Research Note: YouTube Is Infrastructure

YouTube may have started as a platform, but it has become infrastructure.

By 

Ryan McGrady

April 21, 2022

What's on YouTube?

What is YouTube? Those of us working on Media Cloud's International Hate Observatory have been asking that question and coming up with a lot of different answers. Our long-term goal is to understand the hate speech and extremist material hosted by YouTube, but before we start analyzing or quantifying harmful content, we want a clearer picture of YouTube as a whole. What’s on YouTube, who put it there, and who is it for?

What are the sounds and images which first come to mind when you think of YouTube? In 2022, most people probably think of someone playing a video game, a clip from a cable news broadcast, a music video, an explainer video about a scientific concept, or maybe a screaming deer. You probably don't think of a two-hour condo board Zoom meeting, a video taken by someone walking through a house when it was for sale in 2013, a silent one-second clip of a smiley face emoji, a personal message from an individual to their family, an hour of choir practice, or a screen capture of a phone's home screen, but those are out there, too. In fact, we believe there's vastly more of this audiovisual hodgepodge on YouTube than the sort of higher-quality content the site has become known for — there's just no reason you'd encounter it. 

While it's true that nearly anyone can upload a video to YouTube and then share the link with someone, traffic on YouTube is mostly driven by internal recommendation algorithms which organize and prioritize videos according to a wide range of factors (many of which are personalized, based on the information it has about individual viewers). From the perspective of a media consumer, I'm glad YouTube knows I'd be more interested in a very brief history of the Wingdings font than a random board meeting or a realty ad, but as a researcher it presents a challenge. That's because YouTube will not allow outside researchers to audit its algorithms, so we cannot know exactly why some videos are promoted and others exist in obscurity. To make things even more complicated, while YouTube has a useful API, it does not provide any good way to obtain a random sample of its videos, so most studies of YouTube either begin with a list of channels and/or videos or work within the algorithms, which provides a set of videos biased towards more popular videos.

Thankfully, my talented colleagues have figured out this latter problem: a method to obtain a truly random set of videos. In fact, the examples of the less than compelling unpopular videos above come from an early random sample. We will be building and analyzing that set over the next few months, comparing it to other sets, like a sample of the videos YouTube recommends. We think it will allow us to paint a clearer picture of what YouTube really looks like when you peel away the filters and recommendations, and how it differs from what a typical user actually sees. In other words, it will allow us to better answer "what is on YouTube" without relying on YouTube to point us in its preferred direction. 

What Is YouTube?

"What’s on YouTube" still doesn't answer the question "what is YouTube?” The former is viewer-focused, like asking "what's on television?" as opposed to "what is television." Fundamentally, YouTube is a division of a for-profit company and a video-hosting website that people use for a variety of reasons. We want to understand those reasons or, framed another way, we want to understand what people are doing when they upload a video to YouTube. What is YouTube to them?

It's a complex question, and the most obvious answers tend to come up short. The visible, popular part of YouTube in 2022 is full of emergent and established genres, conventions, formats, and tropes that may lend to rough categorization according to stated or unstated purpose. Indeed, in contrast to predictions that YouTube would spell the demise of rigid forms like those used by television, it is easy to get the impression that the uploaders of popular videos are all playing the same kind of "YouTube Creator" game, adhering to similar kinds of technical, social, and economic rules and rationalities. But what about the rest of YouTube? 

In a library, simple categories may be sufficient for the new bestseller shelves, but the rest of the stacks need something like a Dewey Decimal System, the recreation of which is beyond the scope of what we're trying to do with YouTube. Starting to categorize random YouTube videos without such a system feels a bit like Borges' ordering of animals in "The Analytical Language of John Wilkins": "those belonging to the Emperor," "stray dogs," "innumerable ones," "those that from afar look like flies," "those that have just broken the vase," etc. We might begin our project with "those with opinions about Minecraft," "two-hour church services," "with titles containing more than six emojis," and "computer tutorials using a broken microphone."

One distinction I think we can draw has to do with the intended audience: uploads intended for specific individuals or groups vs. uploads intended for anyone else. YouTube allows uploaders to set videos to "private," removing them from search results and making them visible only to the uploader and specific people the uploader shares it with, but in our random sample of videos there are publicly visible videos which seem intended for specific other people, too, achieving a kind of privacy through obscurity: a video birthday card likely for the recipient or perhaps some friends and family, a condo board meeting perhaps uploaded for the board members who missed the meeting, unedited vacation footage, etc. These videos are more about transmission than dissemination, more about storage than broadcasting. We doubt these are a plurality of YouTube videos, but they are a significant presence, and one that is entirely invisible if looking only at the popular YouTube videos.

