The Impact of Data Presentation Practices on Perceptions of Political Polarisation

Abstract

By multiple metrics, the United States is a highly polarised sociopolitical environment. However, some researchers – while acknowledging the presence of some degree of polarisation – have deemed it fair to question whether citizens are as divided as they believe themselves to be. “False polarisation” is a phenomenon wherein true levels of polarisation are consistently overestimated by political partisans, leading to pervasive interparty misperceptions and arguably precipitating a self-fulfilling prophecy which ultimately accelerates actual polarisation. While false polarisation is purported to be driven by well-documented socio-cognitive propensities such as oversimplification and categorisation, the present research questions whether it may also be a result of the manner in which we choose to present data. Chapter 1 tested the impact of data visualisation format on participants political perceptions and cross-party behaviour. A series of four inter-related studies yielded a number of key findings, including 1) type of visualisation plays a highly-significant role in perceptions of polarisation, 2) more “balanced” visualisation formats (e.g., those which depict both group differences and similarities or those which do not visually accentuate group differences) promote more accurate perceptions of intergroup agreement, and 3) providing participants with certain types of “sub-optimal” data visualisations may be superior to providing them with no visualisation at all. Chapter 2 examines the results from a large-scale replication of our work in Chapter 1, which broadened both the scale and scope (i.e., two separate geopolitical contexts) of our original set of four studies. Chapter 3 investigated the longitudinal impact of repeated exposure to different formats of data summarisation on partisan political perceptions over time. Although results broadly failed to identify significant differences between summarisation techniques, a pattern materialised which seemed to indicate that any summarisation technique promoted more positive intergroup perceptions than a no data baseline. Chapter 4 sought to understand whether different visualisation choices and summarisation techniques might impact the way in which an audience A) engages with content or B) perceives sources that present content in a particular manner. Results suggested that while different data presentation choices seem to have little impact on engagement with content, they exert significant influence on metrics related to source perception (e.g., level of trust in source, beliefs about intent, etc.), especially when the relative advantages and disadvantages of presentation methods were made explicit. Finally, Chapter 5 tested whether it was possible to “inoculate” individuals against perceiving exaggerated levels of polarisation when presented with interparty data depicted via truncated bar charts. Results indicated that a short inoculation intervention made individuals significantly more capable of correctly differentiating between misleading and non-misleading visualisations and ultimately significantly less likely to succumb to truncation-induced polarisation effects.