An Empirical and Simulation Investigation of Bounded Confidence and Negative Influence in Opinion Dynamics Using Stochastic Actor-Oriented Model

Bounded confidence and negative influence are two of the most important micro-level mechanisms employed in computational models of opinion dynamics to explain polarization of opinions. However, empirical evidence of both mechanisms is debatable. Two common limitations in existing empirical studies are (1) the limited external validity of laboratory experiments, and (2) the inability in study designs to disentangle negative influence from bounded confidence, as well as from other social influence mechanisms like assimilation. We address both limitations, using the Stochastic Actor-Oriented Model (SAOM) with a longitudinal field data set that tracks adolescents’ social network relations and opinions on a set of issues. Two new SAOM effects are introduced to represent bounded confidence and negative influence, respectively. Results show that for adolescents’ preferences on rap/ hip hop clothing style, the model containing both effects provides a good fit to the data, but only the effect representing negative influence is statistically significant. The results support that negative influence contributes to explaining observed opinion changes, but lend little weight to bounded confidence. Further simulation studies based on the empirically estimated model show that our model implies only low levels of opinion polarization at the macro-level despite negative influence at the micro-level. We conclude that our approach not only overcomes common limitations of earlier empirical work, but also bridges the SAOM and agent-based modeling by offering empirically validated insights into opinion dynamics.