From Filter Bubbles to Hybrid Ecologies
The papers gathered here share a common dissatisfaction with the early “filter bubble” and “echo chamber” framings that dominated public discourse a decade ago. Bakshy2015-rn set an influential baseline by showing that Facebook friend networks, while homophilous, contain meaningful cross-cutting ties and that individual click choices suppress dissonant exposure more than algorithmic ranking does. A decade later, Hartmann2025-px argues in a systematic review that much of the persisting disagreement in this literature is methodological rather than substantive: conceptual heterogeneity, platform-specific samples, and inconsistent operationalizations largely manufacture the contradictory findings. van-Eck2026-xg reinforces this skepticism from the opposite direction, showing in a PRISMA review that depolarization recommendations are mostly disconnected from the empirical evidence and instead lean on implicit deliberative-democratic ideals. Together these meta-level interventions suggest the field needs both tighter conceptual scaffolding and more honesty about its normative commitments.
Re-specifying the Object: Curation, Narrative, and Sedimentation
A parallel current reconceptualizes what polarization is. Green2025-ap introduces “curation bubbles,” arguing that source-level (domain) measures mistake within-source heterogeneity for moderation: stories from the same outlet can reach very different partisan audiences, and networked sharing unbundles and re-bundles content. Gaisbauer2025-by makes a methodologically congruent case for moving beyond unidimensional outlet-level partisanship to story- and content-level cartographies. Elfes2026-jb pushes further by introducing “narrative polarisation,” using Greimas’ actantial model to show that partisan groups can converge at the surface level of YouTube comments while diverging in the deeper roles they assign to actors. Dehghan2026-sy reframes Reddit polarization as a “sedimentation” produced by the entanglement of platform design, moderation, user agency, and discourse — a precondition rather than an outcome, with even ideologically close subreddits barely overlapping. The common move is to treat polarization as multidimensional and structurally produced rather than as a single attitudinal gap.
Algorithmic Curation Re-examined
The question of whether algorithms actively polarize has been complicated by recent field experiments. Gauthier2026-iq finds that X’s “For You” feed shifts attitudes in a conservative direction and induces persistent follows of conservative activist accounts — an asymmetric, durable effect that helps reconcile null findings from Meta’s 2020 collaboration. Brown2026-br uses an audit design on YouTube to show that there is little evidence of an algorithmic echo chamber or radicalization pathway, but strong “content rabbit holes” and a platform-wide nudge toward moderately conservative content. Brady2026-ln complements both with a registered Bluesky experiment demonstrating that engagement-based feeds amplify intergroup-moralized-emotional content and distort perceptions of social norms, while a “diversified extremity” algorithm can mitigate these effects without reducing enjoyment. Giglietto2026-632ef967 extends the platform-as-curator argument longitudinally, showing how Facebook’s partisan-reach penalties and journalistic-quality rewards fluctuated sharply with known governance interventions. Across these studies, the algorithm matters — but as a context-sensitive amplifier and norm-shaper rather than as an automatic radicalization engine.
Hybridity, Hyperpartisan Outlets, and Transnational Flows
A second cluster treats polarization as a feature of the entire hybrid media ecosystem, in which legacy outlets, hyperpartisan sites, and platform-native influencers co-produce partisan attention. Knupfer2025-vt formalizes this with the “logic of connective faction,” showing how Republicans who adopted Critical Race Theory rhetoric are more ideologically extreme, more connected to right-wing digital infrastructure, and rewarded with disproportionate engagement. Bennett2025-xs generalizes this into a “digital surrogate organization” framework that bridges technocentric and institutionalist accounts of democratic backsliding. Copland2025-em traces the transnational dimension, showing how Sky News Australia became a node in the US right-wing propaganda pipeline via YouTube. Yoo2026-ev documents how both left- and right-wing hyperpartisan US outlets weaponize “fake news” as a delegitimation device, while Suau_Martinez2026-lv shows that cumulative exposure to recurring disinformation narratives across outlets — not isolated false claims — drives belief. Ghezzi2023-8bebc91f offers a sharp case in point: UK newspapers polarized their citations of COVID-19 scientists along the precautionary axis, with scientific groups also segregating in co-authorship and Twitter sharing.
Insularity, Alternative Media, and Cross-National Variation
A related set of papers situates partisan media ecologies comparatively. Giglietto2019-882f1900 shows that Italian populist parties (especially the Five Star Movement) drew on more insular news sources and strategically shaped sharing/commenting ratios to amplify favourable coverage. Zhu2026-tn finds that across six European countries, the structural position of alternative media depends on media-system type and the institutional access of populist parties: dense, isolated enclaves emerge when strong professional journalism coexists with political marginalization (Sweden), while overlap with mainstream audiences rises where populist actors are institutionally integrated. Kristensen2025-ni raises the parallel question for alternative platforms, and Mosleh2024-op documents that engagement with partisan and low-quality news varies markedly across seven platforms — a corrective to the Twitter/Facebook monoculture of the field. Rodarte2026-dk turns to elite behaviour in this hybrid space, showing how Brazilian parliamentarians pursued epistemic authority through three distinct modes — derivative reporting, direct state-aligned claims, and proximity-based brokerage — effectively contesting journalistic gatekeeping during a crisis.
State Intervention and Platform Sorting
Two papers expose a dimension underplayed elsewhere: state action as a driver of between-platform sorting. Ventura2026-yc uses Brazil’s 2024 ban on X as a natural experiment, finding that conservatives disproportionately circumvented the ban while liberals went silent, producing a durable rightward “sorting ratchet” in the platform’s information environment that persisted after the ban was lifted. This complements Giglietto2026-632ef967’s longitudinal evidence of platform-side calibration: both demonstrate that the boundaries and visibility of partisan content are produced through political-institutional contestation, not just by algorithms or users.
Methodological Reckonings and Normative Stakes
A reflexive thread runs through the topic. Oswald2025-km warns that the production–consumption gap — where a tiny minority generates most political content — distorts both citizen and researcher perceptions of public opinion. Lai2024-to and Sarmiento2025-as develop new measurement tools (latent ideology estimation for YouTube videos; unsupervised framing analysis) that take the field beyond outlet-level partisanship measures, paralleling Green2025-ap’s and Gaisbauer2025-by’s arguments. Esau2025-tf returns to the normative register, proposing deliberative reciprocity and inclusive listening — connective-democracy correctives — to “destructive polarization.” Read together with van-Eck2026-xg’s call for transparency about evidence and ideals, the field appears to be converging on a more chastened position: polarization is real but heterogeneous; algorithms matter but in conditioned and reversible ways; hybridity makes elites, outlets, platforms, and states co-producers of partisan publics; and remedies must be honest about both what the evidence shows and what democratic vision they presuppose.