Frontier Research Institute for Interdisciplinary Sciences
Tohoku University

In political science studies, policy positions and campaign manifestos have been considered as important factors for voters to decide their voting behavior. Yet, recent election results indicate that factors which are not directly relevant to politics, such as the emotions and prejudices of voters as well as the appearance of politicians, can greatly influence the voting behavior. Candidates indeed advertise themselves through campaign posters and SNS such as Twitter and Facebook, and try actively to manipulate their outlooks and impressions so as to improve their impressions among voters.

In this study, we explore the role of such nonverbal cues (including facial expressions) issued from politicians in politics by using research methods and knowledge in different fields other than political science such as cognitive neuroscience, intelligence informatics as well as psychology. Specifically, we employ photographs of actual candidates and manipulate their facial expressions by conducting face morphology analyses and using Generative Adversarial Network (GAN), which is one of deep learning methods. By using experimental methods, we attempt to understand the extent to which nonverbal cues--factors other than policies-- affect the political judgment and voting behavior of the electorate.