Free Information Disrupts Even Bayesian Crowds
Agent-based model proves free information exchange harms group accuracy, even with perfectly rational 'Bayesian' agents.
A team of researchers including Jonas Stein, Shannon Cruz, Davide Grossi, and Martina Testori has published a provocative paper titled 'Free Information Disrupts Even Bayesian Crowds' (arXiv:2604.01838). The study uses a computational agent-based model to challenge a foundational assumption of modern communication networks: that unlimited, free information exchange is inherently beneficial. The researchers modeled a group of idealized agents—each a perfect Bayesian reasoner, truth-seeking, and cooperative—and found that even in this optimal scenario, unconstrained sharing of information led the collective to form less accurate beliefs about a topic.
The core finding is counterintuitive: more information, freely exchanged among perfectly rational agents, does not guarantee better collective understanding and can actively make it worse. The paper argues that if this detrimental effect occurs in a model with such idealized agents, the risk is significantly higher in real-world settings like social media platforms, where users are not perfectly rational. The authors conclude that the design of high-impact communication networks must move beyond the dogma of absolute informational freedom and carefully consider implementing constraints or structures on information flow to preserve epistemic health.
- Agent-based model shows unconstrained info exchange harms group accuracy, even with perfect Bayesian agents.
- Challenges the core design principle of unlimited sharing on platforms like social media.
- Authors argue for deliberate constraints on information flow in networks with major societal impact.
Why It Matters
Provides a scientific basis for rethinking platform design, moving from 'more information is always better' to considering structured, healthy information ecosystems.