It is tempting to reach for metaphors to think about the uses of YouTube. Pelle Snickars and Patrick Vonderau talked about the metaphors people use in the introduction to their influential 2009 collection, The YouTube Reader. They mention YouTube as a library, archive, laboratory, television, etc. We might add: performance stage, storage locker, distribution channel, greeting card, newsroom, classroom, nanny, production studio, and social network, among others. Instead of comparing to existing technological forms, however, we could ask what type of action (or combination of actions) someone is undertaking or attempting to undertake by uploading a video. After whittling down a long list, I've arrived at nine, any combination of which may apply to a particular video: informing, influencing, entertaining, communicating, expressing, performing, socializing, distributing, and archiving. Breaking it down further could also include things like instructing, promoting, connecting, securing, discussing, analyzing, interpreting, praising, flirting, disparaging, documenting, circumventing, researching, intimidating, judging, and so on. We might also ask not just which of these apply but to what extent each applies. A final approach, which may allow for more nuance, is to try to articulate the likely intended (consciously or not) effects of an upload on the uploader, audience, and/or third parties, several of which will apply to each video. 

Whether we create a vast new classification system, chart combinations of actions, or frame in terms of effects, what makes the task difficult is this: YouTube is not one thing or a small group of things, but rather the default video arm of the internet. It's like trying to categorize the uses of "text-based websites." They span every conceivable reason one might have to record and/or share video (or slideshows or audio, for that matter). Those reasons include the frivolous, but also the vital; repackaged cat videos, but also citizens speaking truth to power; opportunities to experiment and have fun, but also the means to carry out essential activities for the functioning of day-to-day communications, learning, and business practices.

This speculation about the uses of YouTube and the intentions of its users can feel like a purely academic exercise, but how we talk about something shapes how we think about it. If we talk about YouTube as a creator economy, we miss the ways it has become a backend for virtually everything that produces and shares video. If we talk about it as the successor to television, we miss the ways in which traditional television uses it to reach new audiences on demand. Words matter, and the word we've taken to using for YouTube is "infrastructure." 

When Platforms Become Infrastructure

One of the most common words used to talk about YouTube and some other popular websites — and the term companies often use to talk about themselves — is "platform." It evokes an image of a physical platform on which people stand to communicate and express themselves, as well as its meaning from computing. I like the definition articulated by Carliss Y. Baldwin and C. Jason Woodard: a stable system with stable interfaces that supports and encourages the development and use of varied complementary and interchangeable components (a video game system, mobile phone, CD player, and operating system can all be understood as platforms). But the simplicity of the structural metaphor, and even the meaning from computing, belie something much more complex and fraught when applied to companies like YouTube.

More than a decade ago, Tarleton Gillespie wrote a great essay on "the politics of platforms," in which he explains that companies' selection of "platform" was strategic — a concerted effort to define themselves on their own terms, and in a way that means different things in different contexts. Amid the buzz of "web 2.0," with its participatory culture, peer production, and blurring of producers and consumers, a "platform" became a place where anyone with the inclination can stand and speak. Even now, YouTube's about page includes a mission: "to give everyone a voice and show them the world". "Platform" implies a kind of neutrality on the part of the platform's owner, promising to facilitate discourse, dissemination, or expression without the gatekeeping of traditional media companies. 

There is another side to the promise, too: "merely" providing a platform means little or no liability for what people use it for. Gillespie makes the point that going along with the "platform" language, with its connotations of neutrality and egalitarianism, stymies efforts to use more precise descriptions of the companies' practices and their far-reaching effects on the public. It's also just misleading: YouTube allows anyone to upload, but uses its algorithms to determine which of those are actually suggested to viewers. It does not pick and choose which to host, except for the many videos it removes. It's just a stage for other people to stand on, but it pays to produce original content. It's a benign facilitator, but its raison d'être is not to host cat videos but to make money by putting advertisements in front of people.

The framing has broadly been successful: in 2019, YouTube brought in $15 billion in revenue and continues to enjoy a legal framework (Section 230 of the Communications Decency Act) which effectively insulates "providers" from the content users upload. As with other "platforms" like Facebook, it has for years balanced looming regulatory inquiries and moves to repeal Section 230 with reluctant self-regulation. When reports are released about their role in not just transmitting but amplifying harmful content, platforms modify their internal practices (in often opaque ways), further involving themselves with content moderation and monetization to intervene in the most concerning cases.. 

