Introduction: Reading with Algorithms

On March 31, 2023, Twitter (now X) released its recommendation algorithm for public viewing and annotation. The unprecedented move towards transparency included . . . no surprises. Instead, it confirmed what we knew: algorithms read. And the ways that algorithms read affect user behavior. The processes that now structure daily interactions online are cultivated by more-than-human gatekeepers, which apply explicit and implicit logic established by human programmers to determine what kind and whose cultural productions are promoted. And, in doing so, they shift users' reading and production behaviors. In the days after the program was released, developers seeking to gain followers released posts about "how to game the algorithm." But relatively few of these suggested practices were new: users attentive to algorithmic curation already adjusted their speech and productions to be tractable to each platform's algorithms. As algorithms structure the flow of information and rhetoric through daily life, users read with algorithms — and algorithms read for users.

The essays in this cluster investigate the shaping force that social media platform algorithms are having on reading practices and emerging aesthetics. AI and machine learning, data collection and control, are now inescapable facets of how people read: watched over by corporations, interrupted by advertisements, distracted by multiplying devices and open tabs, joined by others engaging with the same content, our practices all observed and reflected in new textual forms. Algorithmic effects are distributed and experienced unevenly, reflecting varying levels of access to social media reading experiences, as well as varying forms of exposure to surveillance and control. What does it mean to read with and through algorithmic structures? How does their uneven distribution affect reading cultures across and within the shifting lines of digital engagement?

When we assembled these pieces, we were particularly interested in approaches from cultural materialist and sociology of literature traditions, with a focus on new sites and styles of reading that have popular appeal beyond academic and high-aesthetic spheres. We wanted to provide space for scholars to consider the well-known realities of social media's capitalist determinants — the "platform capitalism" that drives social media companies to add users whose data they can monetize, whose attention they can sell to other companies purchasing ad space, and whose need for content to consume, and for the networked sociality of online life, ultimately drives the purchase of digital devices and investment and innovation in their production. Put at its most simple: clicks add value, and taking book culture online adds clicks. As Matthew Kilbane writes, in his introduction to a forthcoming collection of essays on poetry and social media, "To 'like' a post on Facebook or to linger with an amusing TikTok video is to express one's self via digital avatar to a network of followers or friends; but it's also to represent oneself as a data point to the underlying audience of the algorithm."1 We are data assets now. We know this. How, then, to marry this knowledge of digital capitalism to the cultural and sociological study of reading practices?

Kilbane argues that "a literary criticism trained-up for the platform age will also require scholars to comprehend how these aesthetic forms, as products of their platforms, express the underlying logics and emergent protocols of those platforms," revealing, as they mediate, the "designed core architecture" of living one's life as a cultural consumer on social media.2 The essays we assembled extend this claim to the study of reading in particular, to consider how the activity of reading, both personal and social, is shaped and moderated online, as well as how it is depicted, curated, debated, and observed. Contributors explore new methods to "read with algorithms" themselves: algorithms as sociotechnical apparatuses that interact with texts to produce meaning. And, understanding reading as a social act, they endeavor to better understand how individuals, communities, and genres have learned to read alongside the algorithms that structure daily practices.

The cluster of works that came together reveals entwining concerns: about the flow and organization of information online; the construction of the data subject through and within reading practices; the gendered and racialized performance of self online; and the forms of labor, often hidden, that undergird it. They were linked, as well, by critical questions about how we study algorithmic reading practices — what is it to understand reading as collective, when algorithms themselves fracture reading communities into digestible trends and data points. Across the works in the cluster, three major methodological approaches emerged in response to the shifting relationships with reading that algorithms introduce: (1) an emphasis on quantitative approaches, which allow us to better think like algorithms; (2) an increase in reflexive approaches that center the researcher's own positionality, acknowledging how algorithmic curation individualizes corpuses; and (3) a quest for more creative modes of expression that can help us consider, across platforms, what it means to talk/read/write/think with algorithms. Interweaving critical and creative frameworks with qualitative and quantitative methodologies deepens our understanding of how algorithms shape reading processes and, consequently, cultural ideas about reading itself.

