Cold Discovery

What is lost when we “watch Netflix” rather than shows and “listen to Spotify” rather than songs?

Books increasingly don’t have covers: The rapid rise of tablets and e-readers has led to more books being read on screens, which de-emphasize the cover as both a visual identifier and a physical delimiter. A cover once represented a book’s tangible individuality, its discreteness. Now, on screens, covers persist as vestigial rectangular images, superfluously ornamenting search results or PDFs. Does that shift in emphasis mean readers engage more directly with texts themselves, rather than judging books by their covers as the cliché warns? Fifty Shades of Grey and self-help books boomed in popularity on post-cover devices. Are we finally free to read what we really want, safe in the knowledge that no one can judge us?

Self-consciousness about what we’re reading isn’t the only thing likely to vanish with book covers. Beyond letting readers publicly signal their identities, covers are part of a whole regime of organizing information and space that is now in danger of disappearing. While the design of libraries and bookstores prioritizes the coherent visual display of book covers and spines so that people can navigate collections and find the singular physical objects the covers signify, the endlessly rewritable surface of the screen dispenses with that arrangement. Screens foreground the digital platform itself as singular, and thereby assimilate any particular text into the theoretically unlimited succession of information they can display. The distinct identity that a particular cover conveys has been traded for a standardizing consistency that unifies everything displayed on a screen as data flowing in a broader stream.

When we watch video on Netflix, we are watching “movies and television” at a less differentiated level of abstraction

Music has undergone a similar transition: While album covers still survive as onscreen companions to digital listening, they have moved from the center of the experience to its periphery. Like the envelope icon that still represents email, album covers have shrunken into stylized reminders of what the technology you’re currently using is making obsolete. Covers aren’t essential for discovering content on platforms. Their images merely decorate the interfaces of menus, icons, and search bars that enable navigation; their original purpose has been overtaken by the algorithmic logic that platforms use to organize and deliver content. That software, more than screens or any other hardware, mediates information access and discovery.

If covers can be construed as misleading or superficial wrappers, platform algorithms are hardly more honest. They introduce their own form of deception, feeding users content according to biases and affordances that are frequently opaque, obfuscated, concealed, or misleadingly represented. At their most transparent, streaming services like Spotify reductively mirror your past choices back to you; Netflix, however, has gone as far as to covertly recontextualize movie screen shots for its menu displays based on individual viewers’ data, in order to entice those viewers to watch more — an algorithmic subversion of the physical cover that is slightly different for everyone.

These dynamics highlight how, on platforms like Spotify and Netflix, specific artists and their works are not the objects offered to the users for consumption — a focus that covers supported. Instead, the object of consumption is the platforms themselves. More than watching certain shows, we watch Netflix; more than listening to songs, we listen to Spotify; more than reading particular books, we read our Kindles. We can choose not to pay attention to the details beyond that. Our peers frequently don’t know what we’re watching, listening to, or reading, but we don’t necessarily know either.

In Understanding Media, Marshall McLuhan argues that every medium contains another pre-existing medium as its content: Speech is the content of writing, while writing is the content of print. Print, along with music and video, has become the content of another medium, the internet and its digital formats. Movies and television, in the aggregate, have become the content of Netflix, so when we watch video on Netflix, we are watching “movies and television” at a less differentiated level of abstraction.

When we step into these universal content libraries, we rely less upon crude clues like covers for navigation and give ourselves over to an environment that offers to do almost everything for us, if we’ll just let it. With nearly any imaginable book, movie, or song close at hand and fewer of the traditional forces that push us to choose one over another, the context provided by those limitations fades and the environment itself gains corresponding prominence — at the expense of any specific thing within it.

