Reproduce “Quantifying Polarization: A Comparative Study of Measures and Methods” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Edoardo Di Martino, Matteo Cinelli, Roy Cerqueti, Walter Quattrociocchi
This study offers a comparative evaluation of five established political polarization measures and introduces an adaptation of Kleinberg's burst detection algorithm for identifying modes in ideological distributions. It demonstrates that existing measures yield varied and sometimes conflicting interpretations on synthetic and real-world YouTube data, while the new algorithmic tool provides a more nuanced understanding of the shape and number of ideological peaks, often revealing unimodal patterns in online discussions.
Key insights
- Existing polarization measures vary significantly in their performance, with some like the Bimodality Coefficient proving unreliable for skewed distributions and others like Van der Eijk's sensitive to mode distance, highlighting the need for careful selection.
- The adapted burst detection algorithm effectively identifies modes, revealing that a significant majority (87%) of online discussions are predominantly unimodal and unpolarized, a finding consistent with DFU and Hartigan's Dip Test.
- Center-leaning video comment sections on YouTube paradoxically exhibited more polarization according to several measures, suggesting these platforms might act as 'meeting grounds' for cross-partisan dialogue rather than strict echo chambers.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.