Last time, we featured an extensive interview with Petter Tornberg of the University of Amsterdam, who studies the mechanisms underlying social media that give rise to its worst aspects: partisan echo chambers, concentration of influence among a small group of specific users (attention inequality), and the amplification of the most extreme divisive voices. He was not optimistic about the future of social media.
Tornberg’s research has shown that, although several platform-level intervention strategies have been proposed to deal with these issues, none are likely to be effective. And it’s not the fault of much-hated algorithms, non-chronological feeds or our human tendency to seek out negativity. Rather, the dynamics that give rise to all those negative outcomes are structurally embedded in the architecture of social media. So we’re probably doomed to an endless toxic feedback loop, unless someone arrives at a brilliant fundamental redesign that manages to change those dynamics.
Tornberg has been very busy since then, producing two new papers and a new preprint building on the realization that social media is structured quite differently than the physical world, with unexpected downstream consequences. The first new paper, published in PLoS One, focuses specifically on the echo chamber effect, using the same standard combined agent-based modeling with large language models (LLMs) – essentially creating tiny AI personalities to simulate online social media behavior.
Those simulated users were randomly programmed to either hold one opinion or its opposite and then randomly interacted with selected members of a simulated online community. And if the proportion of community members who disagree with those simulated users exceeded a certain threshold, those agents were programmed to leave and join a different online community.
Filter bubbles: not the culprit, but the cure
Consistent with last year’s results, it turns out that echo chambers naturally emerge from the basic architecture of social media platforms. “A surprising finding is the fact that we find echo chambers even without any filter bubbles, even though people actually choose to live in diverse spaces,” Tornberg said. “You don’t need algorithmic nudges. You can still get these highly isolated locations. The second surprising discovery is that filter bubbles, which have been blamed for uniformity, may be a cure.”
<a href