The geometry of online conversations and the causal antecedents of conflictual discourse
A new study uses AI to analyze 3 key dimensions of discourse, revealing how thread structure shapes toxicity.
Deep Dive
Researchers Carlo Santagiustina and Caterina Cruciani published a paper analyzing the causal structure of online conflict. Using LLM prompting to annotate three dimensions—stance, tone, and emotional framing—they studied threaded climate change discussions. Key findings show longer reply delays correlate with more respectful language, and strong alignment exists with a parent post's stance and tone. The geometry of a conversation branch significantly amplifies or dampens these alignment effects.
Why It Matters
This provides a framework for platforms to design systems that structurally reduce toxicity and improve discourse quality.