Platform Governance and Research Data Access
The Collapse of an Infrastructure
The papers gathered here describe, from many angles, the unraveling of a particular research settlement: the roughly fifteen-year period in which social scientists could study public communication through relatively open platform APIs. Freelon2024-sc offers the most systematic genealogy, periodizing data access from the laissez-faire era through the current “post-API age” of pay-to-play and walled gardens. Murtfeldt2025-wu supplies the empirical eulogy: a median 25% annual growth in Twitter-based research through 2022, stagnation in 2023, a 13% decline in 2024, and over 33,000 studies whose mode of production is no longer reproducible. Yang2026-tq confirms the pattern at a broader disciplinary scale — the share of empirical social media studies plateaued and fell in 2022–2024 as Twitter/X and Facebook collapsed as data sources. Taken together, these papers establish that what has happened is not a contraction but a structural rupture in the infrastructure of social knowledge.
The DSA as a Partial, Contested Repair
Article 40 of the EU Digital Services Act is the principal institutional response to this rupture, and several papers in this collection assess what it actually delivers. Ohme2026-nv frames Article 40 as a substantive change in practice, while Pierri2025-hm — written from inside the EU pilot process — emphasizes how legal asymmetries, resource gaps, and platform resistance threaten implementation. Philipp2026-tl surveys the patchwork of DSA-mandated APIs and finds uneven scope and eligibility; Entrena-Serrano2025-gw documents that TikTok’s expanded Research API still fails its transparency mandate; and Rieder2025-ju shows that even YouTube’s long-standing search API is “forgetful by design,” systematically under-retrieving older content in ways that defeat retrospective DSA-style auditing. Peters2026-mo takes the analysis a level higher, showing that “data quality” itself was largely absent from the DSA’s drafting and had to be inserted by academic and NGO advocacy against platform pushback — with Meta strategically inverting the argument to claim that low data quality made researcher access pointless. The picture across these papers is consistent: the DSA is the most serious regulatory intervention to date but operates against deep platform inertia, and compliance gestures routinely fall short of substantive transparency.
Industry Influence and the Compromise of Independence
A second strand argues that even where access exists, it is shaped by structural dependencies that compromise the resulting science. Bak-Coleman2025-pm makes the agnotological case at full strength, drawing parallels to tobacco and pharma research; Bak-Coleman2026-mk quantifies it, estimating ~80% “industrial saturation” of high-profile social media papers once author, editor, and reviewer ties are aggregated. Heiss2026-qv develops the same argument in commentary form, while Munger2025-cz turns the critique inward on the celebrated Meta2020 collaboration, arguing that its design tradeoffs — strict preregistration, multiple-comparison adjustments, platform-mediated interventions — biased the project toward nulls convenient to Meta. Allen2025-ot reads the same situation more optimistically, presenting platform-independent browser-extension methods as a way to evade industry capture. The disagreement here is real: whether to repair platform-academic collaboration or to route around it.
Routing Around: New Methodological Repertoires
Several papers exemplify the methodological adaptations that have emerged. Bouchaud2026-lr uses data donations to reconstruct X’s recommender embedding space; Inacio-da-Silva2026-zf deploys volunteer browser plugins to audit Brazilian electoral ads; Efstratiou2025-gs compares algorithmic to chronological feeds via real users’ donated data; Bastos2025-ya and McNally2025-dn (the same study reported twice in the corpus) exploit the now-defunct CrowdTangle archive for longitudinal algorithm auditing, explicitly framing their method as a model for DSA Article 40(4). Jurg2025-ur combines API sampling with browser scraping to audit YouTube’s election-time moderation. These works share a recognition — most clearly articulated in Rieder2025-ju — that APIs are now neither neutral nor sufficient research instruments, and that triangulation across donation, scraping, and limited official access has become the new baseline.
Content Moderation and Its Discontents
A parallel literature tracks what platforms have done — or stopped doing — under their own governance regimes. Donovan2025-ws offers a historical reconstruction of how “misinformation-at-scale” was constructed as a moderation problem between 2016 and 2021 and then deconstructed after political backlash. Moran2025-qn documents the contemporaneous gutting of trust-and-safety professions; Cazzamatta2026-lo examines fact-checkers’ position after Meta ended its Third-Party Fact-Checking Program; Farkas2026-lr and Shi2026-ko (effectively the same study) analyze how European fact-checkers rhetorically defend platform partnerships even as those partnerships collapse. Hurcombe2025-cs and De2026-ld both treat platform discourse itself as an object of analysis, showing how Meta’s Newsroom and platform “changecraft” frame interventions to displace responsibility outward. Tonneau2025-bv documents the linguistic inequities of the moderation workforce that DSA reporting has made newly visible, while Vincent_undated-re illustrates the recurring measurement regimes the DSA now enables. Rieder2026-pp and Simeone2025-vo examine the limits and ripple effects of moderation actions — ambient ideological circulation after deplatforming on YouTube, and durable network restructuring on Twitter.
Political Consequences of Platform Governance
Several papers connect platform governance directly to democratic outcomes. Ventura2026-yc shows that Brazil’s X ban produced a durable rightward “sorting ratchet” in the platform’s information environment; Wang2026-ub documents the failed migration of academic Twitter to Mastodon, suggesting that the costs of governance shocks fall asymmetrically on scholarly communities. Lewandowsky2026-ob makes the broad accountability claim explicit, while Schiffrin_undated-gi extends the gatekeeper-liability frame from misinformation to AI-enabled fraud. Votta2025-xz and Mahl2026-hc situate these dynamics within broader expert and regulatory debates — the latter’s Delphi study finding that scholars now rate platform governance as the single most important intervention area, above literacy and fact-checking.
A Paradigm in Question
The most ambitious papers argue that the very object of study is dissolving. Tornberg2026-lc and Tornberg2025-ir (the same essay) propose a “post-social media studies” reckoning with algorithmic broadcasting, semi-private spheres, and AI-mediated communication as distinct emergent formations. Helmond2026-ll echoes this from the platformization tradition, reading Meta’s retirement of the Like button as the end of the social-graph paradigm and the dawn of “Big AI.” If they are right, the data-access crisis is not merely an infrastructural problem to be repaired by Article 40 or by clever browser extensions: the configurations that “social media research” was built to study are themselves passing. The papers in this collection thus trace a double movement — a methodological and regulatory scramble to restore access to a thing that may no longer be quite what it was, even as scholarship begins to articulate what should replace it.