The Unraveling of “Social Media” as a Category
The organizing provocation across this cluster is that the very term “social media” may no longer name a coherent object. Tornberg2026-lc makes this claim most systematically, arguing that three simultaneous shifts—algorithmic recommendation displacing the social graph, generative AI producing content without human authors, and users retreating into semi-private enclaves—have dissolved the user-generated-content-through-social-networks configuration the field spent two decades theorizing. Boyd2026-op and Marwick2026-ss arrive at kindred diagnoses from the standpoint of lived platform history rather than infrastructural analysis: boyd proposes retiring “social media” for “parasocial media,” since scrolling and influencer-consumption have eclipsed peer reciprocity, while Marwick’s autoethnography of LiveJournal recovers a lost mode of intimate, filtered, reciprocal writing against which contemporary platforms read as alienating and extractive. Baym2026-tr revisits her own decade-old manifesto and finds every threat she named—inequality, opacity, precarious labor, data extraction—not mitigated but intensified, now personified in a “broligarchy” whose wealth has grown by orders of magnitude. Helmond2026-ll supplies the connective tissue: platformization itself, the concept she coined, has entered an AI-centered phase in which the infrastructural project once organized around capturing behavioral traces from the open web (epitomized by the Like button’s 2026 retirement) is being superseded by inference, synthetic content, and the extension of platform logics into new domains—energy, defense, even orbital compute. Read together, these five essays trace a single arc: from social graph to algorithmic broadcast, from reciprocity to parasociality, from an open web of signals to closed AI inference, with each author converging on the claim that none of this was technologically inevitable but rather the sedimented outcome of political-economic choices.
Extraction, Scam Logics, and the Political Economy of Decline
A second thread pushes past diagnosis of decline toward naming its underlying economic logic. Swartz2026-zb is the sharpest provocation here, arguing that scams are not peripheral to platform economies but constitutive of them—that the same infrastructures enabling creator hustle, crypto speculation, and anti-fraud governance are jointly reshaping economic citizenship, and that “scam” itself is a category doing boundary-work between legitimate and illegitimate capitalism. This resonates with Donovan2025-ws’s account of “misinformation-at-scale” as a built-in feature of engagement-driven business models rather than a correctable bug, and with Vertesi2026-lv’s argument that AI critique is systematically distracted by “decoys”—ontological, inevitability, disruption, safety, regulatory—that obscure the underlying “Project of AI” as a world-building assemblage of capital and infrastructural power. De2026-ld complements this political-economy reading with a discursive one: platforms actively narrate their own transformations (“changecraft”) through aesthetics of visible innovation, continuity-framing of ideological pivots, and incremental patchworking, consolidating power while performing responsiveness. Hurcombe2025-cs finds the same rhetorical machinery in Meta’s Newsroom specifically, where floating signifiers like “coordinated inauthentic behaviour” externalize blame while eliding the role of platform affordances (Groups, hyperpartisan verified accounts) in spreading harm—showing how platform self-narration can even seed the vocabulary of subsequent regulation. Against this backdrop, Volpe2026-um’s ethnography of micro-influencers reads as a ground-level case study in how individual creators absorb and reproduce these extractive logics, translating platform demands for constant, algorithmically legible output into performances of care and ethical branding that are simultaneously survival strategies within genuine labor precarity.
Governance, Moderation, and Their Limits
A cluster of papers interrogates whether platforms, states, or researchers can actually govern the systems they study. Rieder2026-pp’s analysis of the “Tate-space” on YouTube shows deplatforming as symbolically potent but practically outpaced by recommendation-driven circulation, an ambient ideological ecosystem that reconstitutes itself faster than moderation can remove it. Donovan2025-ws traces the broader historical arc of platform moderation from non-intervention through COVID-era behavioral intervention to post-2021 retreat, underscoring that trust-and-safety capacity is itself a fluctuating, politically contingent resource. Pierri2025-hm shifts the lens to regulatory infrastructure, reporting from inside the EU’s Digital Services Act implementation on the asymmetries of power and resources that hobble even a landmark transparency mechanism, and flagging large language models as a looming regulatory blind spot. Lewandowsky2026-ob and Bennett2025-xs connect platform dynamics directly to democratic backsliding: Lewandowsky names platforms as accountable actors in authoritarian drift, while Bennett and Livingston’s “communication-as-organization” framework argues that digitally networked publics have become de facto “digital surrogate organizations” pulling parties toward illiberalism—a corrective to both technocentric disinformation research and institutionalist accounts that ignore networked organization. Bruns2026-yv’s history of Twitter’s “death” synthesizes several of these threads into a single platform biography: an initially prosocial, open architecture progressively eroded by mismanagement, weaponized disinformation, and ultimately Musk-era “enshittification,” with successor platforms (Mastodon, Threads, Bluesky) reproducing rather than solving the underlying governance failures. Kristensen2025-ni gestures toward the same question of whether alternative platforms merely displace rather than resolve discursive polarization.
