Deriving an Opinion Dynamics Model from Experimental Data

Opinion dynamics models have huge potential for understanding and addressing social problems where solutions require the coordination of opinions, like anthropogenic climate change. Unfortunately, to date, most of such models have little or no empirical validation. In the present work we develop an opinion dynamics model derived from a real life experiment. In our experimental study, participants reported their opinions before and after social interaction using response options “agree” or “disagree,” and opinion strength 1 to 10. The social interaction entailed showing the participant their interaction partner’s agreement value on the same topic, but not their certainty. From the analysis of the data, we observed a very weak, but statistically significant influence between participants. We also noticed three important effects. (1) Asking people their opinion is sufficient to produce opinion shift and thus influence opinion dynamics, at least on novel topics. (2) About 4% of the time people flipped their opinion, while preserving their certainty level. (3) People with extreme opinions exhibited much less change than people having neutral opinions. We also built an opinion dynamics model based on the three mentioned phenomena. This model was able to produce realistic results (i.e. similar to real-world data) such as polarization from unpolarized states and strong diversity.