Algorithms as Authors: Rethinking Critical Theory for the Age of the Infinite Scroll
For much of the twentieth century, critical theory operated with a reasonably stable set of objects. A novel, a film, a news broadcast, an advertisement — each of these constituted a text in the broad semiotic sense: a structured artifact that encoded ideological positions, interpellated subjects, and could be read against the grain by a sufficiently equipped analyst. Roland Barthes could dissect a wrestling match. Stuart Hall could decode a news photograph. Louis Althusser could identify the state apparatuses through which ideology reproduced itself. The tools were sharp, and the materials, however complex, held still long enough to be cut.
Social media platforms do not hold still. TikTok's For You Page regenerates itself every few seconds. Instagram's feed is a personalized hallucination assembled from behavioral data, advertiser priorities, and opaque weighting systems. The "text," in any traditional sense, has been replaced by something more like a continuous, individuated stream — and the author of that stream is not a human being but a recommendation engine. Critical theory, as currently constituted, is only partially equipped to analyze this transformation. Understanding why that is, and what new frameworks are emerging to fill the gap, is one of the more pressing intellectual tasks facing scholars of culture today.
Where Classical Frameworks Reach Their Limits
Semiotics, discourse analysis, and ideology critique each encounter specific difficulties when applied to algorithmic platforms. Semiotics presupposes a sign system with identifiable codes — shared conventions through which meaning is produced and received. But the For You Page does not present a shared code. It presents a personalized sequence optimized for individual retention, meaning that two users watching the same platform are, in a meaningful sense, inhabiting different semiotic environments. The "text" is not common; it is bespoke. This makes collective ideological analysis considerably more difficult, since the platform does not interpolate a unified subject so much as it splinters subjecthood into millions of individuated engagement profiles.
Foucauldian discourse analysis fares somewhat better, since Foucault's concept of discourse was never reducible to explicit textual content — it encompassed the institutional rules determining what could be said, by whom, and under what conditions. Algorithmic amplification and suppression can be understood as a kind of computational discourse regulation: the platform's content moderation policies, its shadow-banning practices, and its engagement-reward structures all function to make certain kinds of speech more or less possible. Yet Foucault's framework was designed for the analysis of human institutional actors — clinics, prisons, psychiatric establishments — and adapting it to non-human systems of rule-following requires significant theoretical retooling.
Ideology critique in the Marxist tradition faces perhaps the sharpest challenge. The classical model assumes that ideology mystifies real social relations in the service of a dominant class. Algorithmic platforms certainly serve capital — their business model is the monetization of attention — but the mechanism of mystification is not a coherent worldview imposed from above. It is, rather, an emergent property of billions of micro-interactions, each optimized locally for engagement without any central ideological intention. The result is something more diffuse and arguably more powerful than classical ideology: a system that produces political and cultural effects without requiring a directing consciousness.
Emerging Vocabularies: From Media Archaeology to Platform Studies
Scholars working at the intersection of media studies, digital humanities, and critical theory have begun assembling a new conceptual toolkit. Several frameworks deserve particular attention.
Platform studies, developed in part through the work of scholars like Tarleton Gillespie, shifts analytical focus from content to infrastructure. Gillespie's concept of the platform as a "political economy of logics" draws attention to the ways in which design decisions — what gets surfaced, what gets monetized, what gets removed — constitute a form of governance that is neither purely public nor purely private. For critical theorists, this reframes the platform not as a neutral distribution channel for texts but as an active participant in the production of cultural meaning.
Algorithmic accountability research, associated with scholars like Safiya Umoja Noble and Ruha Benjamin, brings critical race theory into direct contact with platform analysis. Noble's Algorithms of Oppression demonstrated that search and recommendation systems encode and amplify existing social hierarchies, transforming structural racism from a human bias into a technical process. This work is significant not only for its empirical findings but for its theoretical move: it insists that computational systems are not ideologically neutral and that their outputs must be read as cultural texts with political stakes.
The attention economy framework, drawing on the economic sociology of Herbert Simon and later popularized by writers like Tim Wu and James Williams, provides a political-economic grounding for platform critique. Attention, in this framework, is the scarce resource that platforms extract and sell to advertisers. Engagement metrics — likes, shares, watch time, comment counts — function as a new form of cultural capital in Pierre Bourdieu's sense: they confer status, amplify reach, and reproduce existing hierarchies of visibility. The influencer who accumulates millions of followers is not merely popular; they have accrued a form of capital that translates directly into economic value and social power.
Feeds as Narrative, Engagement as Interpellation
Perhaps the most productive theoretical move available to critical scholars is the reconceptualization of the algorithmic feed as a narrative form. Traditional narratology presupposes an author who arranges events into a sequence with beginning, middle, and end. The feed has no author in this sense, no fixed sequence, and no end — it is, as theorist Tung-Hui Hu has suggested, structured more like a network than a story. Yet it produces narrative effects: it creates protagonists (the accounts a user follows most closely), antagonists (the content that provokes outrage), and emotional arcs (the affective rhythms of a scrolling session).
Louis Althusser's concept of interpellation — the process by which ideology "hails" individuals into subject positions — can be usefully, if carefully, extended to algorithmic culture. When TikTok's recommendation engine identifies a user as belonging to a particular demographic or interest cluster, it does not merely reflect a pre-existing identity; it actively reinforces and narrows that identity by continuously serving content that confirms the profile. The user is hailed not by an ideological state apparatus but by a probabilistic model of their own past behavior. The interpellating voice is, in effect, the user's own data.
Toward a Critical Theory of Computational Culture
What would a fully developed critical theory of algorithmic platforms look like? It would need to retain the emancipatory commitments of the Frankfurt School tradition — the insistence that cultural analysis is always also political analysis — while abandoning the assumption that cultural power is exercised through unified, intentional ideological projects. It would need to incorporate the technical literacy necessary to read an algorithm as a cultural artifact without reducing critical theory to computer science. And it would need to grapple seriously with the question of agency: if the feed is the author, the user is simultaneously its reader and its raw material.
The scholars doing this work — Noble, Benjamin, Gillespie, and others working in adjacent fields — are not always recognized as critical theorists in the classical sense. But their project is continuous with the tradition: to make visible the structures through which power organizes experience, and to insist that those structures can be otherwise. The vocabulary is still being assembled. The index, as it were, is incomplete. But the mapping has begun.