Simple information constraints can push biological collectives into self-reinforcing consensus that becomes detached from the outside world.
Cornell researchers modelled agents such as animals in groups and cells in tissues, relaxing two unrealistic assumptions: that individuals know the evidence behind neighbours’ decisions and can attend to every neighbour simultaneously. With only actions visible and attention limited, feedback can produce dysfunctional echo chambers. Peripheral individuals with independent environmental information can break the loop, as can a small fraction of agents that occasionally stop following the group.
This identifies a failure mode for decentralised intelligence that is directly relevant to Adaptive Emergent: local agreement is not necessarily evidence of good collective sensing. More interestingly, the model suggests architectural countermeasures. Agents at the network edge can preserve independent environmental input, while occasional refusal to follow social information can restore responsiveness. Diversity of information access and controlled non-conformity may therefore be functional requirements rather than noise in distributed systems.
https://news.cornell.edu/stories/2026/09/study-reveals-how-echo-chambers-form-and-what-prevents-them
