Petter Törnberg of the University of Amsterdam has published two new papers and a preprint concluding that social media's core pathologies, including echo chambers, attention inequality, and extremism amplification, are not fixable through platform-level interventions. The problem is not algorithms or non-chronological feeds. It is structural, baked into the architecture itself.
The first new paper, published in PLoS ONE, isolates the echo chamber effect using agent-based modeling combined with LLMs, treating AI personas as proxies for online behavior. This methodological combination is worth examining closely. It is not just the conclusion that matters here, it is how Törnberg is using synthetic agents to stress-test social dynamics that are otherwise nearly impossible to study at scale.
Three linked works now form a coherent body of evidence that social media is architecturally distinct from physical-world communication, with consequences that compound in ways researchers are still mapping. What comes after social media, if anything does, remains an open question. Törnberg does not offer a clean answer, and that is exactly why the full read is worth your time.
[READ ORIGINAL →]