Future Tense

The Case of the Ornamental Anthropologist

How Netflix puts a human face on Big Data.

Kevin Spacey in House of Cards
Kevin Spacey in Neflix’s House of Cards. He knows when you’ve been binging.

Photo by David Giesbrecht/Netflix

Netflix, if you believe what you read, is more computer than corporation. It famously operates through an interlocking set of finely tuned algorithms powered by the immense amounts of information that the company collects. These algorithms allow Netflix to anticipate everything from what you might want to watch at a particular time of day to the actors and directors whom it should hire for its original programming. This is Big Data put to work, a buzzword come to life.

There are, of course, things Netflix’s algorithms can’t predict. While it’s good at calculating the kind of content we’d like to see, it has less information about our real experiences of it: Netflix knows that I blazed through three episodes of Unbreakable Kimmy Schmidt on Sunday, but it doesn’t know that I was making (and then eating) a pretty great sliced cauliflower salad while I watched. Likewise, it couldn’t listen in as my editor and I chatted about the show the following morning.

Without that information, Netflix’s algorithms can paint only distorted pictures of its users. This, at any rate, is true of most data collection efforts. When reporter Julia Angwin surveyed the information data brokers had gathered about her, for example, she was amused to find that one of them thought her coach-class family were private-jet people.

It’s no wonder then that some have begun to doubt the real reach of Netflix’s formidable data-mining operation. In the New Yorker, Tim Wu has questioned the degree to which Netflix’s original programming really derives from purely metric analysis. Wu writes that he pressed Ted Sarandos, the company’s chief content officer, on the matter, proposing that human judgment played a larger role than the company admits. According to Wu, Sarandos agreed, “It is important to know which data to ignore,” and added that his decisions were 70 percent data, 30 percent judgment.

Wu latches onto this admission, suggesting that it reveals that an actual human mind directs Netflix’s otherwise alien and alienating calculations. By way of connection, he argues:

[The company’s] biggest bets always seemed ultimately driven by faith in a particular cult creator, like David Fincher (“House of Cards”), Jenji Leslie Kohan (“Orange is the New Black”), Ricky Gervais (“Derek”), John Fusco (“Marco Polo”), or Mitchell Hurwitz (“Arrested Development”).

Here, large personalities with clearly defined artistic voices take the place of data we all yield up. This is a comforting hypothesis, but it’s also a little overeager: People still matter, Wu seems to be insisting. The computers haven’t won yet.

Netflix itself sometimes takes a similar perspective, not least of all because it recognizes the limitations of its information-gathering operation. As streaming services began to reshape the ways we consumed media, the data clearly showed that things were changing, but they couldn’t always explain why. “We didn’t know everything that was going on,” Netflix representative Jenny McCabe told me. Despite the company’s reputation as a digital pioneer, one of the solutions it settled on was surprisingly human: It hired a cultural anthropologist named Grant McCracken.

“We’ve been working with Grant for the last two years to find out what’s changing from a consumer’s perspective,” McCabe said. McCracken studies Netflix users from around the world, constructing ethnographic accounts of changing patterns of viewership from his research. Where Netflix’s algorithms parse the ways we interact with the world, McCracken seeks to describe our experience of it. The results of this work have been widely reported—partly because they’ve been widely promoted by the company itself. He’s found, for example, that viewers tolerate spoilers more than they tend to admit. (In fact, Netflix introduced us after I wrote an article on the topic.)

McCracken’s work is as fascinating as it is fun—he’s obviously a serious researcher—but it’s not clear what he actually contributes to Netflix. While Netflix has a sizable consumer research team whose members study the experiences of its users, McCracken’s work is entirely independent from theirs. Though Netflix’s own promotional materials suggest that it hired him in order “to better understand” consumer trends, his actual role seems to be largely promotional. His research is real, but he seems to serve as an external envoy rather than an adviser. When I reached out to Netflix, a representative told me that he had been hired by the company’s PR team to do his own research and talk about it with the public.Though other internal divisions of the company may have looked at his work, its function is promotional.In a separate email, McCracken confirmed that his research is not connected to Netflix’s policies or programming.

Nothing about McCracken is slight: He holds a Ph.D. from the University of Chicago. He’s taught at Harvard Business School and MIT. He’s published studies on everything from the semiotics of hairstyles to qualitative research methods. When we spoke, he was genuinely insightful about changing patterns of media consumption. He gestured, for example, to the television “fidelity contracts” many of us now form with friends and family, agreements formed when we promise not to watch particular programs without our loved ones. Likewise, he is astutely sensitive to the “linguistic markers” and theatrical gestures we use when talking about our favorite programs, such as the “OMG!” that often heralds a spoiler or the way we’ll dramatically clamp our hands over our ears to ward off that unwanted knowledge.

It’s tempting to argue that McCracken’s work with Netflix has more to do with softening its image than it does with research. In a press release about McCracken, Netflix boasts that he “went into the living rooms of several TV viewers across the United States and Canada.” Several is significantly smaller than 50 million, its user base as of mid-2014. Because he studies how individuals see themselves, he puts a more human face on the mechanistic—and arguably intrusive—ways that Netflix really collects data and classifies its users.

Netflix knows that we know that it knows more than we do about what we do in our private time, and it also knows just how unnerving that knowledge is. It’s familiar with much more than our favorite genres and actors. It can also account for everything from the color palette of covers we tend to click on to the rate at which we scroll through its suggestions. By proposing, if only implicitly, that it actually has to come into our homes to really get us, and that it does so through fussy old ethnographic methodologies, it makes its operation seem less all-seeing than it actually is. McCracken is just one man, and one man can see only so much.

Paradoxically, McCracken’s results may be a little more palatable precisely because they’re more general. In his study of spoiler culture, for example, he lists five different ways of relating to spoilers, broad categories into which we’re encouraged to insert ourselves. (Supposedly I’m an “impulsive spoiler.”) We can argue about whether we fit into a profile. It’s harder to fight against a profile that accurately predicts what we’ll watch and when we watch it. Imprecise as they are, McCracken’s categories allow us to maintain a sense of our own will and volition. They leave room for our humanity at a time when we’re increasingly being reduced to numbers.

When I was in grad school, some of my professors would complain about scholars who cite philosophical axioms without acknowledging the underlying arguments. These, they would grumble, are just ornamental quotations. They plump up an essay without adding to it. These citations may still be meaningful and true, but bereft of their real context, they merely adorn. Grant McCracken, likewise, may be an ornamental academic, there to make Netflix look a little less frightening, and to make its algorithmic future seem a little less dystopian.

This article is part of Future Tense, a collaboration among Arizona State University, New America, and Slate. Future Tense explores the ways emerging technologies affect society, policy, and culture. To read more, visit the Future Tense blog and the Future Tense home page. You can also follow us on Twitter.