Interrupting Invisibilities and Bridging Worlds
Special Mention for the 2013 Hannah Arendt Prize.
The Hannah Arendt Prize in Critical Theory and Creative Research is an annual prize competition for those interested in the juncture of art and creative research and in the principles at the heart of the arts and humanities, including sense-based intelligence; the reality of singular, nonrepeatable phenomena; ethical vision; and consilience between inner and outer, nature and reason, thought and experience, subject and object, self and world.
Interrupting Invisibilities and Bridging Worlds: An Essay on the Work of turkopticon.differenceengines.com
by S. L. Irani-Silberman
Amazon Mechanical Turk works by keeping worlds apart. Mechanical Turk is a web-based labor market that draws workers scattered all over the world to perform small bits of digital labor for a few dollars an hour. Amazon’s engineers and managers designed the system to make up for the failure of artificial intelligence to automate fully their data-processing tasks. Failing to replace low-status workers with machines, Amazon simulated machines with hidden, globally distributed, contingent low-status workers. From outside, the technologists and researchers who work “through” Mechanical Turk see the system as a humming, smoothly functioning infrastructure. Workers are organized, or enframed, for the pleasure of technologists. The system’s smooth functioning—employers’ ability to extract cheap, or, if they like, free, labor from workers—relies on keeping workers isolated—from employers, from journalists, from one another, from the administrators of the system itself. “Requesters” (Amazon’s term for employers) write code to delegate information work to this “crowd,” which is figured and organized as technological rather than human infrastructure, as “artificial artificial intelligence,” in Amazon’s cheeky but truthful tagline for Mechanical Turk. Workers do not see their employers, the projects to which they contribute, or one another, and cannot be found by journalists, labor organizers, or watchdogs.
Turkopticon creates a short circuit between these worlds. Workers call their employers to account and engage in mutual aid. Turkopticon offers a forum and format for workers to review employers and to see others’ reviews while browsing Mechanical Turk for work. Unfavorable reviews on Turkopticon have prompted more than a few employers to wonder why their tasks are not being completed—and, eventually, to engage with workers through Turkopticon and other online venues. We have maintained Turkopticon for four years; around 13,000 users have downloaded the browser add-ons since we began counting in August 2012. The website that lets people post and browse employer reviews receives about 50,000 visits a month. About 17,500 people have signed up since 2008. At time of writing, Turkopticon has hosted reviews on the vast majority of employers in the system. We’ve grafted ourselves into Amazon’s infrastructure.
As Turkopticon’s builders, we “intervene” in the scientific discourse on “human computation.” Computer scientists tend to figure workers either as rational self-interest maximizers or as computing machines without needs, desires, or rights. Turkopticon gives concrete voice to the persistently problematic residual of rational-actor models of economic and social life that still dominate the thin and perhaps overly “pragmatic” economic and moral discourse of high technology industries.
In these models, parties transact freely to mutually improve their lots. But how did technologists arrive at a configuration in which workers are better off spending 30 minutes on a menial information task—a task for which they may not be paid, under a labor regime in which they have no rights, no recourse to arbitration in case of nonpayment, and no guarantee that the employer or a system administrator will even read their complaints—rather than looking for other work? What happened before the transaction, in the design of Mechanical Turk itself? In the slow emergence of a libertarian imaginary in the technology industry? In the erosion of the American welfare state? Such questions are largely outside technology discourse, but need not be.
We delivered a talk about Turkopticon at the premier meeting of the human-computer interaction research community to hundreds of computer scientists. This created an opening for others—some computer scientists and industry researchers—to express discomfort with “crowdsourcing” arrangements at the microphone and in the hallways. We present at crowdsourcing industry meetups, standing in the same room as, indeed smiling at, the executives in charge of Mechanical Turk; we find allies among crowdsourcing startup founders in San Francisco and full-time crowdworkers in rusted-out midwestern cities. By publishing in technology research venues and maintaining a working technological system, our work at first seems legible to the community we critique; the work then unfolds in unfamiliar, jarring, disobedient directions, forcing new terms into the world of technology research.
To bridge these worlds—between workers and employers, between techno-libertarianism and other ethical visions, between research and activism—we have, like many others, become hybrid. Our original training was in computer science and engineering; we have branched out to study, to rely on, and be motivated by, feminist theory and social theory broadly; we think “ecologically,” keeping the scope of analysis open where engineers often cut. We have become institutionally hybrid; while we have fed our bellies mostly by the graces of the academy, for two years the maintenance of Turkopticon was supported in part by a quasi-fictional “Bureau of Economic Interpretation”—sleeping on rooftops while attending computing conferences in American cities. We have been engineer and theorist, academic and activist, nomad and lab rat, peacemaker and provocateur, margin and center.
