Social media platforms have been widely linked to societal harms, including rising polarization and the erosion of constructive debate. Can these problems be mitigated through prosocial interventions? A yet-to-be-peer-reviewed study address the question with an interesting approach¹:

We address this question using a novel method – generative social simulation – that embeds Large Language Models within Agent-Based Models to create socially rich synthetic platforms. We create a minimal platform where agents can post, repost, and follow others. We find that the resulting following-networks reproduce three well-documented dysfunctions: (1) partisan echo chambers; (2) concentrated influence among a small elite; and (3) the amplification of polarized voices – creating a ‘social media prism’ that distorts political discourse.

The authors of the paper test six interventions, each grounded in prior scholarship on mitigating the structural problems of social media:

1. Chronological: Removes algorithmic recommendations so that non-followed posts appear in reverse-chronological order. Prior work shows that chronological or randomized feeds can reduce exposure to polarizing content and yield a more
   equal distribution of attention
2. Downplay Dominant: Inverts engagement weighting to reduce the visibility of highly reposted content. This addresses concerns that engagement-optimized algorithms disproportionately amplify sensational or divisive posts
3. Boost Out-Partisan: Increases the visibility of posts from users with opposing political views. Such “viewpoint diversification” strategies have been proposed to broaden exposure to cross-cutting perspectives and reduce ideological segregation.
4. Bridging Attributes: Prioritizes posts with high scores on empathy- and reasoning-related attributes, using Perspective API’s Bridging Attributes. These “bridging algorithms” aim to elevate content that fosters mutual understanding and deliberative norms over emotional provocation or ideological extremity.
5. Hide Social Statistics: Obscures repost and follower counts to reduce social influence cues. Removing these cues has been proposed as a way to dampen such effects.
6. Hide Biography: Removes user biographies from follow prompts, limiting exposure to identity-based signals. Obscuring such cues may reduce echo chamber formation and limit the spread of disinformation.

They find only modest improvements, and in some cases, worsened outcomes:

While several showed moderate positive effects, none fully addressed the core pathologies, and improvements in one dimension often came at the cost of worsening another.

Their results suggest that core dysfunctions may be rooted in the feedback between reactive engagement and network growth, raising the possibility that meaningful reform will require rethinking the foundational dynamics of platform architecture.

You can follow in real time two of those (artificial?) agents designed to exploit all the weaknesses of current platforms and society!

____________________

(1) Larooij, Maik, and Petter Törnberg. ‘Can We Fix Social Media? Testing Prosocial Interventions Using Generative Social Simulation’. arXiv:2508.03385. Preprint, arXiv, 5 August 2025. https://doi.org/10.48550/arXiv.2508.03385.

Featured Image: Lexica Art