Opinion dynamics as associative diffusion

A central paradigm in modeling opinion dynamics is based on the notion of “information diffusion” - the idea that opinions spread like viruses through social networks, diffusing between individuals, communities, and countries. Empirical cases of opinion spread on social networks, such as diverging stances towards vaccines or climate change, cannot be fully explained by these social contagion processes alone. In a model proposed by Goldberg & Stein [1], agents observe each other’s behaviors, but this does not directly lead to simple or complex contagion. Instead, agents only increase their association between the behaviors observed. In this work, we extend the associative diffusion model of [1], designed for the spread of behavior specifically, to study the diffusion and distribution of opinions. Unlike behaviors, people might express negative opinions and these opinions can be negatively associated. In our updated model, we do not find the binary polarized structures observed by [1], but instead, we find various groups of individuals holding similar world views, expressed as the same association structures and matching opinions, see Fig. 1. Investigating the discrepancies between our findings and the model in [1] suggests that their binary polarized structure is largely a result of the assumptions and choices made in the implementation of the Goldberg & Stein model, for which, we argue, we offer a more suitable alternative. By offering a model of opinion dynamics that integrates a dimension of meaning, this research shows that the clustering of opinions may arise due to the emergence of bundled associations, a proxy for meaning or culture, and offers an explanation for the slow but ever-changing composition of opinions in society.

Συνεδρία: 
Authors: 
Anna Keuchenius, Sabin Roman, Chico Camargo, Irene Malvestio, Mikolaj Biesaga, Lea Hildebrandt and Alexandre Bovet
Room: 
1
Date: 
Tuesday, December 8, 2020 - 15:45 to 15:50

Partners

Twitter

Facebook

Contact

For information please contact :
ccs2020conf@gmail.com