As central and essential as YouTube has become to our information and communication ecosystems, and to many people's daily lives, how we talk about it is important. Gillespie saw how "platform" was woefully inadequate back in 2010, but it's still one of the first words we use to talk about YouTube. How about an alternative: YouTube may have started as a platform, but it has become infrastructure. It does still have many qualities of a platform, but at some point a flawed metaphor obscures too much. My colleague Ethan Zuckerman once wrote about an excellent example of YouTube as infrastructure: Wael Abbas, a journalist and activist in Egypt who frequently documents police abuses. He began relying on YouTube when his own website was repeatedly targeted for takedowns and denial-of-service attacks. YouTube-as-platform might be differentiated from Abbas's personal website by things like social features, ease of use, promotional features, a more functional interface, and connections with other kinds of platforms. YouTube-as-infrastructure prefigures the way in which a massive company with vast resources at its disposal, extraordinary technical investiture, a large legal team, and a truly global presence, is not realistically going to succumb to such attacks. (To YouTube’s credit, choosing to host videos of police abuse, even when they violate terms of service against depictions of violence, shows a willingness to use these assets to protect political speech.)

Here’s the thing about the large "platforms": they want to appear as simply one choice among many to share and communicate, but frequently become the only realistic option. As anyone who has quit Facebook (or tried to quit) can tell you, it can be really challenging — even isolating — not because it's enjoyable to have a platform for expression, but because it's part of our basic communications infrastructure.

In 2016, Jean-Christophe Plantin, et al. analyzed Facebook and Google through the lens of "infrastructure studies" and "platform studies," observing a trend they called the "infrastructuralization of platforms." It's an insightful and useful paper, but I can't help but think the phrase still concedes to the preference for "platform." I like the question asked by Susan Leigh Star and Karen Ruhleder in their 1996 study of the development of software infrastructure for scientists to sequence the genes of a worm. They consider what it means to be infrastructure along the way and reframe the question as "when is an infrastructure?" Inspired by Yrjo Engestrom's "When Is a Tool?," they argue that "infrastructure is something that emerges for people in practice, connected to activities and structures." When a platform becomes natural, transparent, and essential due to the ways people use it, it becomes infrastructure. If large parts of society (individuals, organizations, families, businesses, government agencies) organize themselves around it such that a breakdown would have profound economic, social, and operational effects, it probably makes more sense to talk in terms of infrastructure than platforms. Infrastructures are, after all, most visible when they stop working properly.

Changing the language we use to talk about YouTube allows a discussion based on its real place in the world, with real uses by people who depend on it. Looking at its popular videos, it is easy to get the impression that people use YouTube to participate and gain from a specific YouTube Creator economy, with clear genres, conventions, and efforts to build brands and monetize videos. Looking at sets of random and popular videos, however, there are so many more ways that people use YouTube than I would have been able to guess. It's just "where you go" when you want to build a library of church services, to supplement a Twitch.tv-based business, to share political campaign updates, share a clip from your favorite television show, share your thoughts, allow a coworker to catch up on the latest meeting, store a personal video, or document police abuses. It will be worth doing some additional theoretical, historical, legal, and sociological work to better understand the differences between infrastructures that began as infrastructures vs. infrastructures that began as platforms, and what opportunities there may be to shift the public conversation around them, but for now I'm just looking forward to discovering more of the many ways people use YouTube — the many things people do when they upload to YouTube, or the effects people want to have on themselves, audiences, or third parties.

This post has focused on the uploader side of YouTube, but I don't mean to ignore YouTube's no-less-essential role for viewers. Recently, the Washington Post Editorial Board published an opinion piece titled "YouTube must hold the line in Russia" which highlights the way YouTube has become "the surest way for Russian citizens to keep abreast of what’s happening in the world, both just over its border and far away." The Post suggests YouTube (which does not have a spotless record on giving in to censorship, though it has been better than its counterparts) is just too popular to ban entirely, even after Facebook, Instagram, and Twitter have all been blocked. When YouTube becomes the last hope for obtaining information without a government filter, we are no longer talking about platforms where you might choose to speak — we are talking about fundamental infrastructures of human communication.

Hero image by Jim Bumgardner via Flickr. CC BY-NC-SA 2.0.