Joint study of social media and affective polarization in Kenya

This is an observational study, in which we are aiming to demonstrate a reproducible method for quantifying the prevalence of polarizing content (defined later in this preregistration) across five social media platforms (X, Instagram, Facebook, YouTube, and TikTok) in Kenya. In short, we are recruiting a representative sample of participants (with representation specified by demographic quotas) primarily via social media ads. The participants are then asked to complete a survey (the "Survey") about their experiences on social media via Qualtrics, and then to download and install a custom Android app (for a "Field Study"), via which they log in to a social media platform which allows us to scrape the top posts that appear in their feed, the top comments on those posts, and the accounts they follow. We will then subsequently determine the top most followed public accounts on each platform, and likewise scrape the most recent posts from those accounts, and the comments on those posts. This data will allow us to estimate the prevalence of polarizing content both (a) among recommended content, (b) among the largest public accounts on each platform. We use the social graph data to additionally estimate the level of social sorting in the follow graph on each platform. This pre-registration focuses primarily on the Field Study; Analysis of the Survey will be the responsibility of our collaborators at the Neely Center at the University of Southern California. We expect this project will be the first study of cross-platform measures of polarization, allowing major platforms to be ranked according to the degree of polarization and polarizing behaviors present on the platform. Showing platform users the numeric reflection and representation of polarizing behaviour may also spur reflection and potentially behavioural modifications. As such, it will: (1) provide a methodological template for how on-platform polarization could be measured more cheaply in other countries using tested metrics and legal mechanisms for accessing platform data (e.g., Articles 37 and 40 of the DSA); (2) explore how league tables (comparing platforms) can incentivize platform reform and inform user choice; and (3) spur discussion of platform responsibilities with respect to conflict and polarization that informs policy about how the interaction of online platforms with societal conflict should be regulated, including by considering taxation on the polarization footprint as explored in this paper. Given the above goals and the fact that this is an observational study, we are filing this pre-registration primarily to prevent selective reporting, post hoc framing, or motivated methodological choices. For example, to prove that we didn't change what platforms to include, or the metrics used for ranking platforms, to suit a particular narrative.