More than a mere aesthetic shift, this transfer of activity from the physical world to the digital thrusts us into a landscape we haven’t yet learned to navigate, one in which familiar cultural dynamics no longer operate as we expect. Traditional sources of shared meaning and emotional well-being, cultivated in human communities for millennia, fail to translate to these new environments as individuals’ behavior becomes less visible. The euphoria of endless choice and convenience is counterbalanced by a sense of anomie, in which things that used to matter — like taste — don’t seem to anymore and lack an obvious substitute. Meanwhile, the platforms themselves, in their totalizing ambition, take advantage of our disorientation and choose behavior for us.

As Netflix’s cover-image algorithm unintentionally demonstrates, covers don’t just passively represent the content they contain. For centuries, the cover has functioned as the gateway to a work, priming us to receive what’s within.

In the Middle Ages, when manuscripts did not yet have formal titles or distinctive covers, that priming role was fulfilled by the text’s incipit — the first few words that served as an identifying label for the work. Before the printing press, concise titles weren’t necessary to identify manuscripts, as most people owned only a few and could differentiate them well enough by their opening lines. Covers of books, similarly, were made by hand and not standardized. With the advent of the printing press, standardized titles and covers became necessary amid a sudden influx of books to provide context where the incipit had previously sufficed. Those covers and titles linked books to their surrounding environment, not only making them searchable on shelves and in libraries, but creating the very need for shelves and libraries in the first place.

The cover helps to situate a work’s accessibility; it is “infrastructure,” as Sanford Kwinter writes in an essay for Harvard Design magazine:

a book’s cover (or to be precise, its jacket) belonged traditionally not to the book itself but rather to the retail or public environment in which a book is deployed and displayed, in which it claims its place among other books, and in relation to the public eye and mind of the citizen-reader. The cover was therefore conceived as a material part of the extended urban infrastructure within which it was intended to operate, was devised to serve as a communicative but also commutative and conductive surface.

Far more than a container or shell for content, the cover is an interface between that content and human society, the intermediate layer that positions the information in the world.

The definition of infrastructure is often uselessly broad and abstract: One early 19th-century definition  was “the installations that form the basis for any operation or system.” But another approach is to view infrastructure as context — that which establishes a relationship between one thing and other things. Infrastructure creates adjacency where it wouldn’t otherwise exist, frequently in the form of a physical connection. For instance, the massive Denver International Airport, opened in 1995, put an otherwise relatively remote city at the doorstep of the world, replacing a small regional airport with a major international hub. Urban street systems link houses, stores, and workplaces, defining neighborhoods and cities as coherent entities. Airports and roads, however, are only the most tangible examples of infrastructure. Organizational schema like geographic coordinates or the Dewey Decimal System are also infrastructure, as is the internet and everything it comprises, at a global scale.

When we engage with these platforms, however briefly, they envelop us with their logic and their norms

If covers are infrastructure, what adjacencies do they create? They not only make it possible to find a particular printed book or vinyl record in physical space; they also organize the very existence of spaces dedicated to perusing that media. Record stores depend upon covers, as surely as a city depends on roads. Each governs our ability to explore.

A physical cover is a specific analog protocol for engaging with a particular physical object; it’s akin to hailing a taxi by waving your hand in the air rather than ordering a car through a ride-hailing app. Both covers and taxi hailing require being in a specific place at a specific time and making a visual connection. If an app like Uber eliminates the need for mutual eye contact between taxi driver and passenger, thus expanding the range of possible driver-rider matches by orders of magnitude, then media platforms effect a comparable expansion, in the size of their libraries as well as in users’ ability to browse them. In fact, the enhanced browsability permits the expansion. Netflix, Spotify, and Amazon offer far more products than consumers could navigate without the aid of relational databases and screen-based interfaces.

To make these enormous catalogs navigable, platforms reconstitute some of the dynamics and limitations of physical shopping. Algorithms use keyword searches and other data inputs, including correlations with millions of other users’ queries, to shrink the scope of our potential browsing from effectively infinite to something human-scaled and closer in variety to that of a brick-and-mortar store. A manageable array of images (often covers) then appears on the screen, letting us browse the way we always have. But because this array is instantaneously generated rather than predetermined by physical inventory, it requires a different kind of logistical backdrop. Platforms are all-encompassing, in ambition if not yet in practice; the screen, able to display any image immediately, is the ideal endpoint of such a massive system, even if it can only display the tiniest fraction of what’s available at a given moment. The platform’s huge selection doesn’t overwhelm us because we believe more choice is good for us whether we can handle it or not. Regardless, we never have to face the blinding totality of that variety all at once.

