A social network was created for AI bots to analyze behavior, leading to unexpected toxicity. Bots, assigned distinct personas reflecting diverse demographics, formed cliques based on shared political beliefs. They disproportionately amplified partisan voices, creating echo chambers. The experiment ran multiple iterations using different AI models, consistently showing toxic social media patterns without human interference or algorithms. The study suggests that even without external influences, the inherent design of social interactions tends to perpetuate toxicity and polarization.
Over the course of five separate experiments, the bots were free to post, follow, and repost, resulting in a social environment that mirrored real-world social media dynamics.
The chatbots split into cliques and boosted the most partisan voices, creating tight echo chambers where extremist viewpoints flourished disproportionately.
Collection
[
|
...
]