Researchers introduce an information-theoretic framework to quantify the directed flow of semantic content between interlocutors and decompose multi-source contributions into redundant, unique, and synergistic components.

  • The approach uses large language models as probabilistic estimators to compute semantic transfer entropy (STE) for predictive influence and semantic partial information decomposition (SPID) for joint source shaping.
  • Experiments detect reduced information flow in cognitively rigid dialogue and capture the dominant role of persuaders in shaping discourse.
  • The framework distinguishes high- from low-quality psychotherapy by the directionality of therapist-client information exchange.
  • It reveals synergistic premise contributions in argumentative essays, opening new avenues for studying information dynamics in digital discourse and clinical dialogues.