Reading with algorithms changes our categorization processes, while often reinforcing the underlying power structures that normalize certain categories of reading and of literary production. In her essay for this cluster, Claire Parnell studies the centrality of datafication to Wattpad, as a key contemporary publishing platform whose business model is premised upon "personalisation, through sometimes hyper-specific personal genre assemblages and microgenres." Parnell describes how Wattpad users cultivate these "personalised genre assemblages within the platform's curatorial systems," reading with and also away from algorithms. Wattpad material is classified both by "top-down taxonomy and user-driven folksonomy," resulting in the curation of affectively resonant microgenres that orient how readers experience an otherwise dizzying tangle of content pathways. Similar processes shape reading practices, attitudes, and norms differently across different platforms: on BookTok, for instance, the emphasis on recommendation videos creates a focus on the material culture of bookishness, and the performance of emotional connections to books; for recommenders on Goodreads, like the participants in Danielle Fuller and DeNel Rehberg Sedo's study, the knowledge of algorithmic sorting and categorizations leads to particular writing and self-presentation strategies. Fuller and Rehberg Sedo identify genre as related integrally to algorithmic sorting: online reading cultures are highly based in recommendation, and recommendations often sort themselves in terms of genre ("if you liked this work about gay wizards, let me recommend . . ."). The readers they observe experience algorithms as banal realities of online life, which they at times use to their advantage, finding like-minded readers whose interests are never ultimately reducible to one hashtag or genre. As Shweta Khilnani has shown, on Instagram personalization takes the form of decontextualization and abstraction: poems circulate especially well, and become replicable across platforms, when they appear relevant to different contexts, adaptable for anyone seeking a lyric voice with which to identify.3 These forms invite their own reproduction: "it's hard to write a good poem," Jacquelyn Ardam writes, "but it's easy to write a Rupi-ish poem."4 The "Rupi-ish" poem, a micro-genre of Instapoetry, invites imitation, clustering cultural production and facilitating algorithmically-mediated classifications — conditions that make variation itself meaningful. Like algorithms, a genre can sort and classify, but Fuller and Rehberg Sedo see this less as a matter of constraint than of variegated conditions.

At their core, algorithms categorize. In categorizing, they create boundaries and norms. And as the categories through which we read shift, so do the forms through which we relate self to other. This becomes clearest in the way the datafication of reading practices — the individualization of recommendations, entwined with a knowledge of surveillance — shift our ideas about subjectivity alongside our reading practices. Analyzing individual Kindle user data, as the Privacy Settings Collective does, reveals some of the barriers to our understanding of how exactly our reading practices are being observed and monetized. "By bringing digital reading under experimental conditions and then reflecting on the harvested data," they write, we can nonetheless better glimpse the "data subject": "the amalgamation of analyzable metrics, statistics, and crude figures that we become when registered and processed by legally permissible regimes of data rendition." If reading is intimately tied to the construction of the datafied self, then reading for an algorithmic audience requires that users modify their practices to produce the desired sense of self.

Evaluating how our own individual positions — or, in this case, IP addresses — shape our experience of a platform reminds us to challenge the taken-for-grantedness with which we approach shared knowledge. Amazon naturally looms large across the cluster, as the corporation with the biggest stake in the future of reading. Examining Amazon Crossing, which is now the leading US publisher of translations into English, Anna Muenchrath explains how harvested data helps determine what is selected for translation. She reads one representative translated novella, Sora Kim-Russell's 2015 translation of Bae Suah's Nowhere to Be Found, noting how its "algorithmic gaze" queries putative distinctions between the self and the data subject. Subjectivity is shaped, for Meunchrath, by our knowledge that we have been made into subjects of algorithmic analysis. As Meunchrath's essay suggests, our transformation into data subjects transforms our sense of self and subjectivity by emphasizing divisibility, customizability, transformability. This problem of the self emerges in the rise of popular genres like the personal essay: Charles Pidgeon's essay illustrates how the personal essay arises alongside and through emerging ideas about the role of information and information organization — ideas that transform the organizing paradigms of our lives. Pidgeon argues that writers of "online essays" weigh needing to court view-optimizing virality against wanting the prestige that might be threatened if readers perceive them as too interested in SEO. Literary prestige becomes, here, the appearance of disinterest in the power of datafication.

