Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces an information-theoretic framework to quantify the directed flow of semantic content in communicative exchanges, utilizing large language models to compute measures of semantic transfer entropy (STE) and semantic partial information decomposition (SPID). The authors demonstrate the framework's effectiveness through four experiments, revealing insights into information flow in various contexts, including dialogue rigidity, persuasive discourse, psychotherapy quality, and argumentative essay structure. Key findings indicate that the framework can discern the influence of speakers and the contributions of multiple sources in shaping language, highlighting its potential for broader applications in understanding communication dynamics.
Semantic transfer entropy reveals how persuasive speakers dominate discourse, while cognitive rigidity stifles information flow in dialogue.
Quantifying how meaning propagates through communicative exchanges remains underdeveloped in computational linguistics. Here we introduce an information-theoretic framework that quantifies the directed flow of semantic content between interlocutors and decomposes multi-source contributions into redundant, unique, and synergistic components. Our approach leverages large language models as probabilistic estimators of natural language to compute two measures: semantic transfer entropy (STE), which captures directed predictive influence between speakers, and semantic partial information decomposition (SPID), which resolves how multiple sources jointly shape a target's language. Across four experiments we show that the framework detects reduced information flow in cognitively rigid dialogue, captures the dominant role of persuaders in shaping discourse, distinguishes high- from low-quality psychotherapy by the directionality of therapist-client information exchange, and reveals synergistic premise contributions in argumentative essays. This framework opens new avenues for studying information dynamics in digital discourse, pedagogical interactions, clinical dialogues, and any domain in which the structure of linguistic exchange is of research relevance.