Seminar: Using computational methods to study non-verbal political communication - Constantine Boussalis (TCD)
Venue: G316, Newman Building, UCD
Date: 26 February 2020
14:00-15:00 - Connected_Politics Workshop: Images-as-data
Abstract: Recent advances in image processing and machine learning methods have opened a new era of “images-as-data” research in the social sciences. One important application of these methods is the study of how political actors communicate non-verbally and the extent to which multimodal emotive displays of politicians (verbal, vocal and non-verbal) influence voter opinion. This talk will review state-of-the-art computational approaches to the study of images—namely, convolutional neural networks—and discuss findings of a study, which combines image and viewer continuous response measurement data from the 4th Republican Party presidential debate held on 10 November 2016. Further, best practices in model validation will be discussed critically.
Speaker: Dr. Constantine Boussalis is an Assistant Professor of Political Science at Trinity College Dublin. His research interests include environmental politics, political communication, political behavior, and computational social science. Before joining the Department of Political Science at Trinity College Dublin in 2014, Constantine served as Empirical Research Fellow at Harvard Law School from 2011.