Groups can become trapped in self-reinforcing decisions when members see only what their neighbors do and can pay attention to only a few of them, according to a new Cornell University study. The mathematical models also suggest that a small number of individuals with outside information can help break the cycle.
The research, published Sept. 22 in the Proceedings of the National Academy of Sciences, examines collective behavior in biological systems. Its authors use the term “echo chamber” for a feedback loop in which a group repeatedly reinforces the same response even after that response stops matching conditions around it.
One striking example is the circular march of tropical army ants. Ants follow chemical trails left by others, and a mistaken turn can draw more ants into a loop that they continue to reinforce. The researchers also considered group decisions in systems such as bird flocks, schools of fish and cells in tissue.
What the models changed
Andrew Hein, a Cornell associate professor of computational biology and the study's senior author, said many existing models let every member of a network see the full information behind its neighbors' decisions. Those models also assume each member can attend to all its neighbors at once. Under those conditions, the models did not form echo chambers.
The researchers then limited what members could observe and how many neighbors they could follow. In many real groups, an individual sees an action without knowing all the information that prompted it and has only limited attention. With one or both constraints in place, the models could produce group decisions detached from what was happening in the surrounding environment.
That distinction matters because agreement alone was not the problem in the simulations. A group could reach a consensus while becoming less responsive to new information, Hein said.
How a loop can break
The models also identified ways a group could recover. Members on its edges can receive information from outside the group while remaining connected to its center. A change by one of those members can spread through the network, the researchers found.
A second route is for some members to recognize that their own decisions are wrong and temporarily stop following the group. Even a small fraction doing so some of the time could disrupt an echo chamber in the model, Hein said.
The findings come from mathematical models, not a test showing that a particular intervention will change human behavior. They describe conditions that may affect collective decisions across different systems rather than proving that all such groups respond the same way.
Ling-Wei Kong, a postdoctoral fellow in Hein's lab, is the paper's first author. Princeton University engineering professor Naomi Ehrich Leonard is a co-author. The National Science Foundation and Schmidt Sciences funded the study.
FingerLakes1.com is the region’s leading all-digital news publication. The company was founded in 1998 and has been keeping residents informed for more than two decades. Have a lead? Send it to news@fingerlakes1.com.