Polarization manifests in many forms, and has multiple different causes. One of the challenges in polarization research is the lack of precise formal models. A project team led by Lee de-Wit at Cambridge, and co-directed by Jens Madsen at London School of Economics plans to develop a computational model of a particular form of polarization known as opinion polarization, namely the extent to which people are one-sided in the opinions they hold.

The project team will articulate a model of polarization that makes specific predictions about the interaction between individual cognitive and contextual factors. This  will not only provide deep theoretical insight into the drivers of polarization, but also provide a route forward in terms of policy recommendations about the kinds of information environments that could reduce polarization.

To do this, the team plans to test a Bayesian model of opinion polarization. This model has the theoretical potential to bridge individual and societal levels of analysis by using a computational model to generate precise numerical predictions about the extent of opinion-based polarization across societies, over time and in controlled experiments.

The team predicts that in societies where media sources are perceived to be more independent there will be less opinion-based polarization. To gauge how such a process may explain variability, they propose to test this theory in three different ways: cross-sectionally in 12 countries, longitudinally in 6 countries, and experimentally in 1 country.

The model and collected data will add critical insights into the cognitive mechanisms of polarization and explain some of the variability in polarization that we observe across countries and will give a roadmap to better understand how to avoid deepening polarization in the future.