If our interventions have an ancient mascot, it might not be Prometheus, but rather Coyote. When a system becomes too orderly, these transcategorical figures appear, and their transaction across boundaries transfigures those boundaries. We should not mistake “intervention” as an act taken from above, below, or otherwise outside—from a pure position with pure intentions and pure methods. Coyote is a messenger, always between; Coyote “is the spirit of the road at dusk, the one that runs from one town to another and belongs to neither.”  In the early days of Turkopticon, we spread word of our service through a personal connection with a prolific requester. For Coyote there is no outside, no purity, no starting over with the ground swept clean, no unimpeachable subject position, unassailable argument, or finally decisive tactic. As Coyote, we act not from an outside, or even on something outside ourselves, but rather enact situated reconfigurations  in constantly shifting material-semiotic assemblages. Coyote unsettles the existing order not by fighting it on its own terms but by acting in ways that challenge the legitimacy of its categories; by changing the variables salient to the operation of the dynamical system; by changing, in short, the rules of the game. Coyote’s interventions expand “the adjacent possible”; they “multiply possible worlds.”  Yet despite bridging the worlds—in the mythology, the worlds of gods, humans, machines, animals—Coyote is always particular, embodied, finite; Coyote always has a belly to feed.
For four years, Turkopticon has helped disrupt the invisibility, isolation, and silence of Amazon’s distributed workforce. It has altered the balance of power between workers and employers. It has publicly borne witness to worker unease—witnessing carried further by journalists who have used Turkopticon as a starting point to push these labor questions out and by other researchers (one of whom now works for the Department of Labor) who have built on our work.
What now? As our web service has grown more popular, it has also become slow, occasionally impairing its usability. Workers want new features, and we often fail to deliver. We have become part of the taken-for-granted machinery “around” Mechanical Turk. The categories are starting to ossify in the database and in the discourse. When asked why they don’t fix their broken system, the executives, pointing among others to us, say that “the community” handles the problems. Turkopticon began as a gesture, but like any organism in a coevolving system, we have to run as fast as we can just to stay in the same place.
We have not called ourselves artists; we have not needed to do so. In our work, we have found maintenance—of social relationships and computer code alike—to be as essential as intervention or (that term celebrated so widely it risks becoming emptied of meaning) “innovation.” But we might ask if the spirit of intervention today—in art, in activism, in creative research—is embodied precisely in the act of imagining and making possible “otherwise.” Bounded agencies still slip the trap of the channels carved out for it by existing institutions, infrastructures, norms, and practices; the difference that makes a difference lies in the development and exercise of the disobedient imagination. Opportunities for this exercise turn up without fail. As Goodman wrote of anarchism, the relativity of critical and creative practice to the actual situation is part of its essence; like the anarchists, like Coyote, “we are not in charge of the world”—“we just live here and try to strike up non-innocent conversations.” 
1. Lewis Hyde, Trickster Makes This World: Mischief, Myth, and Art (New York: Farrar, Straus, and Giroux, 2010), 6. ↩
2. We mesh terms from the works of Donna Haraway, Karen Barad, and Lucy Suchman. ↩
3. Stuart A. Kauffman, Investigations (Oxford, UK: Oxford University Press, 2002), 244; Michel Callon, “What Does It Mean to Say That Economics Is Performative?” in Do Economists Make Markets? On the Performativity of Economics, eds., Donald MacKenzie, Fabien Muniesa, and Lucia Siu (Princeton, NJ: Princeton University Press, 2008), 352. ↩
4. Donna Haraway, “Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective,” in Simians, Cyborgs, and Women: The Reinvention of Nature (New York: Routledge, 1991, 199. ↩
About the MA in Critical Theory and Creative Research
The MA in Critical Theory and Creative Research (CT+CR), the first of its kind in the U.S., is an accelerated, 45-credit, seminar-based program (one year + summer intensive) that prepares students for opportunities at the intersection of art, theory, and research. Located in the metropolitan heart of the Pacific Northwest, a center of creative risk-taking and social experimentation, the program combines the study of critical theory as a mode of socio-political critique and creative research as a process-driven form of inquiry, pushing both theory and research in new directions within the context of a 21st-century art school. The program is devoted to people and ideas and to a rethinking of the present and future of cultural production; of arts-based research and research-based arts; of curatorial practice, documentary, and the Archive; and of social and political reconfiguration in relation to major sites of contemporary contestation. See additional information on the MA CT+CR Program page. Founding Co-chairs: Anne Marie Oliver and Barry Sanders.