Digital platforms are thus a more refined, rationalized, and totalizing kind of infrastructure. Because they accomplish their specific goals so effectively — efficiency, legibility, and breadth of selection — they tend to rapidly replace their analog predecessors once they appear. When we engage with these platforms, however briefly, they envelop us with their logic and their norms.

Legible, contextualized environments are something humans deeply desire. Urban theorist Kevin Lynch, in his 1960 book The Image of the City, called this “imageability.” He argued that “a vivid and integrated physical setting, capable of producing a sharp image, plays a social role.” Such a setting provides “the raw material for the symbols and collective memories of group communication,” which in turn “gives its possessor an important sense of emotional security.” So a bookstore, from this point of view, is not just a place to find books, but a place to confirm a sense of belonging, one rooted in a shared sense of how things are categorized and organized.

By making intentional curation and random digital output equivalent, the platform gains relevance at the artists’ and users’ expense

Platforms deplete those forms of imageability that we are used to. They are effective, to an unprecedented degree, at placing global-scale variety and informational agency at their users’ fingertips, but this often fails to foster “emotional security” as we know it. Though we spend much of our waking lives in a state interpenetrated by the analog and digital — checking Twitter at the dinner table with friends or family; sharing a day at the beach on Instagram; streaming Netflix movies while falling asleep — we still value that stabilizing power and imageability of physical space, which offers familiar signposts to help us navigate the world as an unprecedented flood of information rises and swirls around us. We may adapt and get better at reconciling the two domains, but we haven’t yet internalized all the assumptions and instincts that digital landscapes require. The displacement of the analog by the digital remains particularly disorienting (or “disruptive”) as software “unbundles” and “bundles” the world.

A world increasingly pervaded by this platform logic might resemble the world Chenoe Hart imagined in speculating on the future of self-driving cars (another digitally oriented infrastructure): “Once physical locations are rendered as abstract coordinates in a user interface, they effectively become arbitrary, as interchangeable as the retail spaces of big-box stores. The experience of inhabiting any particular interior space might become decoupled from its existence within a specific place, free from the baggage of associated historical and geographic context.” As self-driving cars might one day make locations more interchangeable, streaming services have made content more interchangeable, by abstracting it as instantly queryable data.

Digital platforms introduce a new type of context that relies on branding, platform unity, and network topology more than geographic proximity. To software, locations are just spatial coordinates separated by quantifiable distance (as demonstrated by geographically situated apps like Uber or Yelp). The existence of adjacent, unrelated things whose fates become intertwined by their coincidental proximity is a foreign concept within relational databases, where any two things can become adjacent given the right query.

Platforms posit space without imageability, or geography without adjacency. Inhabitants of that space, not grounded within it by any legible organizing principle, nor able to identify with any specific place above any other, lose that fundamental source of shared meaning along with the emotional security that it fosters. We may recoil at the world Hart describes and affirm our desire to hold on to geographic context, even though we dip in and out of such a world constantly via our phones.

Earlier versions of the internet, which still survive in residual pockets, once replicated the adjacency of the physical world more faithfully: “Surfing the web” involved bouncing between personal websites and blogs linked together by webrings and blogrolls, a structure that encouraged adventures down unplanned rabbit holes. In the past five years, though, Google and Facebook gained direct influence over the majority of internet traffic, and computer-generated certainty has encroached upon human-generated randomness. In this environment, adjacency is replaced by a software-driven prescriptiveness that bypasses the peripheral and unrelated. “Rabbitholing” now refers to YouTube’s algorithmic suggestion of increasingly inflammatory content, a top-down phenomenon still influenced by individual agency but shorter on serendipity. On Google and Facebook there are fewer happy accidents; our proximity-derived relationships give way to a quantum entanglement that knows no distance.

