« Tucows Not Liable for Spam Judgment Against 3rd Party Based on Private Registration Services -- Balsam v. Tucows | Main | Court: Prosecutors Can't Rummage Around in a Defendant's Gmail Account -- U.S. v. Cioffi »
October 27, 2009
Zittrain on the Dark Sides of Crowdsourcing
By Eric Goldman
Last week, Cyberlaw expert/rock star Jonathan Zittrain of Harvard Law School (visiting at Stanford Law School this term) spoke as part of SCU's lecture series on IT, Ethics and Law. An overflow crowd of over 100 people came to the talk, and as usual Jonathan did a wonderful job. You should attend his talks if you have a chance. They are always highly entertaining and thought-provoking.
Jonathan's talk was opaquely titled "Minds for Sale: Ubiquitous Human Computing and the Future of the Internet." I think his talk really covered the Dark Sides of Crowdsourcing. Normally, such a talk would immediately raise red flags about the speaker’s intentions and net-savviness. However, Jonathan has sterling credentials as a crowdsourcing enthusiast, so he can raise concerns without sounding shrill or regressive.
He pointed to numerous examples throughout his talk, but one of his principal targets is Amazon.com's Mechanical Turk. Mechanical Turk describes itself as "a marketplace of work. We give businesses and developers access to an on-demand, scalable workforce. Workers select from thousands of tasks and work whenever it's convenient." LiveOps is also on his mind; the site hires individuals to provide live customer support for third party businesses from their homes on a flexible and dynamic basis.
From the worker's perspective, Jonathan raised several potential concerns about Internet marketplaces for labor, including:
* surveillance. LiveOps can track and monitor every aspect of workers’ performance, including when they were online and where they connected from. This does depend on the employer’s specific technological architecture; I’m not sure if Mechanical Turk gets such deep looks into its workers' behavior.
* alienation. Online workers are often retained to do pieces of a larger project without understanding the whole project. Thus, they may apply existing rote skills but not broaden their expertise.
* moral valence. Because workers may not understand the greater project, they could help projects that are morally objectionable without realizing it. Jonathan gave some terrific examples, such as spammers distributing the work of breaking CAPTCHAs and the government asking citizens to identify folks in group/crowd photos for law enforcement or possibly dissident-busting purposes.
Jonathan also raised some systematic concerns that could arise from these online labor marketplaces, including:
* without uniform labor laws, employers could trigger a race to the bottom where the work flows to jurisdictions with the laxest worker protections. We’ve already seen an analogous situation as states try to enact “Amazon affiliate sales tax” laws designed to trigger sales tax collection obligations based on the presence of marketing affiliates in the state, which has prompted some Internet companies to dump affiliates overboard in some of those states based solely on their regulatory policy.
* crowding out. As it becomes easier for workers to monetize their time in smaller units, they may become less willing to contribute their labor to non-paying enterprises like Wikipedia. I address some of these dynamics in my Wikipedia paper.
Jonathan identified some possible solutions, including:
* revitalized labor standards, such as minimum wage laws or anti-child labor laws. As one example, he proposed that worker reputation should be “portable” so that good workers for an online enterprise can have their accomplishments follow them to future employers. Ironically, this may be unintended Internet exceptionalism because IMO worker reputation isn’t portable in physical space due to the collapse of the job reference market.
* unionize online workers. Unionization was a natural proposal given the talk’s tenor, but I found it anachronistic. How can unions be relevant to online labor markets where people can "change jobs" with a few clicks of a mouse? The Internet has much more competition among employers than any geographically restricted labor market, and the employees’ friction to change jobs online is so much lower than it is in physical space. For example, unhappy Mechanical Turkers can easily click to a large number of competitive websites that will gladly pay them for their contributions. At the same time, the Internet globalizes labor forces in ways that are unprecedented in physical space, so the value for “commodity” labor plummets. Both dynamics seem to doom any efforts to unionize online labor.
* disclosure. Employers should have to disclose who they are and why they are asking for the work to be done.
* opt-out. When people are doing work without knowing it, they should have the opportunity to opt-out. For example, the RECAPTCHA project uses human CAPTCHA-solving to correct OCR scanning errors. Zittrain thinks people should be given a choice not to provide those services.
Jonathan's final takeaway message was to express a general reservation that money is pervading our relationships and activities. He gave the example of how a Mechanical Turk employer offered to pay other Turkers to do kind acts--something, Jonathan pointed out, is oxymoronic because paying people to be kind means they aren’t actually being “kind.”
Whether intended or not, Jonathan’s talk had a strong Marxist undercurrent that is tough for many of us to embrace. The Internet makes labor markets more efficient. It also increases the heterogeneity of ways that people can find gainful employment they can perform at the time and place of their choosing. Both generally sound like strongly positive developments to me.
I thought Jonathan’s strongest point was his dystopian view of bad actors (e.g., repressive governments and spammers) crowdsourcing socially detrimental work without workers knowing it. Disclosing the employer’s identity and motivations would be a partial but necessarily incomplete solution. More transparency would prevent people from inadvertently contributing to bad projects, but some people need cash so desperately that they will take it regardless of the moral valence. (I'm ignoring the malcontents who would gladly pay for the chance to facilitate social disruption).
Jonathan’s talk became harder to follow as he addressed examples well beyond online labor marketplaces. Specifically, a number of his examples seemed to conflate work activities, play activities and other ways that people voluntarily allocate their time. (FWIW, I’ve been accused by Timothy B. Lee of doing the same thing in my Wikipedia paper). For example, he treated Google’s use of website links for its PageRank algorithm as a form of “work” where Google gets the benefit of individual linking decisions. (In turn, he lauded Google’s nofollow link as a good example of how laborers can have an opt-out mechanism). While we collectively create value for Google by establishing links to third party sites, I don’t see how linkers are “working” for Google. Instead, I think the links are a positive externality captured by Google. We create positive externalities through our online (and offline) activities all of the time, and I think trying to characterize those as “work” is not the right direction. See, e.g., Frischmann and Lemley’s critique of regulatory overresponses to spillovers.
Similarly, Jonathan gave examples of “games” where the players provide valuable outputs to the game organizers through their ordinary gameplay. An example is where people are asked to tag photos as part of a game, where the tagging can become commercially valuable metadata. In this situation, the game organizers get an undisclosed private benefit (the beneficial work) from what is otherwise a fair market transaction (people voluntarily enjoying the game). Normally, we don’t worry about undisclosed private benefits; in fact, they occur in most economic transactions, even in efficient marketplaces. While increased disclosure about the game organizer’s motivations might help game players make more informed decisions about whether to play the game or how to price their participation, it’s not clear to me that such disclosures would actually help the game player’s make better decisions or enjoy the game more.
Terri Griffith, a colleague of mine in the SCU business school, wrote up her perspectives on the talk.
UPDATE: Mike Sardina, an SCU Law student, also wrote up a recap with commentary.
TrackBack URL for this entry: