The Bayesian Echo Chamber: Modeling Influence in Conversations

Motivation: mapping influence from linguistic accommodation Question: Who influences whom in a conversation? • The influence relations are usually implicit in social interactions. Can we infer them from “who said what at what time”? (availability of many transcript data) • Existing approach — modeling the dynamics of turn-taking behavior, i.e. “who speaks next?” (e.g. you are influencing me if I respond quickly to what you said) • Yet, content information should tell even more. Can we infer influence from content + dynamics? • Linguistic accommodation phenomenon in sociolinguistics