We still value serendipity, though — on the internet as well as on city streets — as a form of randomness that, by definition, lies beyond the scope of a system’s rules or objectives. As systems governed by explicit rules and objectives increase their influence over daily life, threatening to radicalize us in strange rabbit holes or just sell us things we don’t really want, serendipity is a way to be something beyond just a user.

The replacement of serendipity with algorithmic flow affects how we discover and engage with art. The same contextual shift that makes covers less important also demotes individual works and artists to a lower position on the value chain. Again, the phrases “watching Netflix” and “listening to Spotify,” as opposed to watching or listening to something specific on them, suggest that these platforms denature their content and assimilate its identity into their own. While a book cover wrapped an individual work — an independently defined, freestanding unit of content — a platform interface wraps the entire collection of works that users can access through it. In the process, that collection becomes a slurry of fungible content that fuels the platform. Volume supersedes specific quality. It matters less which movies and shows are on Netflix, and more how many are on it. Past a certain threshold, most users can live without any particular movie. Platform content thus becomes raw input flowing into a factory toward a destiny as a single, homogenous product.

No form embodies the platform’s ascendance as well as Spotify’s fundamental organizing structure, the playlist. Playlists can be handmade by another person or algorithmically generated from one’s own past listening behavior. In either case, the unifying theme is no longer the artists who made the music. The juxtaposition of artists on a playlist might be a savvy act of intentional curation or a random digital output, and Spotify doesn’t necessarily want you to distinguish between the two. By making both equivalent, the platform gains relevance at the artists’ and users’ expense. As Liz Pelly notes in an essay for the Baffler, “It turns out that playlists have spawned a new type of music listener, one who thinks less about the artist or album they are seeking out, and instead connects with emotions, moods and activities, where they just pick a playlist and let it roll.” At the same time, the identity of content is unbundled from its makers and curators and rebundled with the brand of the platform that streams it. Like the placeless spaces the self-driving car will create, Spotify minimizes friction and frees songs of contextual baggage so they can exist anywhere.

Services like Netflix and Spotify have embraced the not-quite-familiar context that digital environments offer as a replacement for physical adjacency. Despite whatever disorientation they induce, those services offer some compensating comforts: The self-expression of owning a record collection and the elitism it once enabled give way to an egalitarian tropics in which everyone theoretically listens to a bit of everything and nobody’s taste is entirely legible. We can safely read Fifty Shades of Grey without judgment. The algorithmic curator we each personally train validates our preferences by reflecting them back to us, allowing us, as Rob Horning writes, to consume our identities rather than perform them.

We’ve only just begun to assimilate these platform landscapes into our understanding of the world. Kevin Lynch believed that an intelligible physical environment — the kind that Netflix and Spotify have left behind — could, at its best, provide a foundation for “the symbols and collective memories of group communication.” Imageability guards against solipsism, providing cues that force us to acknowledge a reality not specifically tailored to our own preferences and enabling a joy beyond what individual experience alone can bring. At their worst, platforms threaten to send each of us down our own decontextualized rabbit hole, weakening collective narratives and exposing us to active attacks on what Renee DiResta calls “society’s ability to operate with a shared epistemology.”

While many lament the transition to a digital realm that doesn’t naturally accommodate those familiar communal symbols and many others uncritically embrace it, a healthy synthesis of the two positions is possible and even necessary. We can acknowledge the widespread arrival of a new technological medium and learn how to create shared meaning within it that is as robust as more established forms of meaning, though not subject to its constraints. This will likely require different platforms, norms, and behaviors than those we have now. Lynch coined the concept of imageability as it was receding from automobile-stricken urban environments. We may need more theorists like him to help us recognize new forms of digital imageability, ones worth preserving and developing before they too start slipping away.

Drew Austin writes about technology and urbanism on the blog Kneeling Bus.