Understanding the Dynamics of Contemporary Political Polarization Through Agent-Based Models

According to recent popular narrative, Americans and the politicians they support are developing increasingly extreme opinions across a wide range of issues, and, as a result, citizens and their representatives are much less likely to interact and collaborate with those with whom they disagree. Many have gone so far as to vilify others who identify with the opposite party, further hampering social cohesion and meaningful policy progress. In the aftermath of the 2016 election, Professor Matthew Gentzkow wrote a summary of much of the contemporary literature on the nature of political polarization, describing several key phenomena that characterize the current state of polarization in the US. First, the distribution of individuals’ self-identified positions along the liberal-conservative spectrum has not actually changed in the last 40 years. In fact, the overall distributions of beliefs about specific issues have remained singly-peaked rather than becoming bimodal as a result of people developing more extreme opinions. However, the median Democrat is much more liberal and the median Republican is much more conservative than they both were 20 years ago. How is this observation possible if the distributions of beliefs have not changed? Gentzkow points to the increase in the correlation between beliefs and the correlation between any one belief and identifying with a certain political party as the primary causes [1]. Political processes occur primarily in a social context, meaning that social structure, individual behaviors, and the landscape of political beliefs are intimately intertwined in a way that is not very well understood. In this paper, we study the bidirectional relationship between network structure and agent beliefs by defining and simulating several theoretical models to answer the following questions: