How Social Networks Play Favorites

To the partisans of social media, a sudden burst of viral fame is a path to success. The contours of this path may be elusive, but that is all the more reason to spend more time on social media, honing your brand, developing your voice, cultivating your audience, and all of the other vaguely buzzy practices around which a flourishing consulting industry has developed. But as with semantic analysis, the people achieving success here are not those trying to go viral at all costs, but rather those selling the shovels. It’s the writers of how-to books, the leaders of viral-marketing seminars, and the highly paid advisors, experts, and keepers of metrics who have profited handsomely from the hunger to go from unknown to famous through some social media alchemy. And most of all, it’s the platform owners.

Some of us have something tangible to sell—the rise of self-publishing has spawned an ancillary industry telling newly minted authors how to get their books noticed through social media—but many of us are just selling ourselves. And so the murky quest to go viral has filtered down to the proles, who ask celebrities, for no reason but for the distinction of the thing itself, “Can I get a retweet?” A related custom: pushing to get someone trending on Twitter. At its most pathetic, this request is made on behalf of someone sick or dying, as if Make-a-Wish had established an arm to achieve viral fame for the terminally ill. Ephemeral renown for an ephemeral life. Let’s get him trending, his supporters say. Let’s get his name out there. But why? For what? So that it can pop up in the list of trending topics? People should be allowed their particular (or peculiar) forms of solace, but there’s something very sad about this one: how it depends so much on the goodwill of a disinterested higher power, how meaningless and contingent the desired result is.

Like the concept of going viral, a “trending” topic (a term native to Twitter that has since percolated throughout social media) is treated as a good unto itself, a sign of public interest in a subject, without much deeper exploration of the technological and commercial forces that determine what trends. As Jared Keller, the director of programming at Mic, a media company catering to millennials, told me: “Just because it’s trending doesn’t mean it really matters. But we think of social media as a really great mood ring for popular culture.” But there are some unexamined assumptions with this kind of thinking. Twitter is “self-selecting,” Keller said. “People opt in to social, so they opt in to those kinds of conversations. Is this really what matters, or is it just what matters to the Twitter audience?”

There’s nothing inherently good, worthwhile, or organic about a supposed trend. One study found that “most trends in Sina Weibo [a Chinese micro-blogging service] are due to the continuous retweets of a small percentage of fraudulent accounts.” As Evgeny Morozov writes, “The hidden initial manipulations of the PR industry are only made worse by the business incentives of platforms such as YouTube and Facebook, which have their own reasons to promote memes: they create some shared culture and, more important, lead to more page views, more user interaction (i.e., users reveal more about their interests to the company), and, eventually, more and better advertising. Memes, then, are what happens when one greedy industry meets another.”

In other words, social networks play favorites, while PR companies, producers, journalists, and others have the power to influence what rises to the top. A video featured on YouTube’s home page will accrue tens or hundreds of thousands of views simply by virtue of being placed there. Kickstarter, an ostensibly meritocratic crowdfunding platform, regularly features selected projects in its newsletter and on its home page. Such a distinction can bring thousands of dollars into a project’s coffers, which is often the difference between achieving a fund-raising goal—and being seen as a success—and walking away with nothing. In the same way, when Twitter shows “related headlines” below tweets about news events, the site is making deliberate choices—choices made in the design of an algorithm, rather than by an editor working on the fly—about which news outlets to privilege above others. Sponsored posts jostle alongside regular editorial material, the difference between the two barely marked, if noticed.

Views, likes, mentions, hashtags, followers, all the various indicators of virality—these can be bought. Seek out the right forums or tucked-away online shops, and for about the cost of dinner at a fancy restaurant, you can create the illusion that your previously unknown video or Twitter account has a huge fanbase. A site called YTView has a full menu of purchasable popularity; for YouTube alone, it offers views, likes, subscriptions, or comments. You can buy Facebook likes or raw Web traffic—indeed, the digital media industry is rife with rumors and accusations of various publishers using click farms to boost views and, by extension, advertising revenue. If you want to raise your status among your peers, you can send some cash through PayPal for Instagram or Twitter followers. It’s a large, albeit shady industry, believed to generate millions of dollars in revenue. Relying on bots, purveyors of views and likes—call them sellers of attention-on-demand—argue that theirs are real followers, while the other guy’s are phony. Or, they promise, they can provide better engagement or views only from the United States, which are less likely to attract unwanted scrutiny. (Fake views often come from eastern Europe and east Asia, part of the broad archipelago of the Internet’s black market.) That is, assuming that YouTube, Twitter, and other platform owners care about the fakery. While these sites are known to purge bots, fake followers, and fake sentiment—a practice that serves as a public affirmation of their networks’ authenticity—they probably also don’t mind the traffic.

Wanting to find out how this process worked, I went in search of followers of my own. A simple Google search for “buy Twitter followers” led me to dozens of sites. Some looked like hastily thrown together operations; others had more sophisticated graphics, bright logos, and fancy WordPress templates. Despite the differing aesthetic presentations, it was hard to sift out which was more reliable, given that the entire industry operates in the shadows and is built on a certain amount of deception. Even those that claim to provide more fully fledged marketing services don’t seem like more than fly-by-night operations. I decided to let my wallet guide me. A site called Social Burst offered 2,000 Twitter followers for $15—a few dollars below the going rate on some other sites. I entered in the name of my Twitter handle, @silvermanjacob, and sent $15 to someone named Matthew Nuttall. Because I never had to enter my password or authorize the site to access my Twitter account, and because I paid the fee through PayPal, it all felt relatively secure. It was also astonishingly easy. Soon a confirmation e-mail arrived, promising that my followers would appear within one to two working days.

A few hours later I had forgotten that I had even made the transaction, until I opened my Twitter page and realized that my two thousand followers had already appeared (almost 2,200, in fact). Then I embarked on that peculiar, Internet-era game: Are they real? What does real even mean for these purposes? I scrolled through the list, clicking on some individuals, and gathering a sense of the rest. Most had full names, avatar photos, and had tweeted hundreds or thousands of times—signs of verisimilitude but not necessarily of life. A number of them appeared to be from overseas—Latin America, the Middle East—and followed far more people than followed them, though that may have been a sign only of their relative lack of popularity. What seemed like a disproportionate amount had locked their accounts, and I did find the occasional account that was tweeting spam or outright gibberish. Other messages were simply incomprehensible to me, shared as they were in languages I didn’t understand.

Given the speed with which they appeared and this sketchy collection of characteristics, I’d have to say that most of the accounts were fake. But did it really matter? Is someone who tweets once a month in Tagalog more real than someone else who tweets in English about cricket ten times a day? On the other hand, maybe there is a meaningful distinction between a bot that tries to emulate its followers and one that spews out deals for porn sites. Still, there’s a range of behaviors we expect from not only our followers, but also those we consider real and those we want to interact with; someone might be quite real but totally uninteresting to me, and I to them. Consider also that Twitter, in March 2013, reported that it had 200 million active users out of 500 million registered accounts—meaning that 60 percent of registered Twitter users no longer logged in (a sedentary posture that could also be confused for bot-like behavior). A year later, Twopcharts, an analytics firm, reported that 44 percent of 974 million registered Twitter accounts had never posted anything. What’s more, past studies have found that many of the most popular Twitter accounts of politicians and celebrities are filled with followers that are actually fake or bots pushing spam. During the 2012 GOP presidential primary, Newt Gingrich trumpeted his 1.2 million or so Twitter followers as a sign of his popularity with the electorate, until several media outlets reported that they were fake and that his campaign used its war chest to pay companies to push followers his way. In short, the range of what’s real/fake/active/inactive/a bot/not a bot becomes more complex the more you dive into the data and into varying definitions of what should be expected from a Twitter user.

For my purposes, questions of authenticity didn’t really matter. I could imagine some other journalist scolding me for misrepresenting the size of my audience, but that also would strike me as picayune. I had no boss to answer to, and no one really cared about how many people follow me on Twitter. But I admit that I cared a little bit; no matter how much I tried to avoid the implicit measurement contest of comparing follower counts, it was hard not to think that this information did matter, even if for the wrong reasons. These are kinds of metrics that get dropped in hiring memos or that writers, with some shallow jealousy, marvel over when talking about competitors and colleagues. My inflated follower count gave me a small lift while also confirming my suspicion that these theatrics over metrics of popularity were both ridiculous and easily manipulated. Now that I had so easily, if perhaps irrelevantly, gamed the system, I was able to share in the sense that it was all a farce. But at the same time, I had this nice beefy follower count, cresting over 4,600. It felt like what I imagined Botox might feel like: an artificial-but-hard-to-detect boost to both my ego and the face that I showed the world. That is, until the face started to droop. Over the next several days, I lost about 400 followers, though it slowed in the following weeks. Some services claim that they allow you to “top up” followers in instances precisely like these. I decided to check on Social Burst and see what its policy was. Social-burst.com now redirected to instapromotion.co.uk, which in turn claimed a copyright in the name of Get Socials 2014. It was becoming clear what kind of world this was: one in which Web sites and followers alike flitted in and out of existence, changing their names and profiles as the opportunity presented itself. For a customer like me, it still didn’t matter much. I had pretty much gotten what I paid for, and besides confirming for myself how easy it was to gin up attention on demand, I had the mild satisfaction of knowing that I was fooling those people who might care about things as arbitrary as follower counts.

For musicians, actors, and other artists, as well as their managers, the goal of botting, as this practice is known, is to quickly acquire enough attention to create the impression of legitimate interest. And it’s important that that interest seem organic and rapid, as if the media object is already incipiently viral. It then only needs to catch the attention of an “amplifier”—a YouTube moderator, a popular Twitter account, a celebrity, some bloggers or journalists trawling for content—after which it can take off and be truly viral. The despicable “Innocence of Muslims” video, which eventually caused riots in several Muslim-majority countries, lurked unnoticed on YouTube for months until Arabic subtitles were added, and it then found amplifiers in an Egyptian newspaper and the Islamophobic pastor Terry Jones. The “Harlem Shake” only became a world-spanning meme after being promoted by the popular DJ Diplo and the site CollegeHumor.

It’s a form of “fake it until you make it,” although the fakery of buying attention is, in some regards, a more honest form of PR. Instead of flattering journalists or sending them swag, or cozying up to the owner of a popular YouTube channel, or exchanging favors with a celebrity with a big Twitter following, the attention is bought directly. No matter that the audience is likely composed of bots: the assumption is that the real views will come later. And if not, the agent or PR rep can still tell his or her boss that they got x number of views in such a short period of time; no need to say that these weren’t real people, if it’s all the same to the higher-ups. In the meantime, if the party buying YouTube views is a YouTube partner, they might even make back some of their investment based on the ad revenue that the fake views generate.

Botting and related practices are just the seamier side of how attention is traded, bought, sold, or pilfered on social media. You could pay a celebrity such as Paris Hilton $4,600—the reported going rate for that archetypal viral star—to tweet about your video or product. There are Internet-famous social-media users who charge far less. You can try to piggyback on an already popular hashtag, as some corporate marketing departments are wont to do, particularly with hashtags cel- ebrating national holidays (Memorial Day is a great opportunity for a fast-food chain to tweet that it supports veterans). A range of Web sites selling piecemeal labor, such as Fiverr or Mechanical Turk, can also bring you “legitimate” traffic.*

Even Twitter engages in a version of this. Its algorithms can pick up which topics people are talking about and determine if they should be declared “trending.” But there’s a lot they can’t do. Like semantic-analysis programs, they have trouble with tone or unorthodox meanings. And Twitter’s algorithms can’t always correlate breaking news events with the queries that people are searching for on the service. So the company built what they call a “human-computation engine”—a fancy term for a distributed network of low-paid contract workers, some of whom are guaranteed to be available at any time of day. When new search terms are spiking, the system sends questions about the terms to a team of “human evaluators” recruited from Mechanical Turk, an Amazon-owned company that allows people around the world to perform menial tasks, such as transcription or tagging images, from their computers for a tiny fee. The Turk workers then determine that, say, #bindersfullofwomen (the example provided by one Twitter engineer) is a reference to Mitt Romney’s odd elocution in an ongoing primary debate being watched, and tweeted about, by millions of people. The workers classify the term appropriately and send it back to Twitter, which allows the company to determine whether the topic qualifies as trending.

Twitter offered some reasons why they work with Mechanical Turk: “In-house judges are unfortunately hard to scale as they require standardized hiring processes to be in place. They also tend to be relatively more expensive, it can be harder to communicate with them, and their schedules can be difficult to work with.” In short, Mechanical Turk workers are cheap and always on call. They don’t demand much, such as regular work hours or minimum wage. Cheap scalability is paramount, even for a corporation worth more than $20 billion. The company adds: “Our custom pool of judges work virtually all day. For many of them, this is a full-time job.” But it’s full-time work for Twitter Inc. without any benefits, institutional support, room for advancement, or pay that should come with a full-time job. The tasks available for Mechanical Turk do offer some people important flexibility, especially for stay-at-home parents, as well as a source of income for people who need a little more cash or are unable to find permanent work. But for a company as large as Twitter, hiring an in-house, full-time “human computation” staff shouldn’t be prohibitively expensive. Twitter would presumably gain from bringing the staff closer to home; it’d be easier to respond quickly to breaking news with on-site employees and to learn from their experiences. These employees would in turn have opportunities to learn new skills and gain responsibilities, and presumably, some would be happy to work night shifts, as the company clearly requires. It would also help affirm Twitter’s place as a media company, as many tech commentators have long considered it. Any legitimate media organization has staff working a night shift to respond to events around the world. Twitter has such a staff, but they’ve chosen the easier, cheaper option—one that ensures that one of Twitter’s most distinctive features will always have a whiff of exploitation underlying it. At least until Twitter decides that its human computation engine can be replaced with automated algorithms.

* Or at least they promise it. As I write this, one of Fiverr’s top users is a UK man calling himself Giblerto Samba, who for $5 will promote your music to his 100,000-odd Facebook fans. He also will include various hashtags to help get the post trending on Twitter and Facebook. His disclaimer cautions: “We do not guarantee activity on FB/YouTube/Twitter as we don’t force anyone to react.”