Human-AI adaptive dynamics drives the emergence of information cocoons

Jinghua Piao, Jiazhen Liu, Fang Zhang, Jun Su, Yong Li

Abstract: Despite AI-driven recommendation algorithms being widely adopted to counter information overload, substantial evidence suggests that they are building cocoons of homogeneous contents and viewpoints, further aggravating social polarization and prejudice. Curbing these perils requires a deep insight into the origin of information cocoons. Here we investigate information cocoons in the real world using two large datasets and find that a large number of users are trapped in information cocoons. Further empirical analysis suggests that two ingredients, each corresponding to a fundamental mechanism in human–AI interaction systems, are correlated with the loss of information diversity. Grounded on the empirical findings, we derive a mechanistic model for the adaptive information dynamics in complex human–AI interaction systems governed by these fundamental mechanisms. It allows us to predict critical transitions between three states: diversification, partial information cocoons, and deep information cocoons. Our work not only empirically traces real-world information cocoons in two representative scenarios, but also theoretically unearths basic mechanisms governing the emergence of information cocoons. We provide a theoretical method for understanding major social issues resulting from adaptive information dynamics in complex human–AI interaction systems. It is widely known that AI-based recommendation systems on social media and news websites can isolate humans from diverse information, eventually trapping them in so-called information cocoons, where they are exposed to a narrow range of viewpoints. Li et al. introduce an adaptive information dynamics model to uncover the origin of information cocoons in complex human–AI interaction systems, and test their findings on two large real-world datasets.