New Criticals

Digital Labor’s Blackboxed Supervisors

Two YouTube vloggers live in the United States, earning a substantial portion of their yearly incomes from their platform-distributed content. One sits in front of her laptop’s built-in webcam, spinning tales of awkward Tinder dates. 5,000 sets of eyeballs watch an advertisement and then watch her video. Another 10,000 do the same. The next day, she may reasonably expect to see a tally of 15,000 views, earning between $1 and $13 per each thousand of these views ($15-$195)[ii]. The other vlogger stands in his basement in front of a green screen and explains how to create a functional air cannon or “potato gun” to a large camera mounted on a sturdy tripod. Through this video, he produces 10,000 views in a day, then another 500 the next day. He may reasonably expect to see a tally of 10,500 views the next time he “deep dives” into data calculated and displayed by YouTube’s content management system (CMS). Curiously, the woman may instead find 12,000 or 17,000 views and the man may find that only 9,000 views actually counted in the way that really counts, the way that earns money. They work for a piece-rate, yet both the piece and the rate remain mostly unknowable and, often, unpredictable.

In this essay, I want to highlight platforms’ blackboxed inner-workings as a labor issue. This follows on the heels of rather well-developed discussions of the political dimensions of platforms wherein algorithms and incentive structures mediate, alter, or deform the public sphere [iii]. Likewise, I want to emphasize that the heightened “media participation” that platforms enable is, at the same time, an issue of labor. Recent discussions of digital labor focus on the unpaid work of users, arguing that all users produce a valuable good in the form of their attention (the “audience commodity”) [iv] or data captured by the platform. Instead, this essay focuses on the paid digital labor of YouTube “Creators” or content-producing users that earn substantial portions of their income from the platform such as the two vloggers above.

As is well known, YouTube and other social media platforms play role of audience surveiller and auctioneer, selling users’ eyeballs and data to advertisers. Above, the platform also plays the part of an inspector on the shopfloor that monitors the quality of the views manufactured by content-producers. If the views meet the platform’s standards, YouTube doles out fractions of a cent for each qualified view much like a machinist might receive a piece-rate for each machined part that meets pre-defined, relatively stable criteria. For the hypothetical machinist, these criteria could be known, seen, and understood. The production of fixed, largely material or “closed-box” commodities in a piece-rate system tend to be regulated by clear, mostly settled definition of what constitutes a piece.

No such clarity exists in the production of views, the definition of which varies from month-to-month, platform-to-platform, and, on YouTube, view-to-view. For a video with less than 301 views, the platform simply counts the number of times the video has been loaded. After 301 views, this mode of counting changes and the platform begins to count only views that lasts roughly 30 seconds or more. Thus, the definition and value of a view appears significantly less settled and subject to frequent change – more like the position of an electron than a car. This should comes as no surprise to scholars of audience measurement systems, but the stakes change dramatically when we consider this not as a form of organizational sense-making, but as a measure of labor output.

Put differently, the product of a content-creator’s labor appears ontologically unstable and fluid. As I’ve said elsewhere, views lack the fixed materialities of more “closed-box” commodities such as a pin, a car, or even a film or recorded song [vi]. This instability exacerbates the already uncertain value of media products (“nobody knows”) insofar as the definition and value of a view varies based on when it occurred and audience demographics. Content producers tend to experience this as the platform’s unknowable logic and this tends to produce anxiety and uncertainty. As a man earning nearly $40,000 a year from YouTube and other content platforms explained to me, “I don’t know that I would ever really feel comfortable [doing this as my sole job] because of the way that YouTube changes and Google changes and Facebook and all these ways that you make money change.” As in the examples given in the opening of this post, views may appear, disappear, and reappear without notice or explanation.

