is doing a project looking at how drug companies are moving from products that cure to products that reduce risk since you can never be risk free. I wonder if a strange side effect will be a popularization of risk culture that changes some of these labor debates.

  5. zelda says:

    “It’s something that can make us ask why Amazon didn’t design these informational safeguards in to begin with.”

    Especially since these safeguards are a notable feature of the success of ebay, not an untried mechanism…

    “The thing we need to worry about is not only what choices people make, but what choices people have.”

    So true. In my early twenties I spent several years working as a permatemp–in my case, as an ‘independent contractor’, meaning I was paid a flat fee per hour with zero benefits. In theory, I could have paid for those benefits myself, since I was self-employed and thus technically the owner of my own business (with one employee–me). In practice I got nothing–no health insurance, no savings, no pension plan, no paid vacation/sick leave–because I didn’t earn enough to afford any of those things. (At the same time, being the employer, I did have to pay extra Social Security tax.) In at least two of the places where I worked as an independent contractor, it was clear that this route was being taken to cut costs and that the business was skirting legality to do so.

    The reason I bring this up is because I was aware even then that working this way was a matter both of having the choice to do so and of how I perceived the riskiness of my choice. Bluntly, I was playing the odds and I got lucky– I didn’t go bankrupt as a result of some serious misfortune. What galls me as I think about this now in the larger context of American culture is the degree to which people take such risks at least in part *because they are misinformed about risk in general*. I know I was pretty clueless at the time. It’s likely that people everywhere want to believe they are lucky, but it’s also true that the less you educate people about risk in general, especially real risks (no health insurance) vs. perceived risks (serial killers), the more likely they are to make stupid gambles. Or to support politicians and others who want to help them make stupid gambles (balloon ARMs, privatizing Social Security). I got a lot out of Peter Bernstein’s book called _Against the Gods: The Remarkable Story of Risk_, but I am no expert in this area– anyone have other reading recommendations?

  6. Drew says:

    Turkopticon is a great idea and should be built into the AMT system. There should be ebay-like reputation mechanisms on both sides.

    While it might be troubling that people choose to do “HITs” on AMT despite the low pay, the question is what to do about it. AMT could enforce a minimum wage (which would likely lead to HITs becoming more difficult) or AMT could be shut down (which would deprive a bunch of people of work that they apparently thought was worth doing). Neither seems very desirable.

    As you point out, it’s about much more than AMT — it’s about the options people have that lead them to AMT in the first place. For many workers on AMT, the next leading option is playing games. (That’s what Panos’s survey says.) According to Luis von Ahn, 9 billion hours were spent playing solitaire in 2003. If we’re going to wring our hands about people performing rote tasks for low pay on AMT, what do we think about people spending that much time performing rote tasks for _nothing_ on Spider Solitaire?

    Ultimately AMT is a labor market in the purest form. There is no way for workers to organize, but there is also not really much opportunity for building market power as a requester. The right thing to do is to remove information asymmetries, as you guys are doing with Turkopticon, and let’s hope that as the market develops it will become easier for serious workers to find rewarding jobs on AMT.

  7. Matt, very interesting re: elance and rentacoder. I think part of the issue is that AMT is just much less developed/mature than those sites.

  8. Matt says:

    Wow– awesome post, and awesome extension. It’s crazy that there aren’t built-in protections of this sort.

    I’ve been using elance, rentacoder, etc. to find subcontractors. (The best case with those sites, AFAICT, is to find someone to work with over the long term, using it like an interview process.) I’ve heard complaints that they’re employer-biased, but they do all require placing funds in escrow before the work starts. There are arbitrators in case there are disagreements about payment. Does Amazon not require funds to be placed into escrow? Or is the problem with their conflict resolution?

    I’m not surprised that the legitimate employers would endorse Turkopticon. Scammers simply drive potential turkers away.

    Re: permatemping, e.g. Walmart, that simply strikes me as a “smart” (from a game standpoint) exploitation of flaws in the labor code. With the US health care system bizarrely connected to full-time employment (so much for encouraging enterpreneurship), power shifts even further to the employer. You’re going to see many of these exploits emerge from the largest corporations, because they’re the ones who can afford to invest in the exploits, and make the money back over their large number of employees.