Our experiment creates “multiple worlds,” in which informational messages can be shared among individuals that are connected through network ties. The setup resembles many online platforms, and it allows us to vary network architecture: In half of these online ecosystems, networks were integrated such that contacts were 50:50 liberals and conservatives. In the other half, networks were segregated, such that contacts shared similar political views. The macrosociological design compares multiple realizations of a message’s diffusion process between segregated and integrated networks (see figure). We conducted the online experiment with participants from the United States, where political polarization, in a two-party system, has been on the rise for many years.
Diffusion of a false conservative-leaning message in an integrated (left) or segregated (right) experimental network. Red (blue) nodes indicate conservative (liberal) participants (N=96).
Experimental results reveal that partisan sorting structurally undermines the veracity of information circulating: Network segregation brings together supply of and demand for ideologically aligned news that is believable to biased partisans but otherwise too implausible to propagate. Network segregation thus favors false over true news, increasing the spread of false information in a population. Agent-based models show robustness of this finding across different network topologies and sizes.