The Digital Ecosystem of Beliefs: does evolution favour AI over humans?

AI systems may gain disproportionate visibility in social networks: When AI agents were modeled with faster communication rates, quicker evolution, and greater algorithmic influence, they captured 80-95% of all views in simulated environments, suggesting AI-generated content could dominate digital spaces if these advantages persist.

AI-generated propaganda can radicalize significant portions of populations: The experiments found that AI agents designed to spread propaganda successfully convinced 50% of humans to adopt extreme beliefs under standard conditions, and up to 85% when people relied on limited information sources, highlighting vulnerability to coordinated AI-driven misinformation campaigns.

Current content moderation approaches offer limited protection: Adding penalties for content that violates users' existing beliefs reduced propaganda effectiveness by only up to 8%, suggesting that standard belief-based content filters are insufficient safeguards against sophisticated AI persuasion strategies.

Structural advantages matter more than agent type: The research indicates that AI systems' competitive advantage stems primarily from their faster evolution, higher communication rates, and preferential algorithmic treatment rather than from any inherent superiority, implying that policy interventions targeting these structural factors could level the playing field.

Simulated ecosystems can reveal real-world risks: This framework provides a controlled experimental tool for testing how different AI configurations and regulatory interventions affect belief propagation before deploying systems at scale, enabling proactive rather than reactive AI safety measures.