There is a recourse to literary traditions within these spaces that does not quite equal, or that exceeds, a simple matter of ongoing veneration via the circulation and accrual of cultural capital. It is certainly true that, as in the phenomenon of dark academia, online readerships can uphold a venerated literary tradition that is remarkably white and politically outmoded — constructing engagement with the literary canon, for instance, as a sign of one's superior cultural status or apprehension. This is one of the functions of bookishness, as Jessica Pressman describes it,5 in which digital expressions of interest and investment in the physicality of books help the user curate a particular kind of self for online consumption. Professions of nostalgia for a time before social media can be an attempt to form communities of opposition to the pressures of the present. These communities are of course reliant (knowingly, despite themselves) on the very networks of platform capitalism that cause unease and exhaustion: they can't be avoided if one seeks the kinds of online life that can also console and enliven. In this light, bookish social media is simultaneously a stance of disavowal and a means of participation — or, a means of ambivalent participation that derives something like enjoyment from disavowing the medium itself.

The socialization of reading practices online — whether through cross-platform reading challenges, review integration on reading sites, the production of brand-specific tags, and across it all, the erasure of much of the labor of producing bookish content — turns reading into even more of a performance of cultural capital. Reading with algorithms means reading publicly: as Andrew Piper wrote in 2012, our reading "may not be 'public' (i.e., on display), but it is being read."6 And our knowledge of being read (by algorithms, for example) regularly shapes not only residual and dominant reading practices, but emergent practices, too: as any well-versed social media user knows, ensuring the algorithm "feeds" you what you want to see requires reading, and scrolling, strategically. As reading's place as a form of socialization and a marker of cultural capital takes center stage, most reading now mirrors the elementary school reading challenge: made quantifiable and performative, and implicitly public or publicize-able, what and how we read is intimately tied to the endless construction of selfhood online, often compulsively and anxiously, within broader social norms.

That strategy, in turn, requires a deliberate reflection on who you want to be, or to be seen to be. If, as Leah Price argues, it is possible to "Show me how you want to read, and I'll show you who you want to be," then what we show to our algorithmic readers reveals who we want to be, as data subjects — and who we want to be, unsurprisingly, reproduces hegemonic standards.7 In her essay for the cluster, Rachel Wilson studies the Rory Gilmore Reading Challenge in its current BookTok-driven phenomenon. Thousands of TikTok videos prompt viewers to read all (a term that Wilson shows has a surprisingly unstable meaning) of the books featured on Gilmore Girls across its many seasons. An activity upholding certain canonical literary verities, whose appeal is burnished further by Rory's own (white, middle-class, Yale-bound) beauty and intellect, the media attending the challenge offers its participants access to the performance of that same vision of reading as a space apart from life — a space of middle-class leisure and comfort. On TikTok, though, books also become props in another kind of artistic collage, the video whose construction requires aptitudes well outside the usual literary skill set. This includes activities such as, in Simone Murray's words, "editing still images, film snippets, soundtracks and captions," to figuring out how to "tag their posts profusely to maximise discoverability" and to "game a platform's algorithm through scheduling optimal release times" combined with intensive sustained engagement with users.8 This makes the valence of concentration itself unique, emphasizing books' materiality in the face of digital abstractions.

Producing content for social media's book cultures, including the simple acts of liking a video or recommending a book, can also be a labor of love: as an expression of an authentic commitment to certain cultural practices, it is exploitable — monetizable in the form of info about the "data subject" — precisely because that commitment reflects real needs to find help in managing the nature of daily life. In a meditation on turning to TikTok in the early days of the pandemic, Charlotte Shane writes that "There is no escape from what life is like, but there are momentary reprieves from the worst of it, and those reprieves supply the gladdening required to endure." However hard she found it to be consoled by real present humans in the moment of tragedy and loss, precisely because of the algorithm's affordance of predictability and surprise, its scripted flexibility, its "specificity of variation," TikTok could find you, despite yourself, "tricked . . . into happiness." You don't know when the right video will come along, the one that can serve as that "depression prophylactic" and induce the laughter that reminds you that "the conditions for sadness are abundant but not total."9 But you can count on its eventual arrival.

