Reproduce “Information Dynamics of Language Communication” 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
Leonardo S. Goodall, Andrea I. Luppi, Pedro A. M. Mediano
Researchers from Oxford, Cambridge, and Imperial College London developed an information-theoretic framework using Large Language Models to quantify semantic information flow and multi-source contributions in language communication. This framework, leveraging Semantic Transfer Entropy (STE) and Semantic Partial Information Decomposition (SPID), was validated across synthetic and real-world dialogues, revealing specific patterns of information exchange in various contexts.
Key insights
- Semantic information flow is sensitive to communicative styles, with flexible cognitive styles leading to higher information transfer.
- Communicative roles exhibit distinct directional asymmetries in semantic influence, such as persuaders dominating semantic flow towards persuadees.
- Information dynamics in therapy indicate that lower therapist-to-client semantic transfer and client-driven flow correlate with higher therapeutic quality.
- Argumentative premises can synergistically combine to provide more predictive information about a claim than their individual contributions, though redundancy increases with more premises.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.