In addition to the platform’s ever-changing views, fluctuations in a content producer’s revenue may occur for at least three reasons. First, the value of a view tends to be mitigated by whether or not the view generated an “impression,” industry parlance for the number of advertisements “served” to a consumer in a particular view. For example, a video generates two impressions if a consumer sees two advertisements, but this may only count as one view. Likewise, a viewer that sees no advertisements generates no impressions and thus no revenue. This may affect the calculation of views and revenue. Second, views may be deemed illegitimate if they stem from content that violates the platform’s copyright rules or community guidelines. In those cases, the platform may strip the content-producers’ revenue from those views. The following explanation from a manager at a YouTube production network was echoed by a number of Creators with whom I spoke:

YouTube is very coy about their algorithms. They don’t want to give us too much information. For instance, you know, there’s different kinds of strikes that you can get on videos. … [YouTube] will choose either to leave it up or take it down. So when I contact YouTube and I’m like “Hey, this video received a community guidelines strike, like, I reviewed the video personally and I don’t really see anything in there that I believe violates the community guidelines strike, can you give me a little more information on that.” YouTube will generally be like “no, we can’t, but you know, just know that we did review it and we found it to be in violation and it will expire in two weeks” so then we say “Okay, that’s that.”

Third, views may be removed if the platform detects “suspicious” activity. Here, creators described equally ambiguous responses from the platform. As a creator said, “It just said in the initial email that [YouTube] sent me that there was suspicious click-activity on my channel.”

This puts the products of YouTube labor in a doubly precarious situation. Like many other contingent media workers, they are not technically in the employ of any company. Unlike a contract worker, they are only paid for views garnered by their content and the definition and value of those views remains in flux. This would be a bit like being a software programmer that was paid based on how many people used their lines of code and then that payment varied by the demographic information of the software users. Imagine television actors’ pay rates being tied not to union agreements but to Nielsen ratings. Now imagine if those ratings changed, not from broadcast to broadcast, but retroactively and without warning.

For both content-producers and researchers alike, the platform’s fluxing enumeration of views and opaque disciplinary actions presents at least two distinct problems. First, sociological research suggests that the cultural logic by which labor comes to be measured and thus valued shapes both the meaning of work and the potential for collective labor mobilization [vii]. How then might the fundamental instability of a view affect how content producers understand both their position vis-à-vis the technological (often global) capital of platforms? Second, and of more practical concern for creators, how does one mobilize against power that remains partially unknowable and, often, unreachable? This is not solely a problem for YouTubers. As I’ve shown elsewhere, this affects office workers in digital media just as much as it affects vloggers, bloggers, journalists, and other workers and organizations downstream from media distribution platforms. They all occupy similar structural positions vis-à-vis platforms and from below, the platform appears inscrutable, high above in the cloud.

1. This essay stems from research funded, in part, by the National Science Foundation’s Dissertation Improvement Grant in Sociology (Award Number 1636662).

2. I report a range because YouTube’s “users” may not discuss their exact payment rates, less they violate the platform’s user agreement.

3. E.g., Tarleton Gillespie, “The Politics of ‘platforms,’” New Media & Society 12, no. 3 (May 1, 2010): 347–64.

4. E.g., Dallas Smythe, “Communications: Blindspot of Western Marxism,” Canadian Journal of Political and Social Theory 1, no. 3 (1977): 1–27; Lee McGuigan and Vincent Manzerolle, eds., The Audience Commodity in a Digital Age: Revisiting a Critical Theory of Commercial Media (New York: Peter Lang Publishing Inc., 2014).

5. See Trebor Scholz, ed., Digital Labor: The Internet as Playground and Factory (Digital Labor: Routledge, 2013).

6. Michael Siciliano, “Control from on High: Cloud-Computing, Skill, and Acute Frustration among Analytics Workers in the Digital Publishing Industry,” Research in the Sociology of Work, Research in the Sociology of Work, 29, no. 1 (August 19, 2016): 125–53;
See also, Karin Knorr Cetina and Urs Bruegger, “The Market as an Object of Attachment: Exploring Postsocial Relations in Financial Markets,” The Canadian Journal of Sociology 25, no. 2 (April 1, 2000): 141–68.

7. Richard Biernacki, The Fabrication of Labor: Germany and Britain, 1640-1914 (Berkeley: University of California Press, 1997).