Influence is often studied by examining explicit structural links in observed interaction networks, such as paper citations or bill co-sponsorship. However, for many domains, such declared links do not exist, are unreliable, or fail to reflect behaviors of interest. In these situations, observed temporal dynamics can instead be used as a proxy by which latent structural information can be inferred. In this talk, I will present the Bayesian Echo Chamber, a new Bayesian generative model for social interaction data. By modeling the evolution of people's language usage over time, using a discrete analog of the multivariate Hawkes process, this model is capable of discovering latent influence relationships. Unlike previous work on inferring influence, which has primarily focused on simple temporal dynamics evidenced via turn-taking behavior, the Bayesian Echo Chamber captures more nuanced influence relationships, evidenced via linguistic accommodation patterns in interaction content. I will present results validating the model's ability to discover known influence patterns using transcripts of arguments heard by the US Supreme Court and the movie "12 Angry Men." Finally, I will showcase the Bayesian Echo Chamber's capabilities by presenting latent influence relationships inferred from Federal Open Market Committee meeting transcripts, thereby demonstrating state-of-the-art performance at uncovering social dynamics in group discussions.
(Joint work with F. Guo, C. Blundell, and K. Heller.)
Hanna Wallach is a researcher at Microsoft Research New York City and an assistant professor in the School of Computer Science at the University of Massachusetts Amherst. She is one of five core faculty members involved in UMass's recently formed Computational Social Science Institute. Hanna develops machine learning methods for uncovering new insights about the structure, content, and dynamics of social processes. In collaboration with political scientists, sociologists, and journalists, she analyzes publicly available interaction data, such as public record email networks, declassified document dumps, press releases, meeting transcripts, and news articles. Hanna's research contributes to machine learning, Bayesian statistics, and the nascent field of computational social science. Her work on infinite belief networks won the best paper award at AISTATS 2010. Hanna holds a B.A. in Computer Science from the University of Cambridge, an M.S. in Cognitive Science and Machine Learning from the University of Edinburgh, and a Ph.D. in Physics from the University of Cambridge. Most importantly, however, she is (to her knowledge) the only person to have appeared in both Glamour magazine ("35 Women Under 35 Who Are Changing the Tech Industry") and Linux Format.