Algorithms, Attitudes, and the Politics of Feed Design
Several papers offer empirical purchase on how algorithmic curation itself produces political effects. Gauthier2026-iq’s field experiment on X finds that switching users onto the algorithmic feed shifts attitudes rightward in an asymmetric, persistent way—because the algorithm changes who users follow, not just what they see in the moment—helping reconcile earlier null results from Meta’s 2020 studies. Efstratiou2025-gs complements this with an audit of the pre-rebrand Musk-era Twitter algorithm, showing that apparent right-leaning amplification is better explained by rewards for “agitating” content and proximity to Musk himself than by partisanship per se, with legacy-verified institutional accounts losing visibility as Musk’s own centrality surged. Both papers substantiate, at the level of causal mechanism, the broader narrative traced by boyd, Törnberg, and Bruns: that recommendation systems actively reconfigure the social fabric rather than neutrally reflecting it. Oswald2025-km supplies a structural caveat relevant to all of this work—the “production-consumption gap,” in which a small minority of highly active posters generate the visible discourse that both citizens and researchers mistake for public opinion, meaning even well-designed algorithmic audits risk studying a skewed, unrepresentative sliver of the platform.
The Crisis of Studying Platforms
A distinct but related concern running through this set is methodological and institutional: can research keep pace with, or maintain independence from, the platforms it studies? Munger2025-cz’s meta-scientific critique of the Meta2020 partnership argues that even the most ambitious platform-cooperative research produces narrow, rapidly decaying estimates and inconsistent operationalizations of core concepts, concluding that reactive social science is structurally incapable of governing platforms and that proactive regulatory disclosure is the only viable path. Bak-Coleman2026-mk substantiates the independence worry empirically, finding that roughly half of high-profile social media research has undisclosed industry ties, concentrated among a small elite of repeatedly-engaged scientists and editors, with an estimated 80% “industrial saturation” of the literature once authors, editors, and reviewers are combined—and with industry-tied work skewed toward user-blaming topics like misinformation-sharing rather than platform-level dynamics. Bastos2025-ya and Wang2026-ub document the collapse of research infrastructure from a different angle: Bastos eulogizes the open API ecosystem that made “Twitter studies” possible, while Wang’s survival analysis of academic Mastodon migration shows that even a highly motivated, well-networked cohort of early adopters largely failed to sustain the platform switch, undermining hopes that decentralized alternatives can simply inherit displaced scholarly communities. Pierri2025-hm, again, situates the DSA’s Article 40 researcher-access provisions as a fragile institutional response to exactly this crisis of dependency.
AI as the Emergent Frontier
Nearly a third of this collection reframes platformization’s next decade around generative AI, positioning these essays as a bridge to a successor research agenda. Helmond2026-ll and Tornberg2026-lc both identify AI-driven inference and synthetic content as the mechanism displacing the social-graph paradigm outright. Stanusch2026-ec examines how the AI industry manages its own accountability discourse through “premediation” and “preclusion” during the Sam Altman controversy, while Dodds2026-df extends this critique to journalism, arguing that AI hype functions as a systemic infrastructure redistributing legitimacy and resources rather than mere narrative excess—journalists occupying a contradictory role as both hype-makers and hype-watchers. Hepp2026-oi adds ethnographic depth, distinguishing Silicon Valley’s “loud futuring” (CEO pronouncements) from a “quiet futuring” rooted in Effective Altruist and Rationalist communities that supplies the ideational reservoir for AGI, job-loss, and geopolitical narratives while presenting itself as apolitical. Weinbrand2026-sf traces the same dynamics into Google’s AI Overviews, showing search itself being redefined as a contested site of epistemic authority. Vertesi2026-lv provides the most totalizing framing, urging accountability researchers to look past AI’s discursive “decoys” toward the underlying financialization and network power. Finally, Goldberg2026-eb offers a rare constructive counter-note: a multi-stakeholder position paper proposing that LLM-enabled bridging systems, collective dialogue tools, and AI-assisted community moderation might—cautiously, and not without new risks of synthetic participation and legitimacy erosion—reopen possibilities for more deliberative digital public squares, even as it acknowledges the same structural asymmetries of power and access documented throughout this collection.
Taken as a whole, this body of retrospective and position work narrates a single decade-long arc: an initial optimism about networked sociality curdling into extraction, parasociality, and scam-saturated economies; a governance apparatus perpetually outpaced by the platforms it tries to regulate or study; and a research field forced to reckon with its own structural dependency just as the object of its inquiry—“social media” itself—dissolves into an AI-mediated successor landscape whose contours are only beginning to be theorized.