This "trickery" of joy is a key part of algorithmic reading practices. Comfort requires erasing the labor that produced it. Production disappears into consumption, as the many hours required to produce a polished video disappear into a performance of ease aligned with the frictionless ideology of the digital. In her cluster essay, Lindsay Thomas writes that a large niche of BookTok videos feature "a sentimental white feminine aesthetic," gesturing toward "a throwback world in which book buying, particularly purchasing fiction, still functions as a marker of white middle-class belonging (and belongings)." At the same time, though, Thomas indicates, BookTok videos are quite simply work, often passionately undertaken. They take tremendous time and effort to construct. Through them, creators can receive money and products (especially boxes of books and reading accoutrements). Unpaid users and barely-compensated readers do more and more of the work that the book industry once assigned to marketers. In this respect, any social media production of book cultures inevitably reflects the withering away of the conditions that once secured middle-class "belongings," alongside the kind of literary prestige that was built upon that material foundation of relative wealth and leisure. These conditions, for many in Instagram's or TikTok's vast userbase, are sought after through participation in the digital hustle. The physical book becomes, again, a prestige display object remediated for digital content production and circulation. Even as they avoid thematizing work itself as a facet of online life, the BookTok videos that Thomas studies distill "hours or days of reading into just tens of seconds of carefully edited footage." In this, they show reading: thinking about one's reading, and cultivating a community around reading, as activities that take time — in conditions in which time itself is such a diminished and unevenly distributed resource.

A fantasy of effortless engagement, tied to particular models of white, middle-class femininity, dominates at the surface level of online reading cultures. Beth Driscoll and Claire Squires explore the feminization of reading practices in their creative contribution, in which their "post-data, activist and autoethnographic" Ullapoolist method sees them "drawing on situated knowledge that enacts creative critique and playful experimentalism." Written as a television chat segment called "Breakfast with AlgoBooks," their piece reflects on the culture of TV book club recommendation as "domestic, feminized, and warmly relational." Such algorithms create a vision of easefulness that functions only insofar as it erases the laboriousness of its own production. An algorithm's work, like any good celebrity's, is to model a good life — a life whose quality is ensured by purchasing sponsored products, or engaging with the proper platform. 

The fantasy of algorithmic decision-making reflects broader shifts over the past decade, as the dominant image of the digital age has moved from the network to the cloud, producing visions of floaty, effortless immateriality. Algorithmic decision-making carries this imagined effortlessness and relief from agency one step further, promising to reduce friction in everyday life by making decisions for us. Can't decide what news source to trust? Google will mark the best options. Can't decide what to read next? Check out a Twitter trend. Algorithms promise to take on decisions for us, reducing cognitive overload, reducing friction in everyday life — the small pulls of care and concern that take us away from the task at hand — so that we can enter flow space. The human labor that goes into these decisions, whether in the manual curation of trends or the slow automation of machine learning, is erased, implying a perfectly automated cultural world that can ease our daily life.

This invisibilized labor is linked to a broader devaluation of femininized cultural production and consumption, a long-standing phenomenon that has been reinforced online, and in particular in online reading communities. The hysterical BookToker — a figuration examined in Lindsay Thomas's and Rachel Wilson's pieces — moved to tears over her favorite book, becomes a channel for the overwhelmingly affective consumption of "essential" reading. Like books, algorithms grab us, and in doing so, they promise to release us so that we might continue searching, reading more and thus becoming more like what we (and the algorithms) imagine we are. Reading with algorithms, then, requires experiments in power and release, narrative and subjectivity, connection and isolation, categorization and differentiation. Together, the essays in this cluster suggest new approaches to reading with algorithms, and of understanding the production of the self within and against the datafication of daily life. 

---

---

Susanna Sacks (@susannalsacks) is an Assistant Professor of English at Howard University, and her research examines the influence of digital publication and transnational institutions on African poetic forms and networks in the twenty-first century.

Sarah Brouillette (@brouillettese , @sarahbrouillette.bsky.social) is a Professor in the Department of English at Carleton University in Ottawa, Canada.

---

References