Identify the mechanisms driving the creator–reply distortion gap

Determine which mechanisms—ranked-reply selection, audience composition, community norms, and the aggregation of multiple replies—produce the observed excess of rhetorical distortion in X/Twitter replies relative to creator tweets.

Background

The study finds that X/Twitter replies exhibit substantially more measured rhetorical distortion than the creator tweets they answer, even after controlling for text length. The authors propose several possible explanations: platform ranking may preferentially surface emotionally intense replies; the composition of audiences replying to political and media accounts may increase partisan or hostile language; and established comment-section norms may generate incivility independently of ranking. Aggregating up to five replies into a single reply unit also contributes mechanically to the observed gap.

The authors explicitly state that their data cannot yet distinguish among these mechanisms. They indicate that resolving the issue would require collecting replies in chronological order as well as the ranked replies displayed by the platform, thereby separating ranking effects from audience and conversational effects.