Examining Anxiety Contagion in Computer-Mediated Communication using Machine Learning

Abstract

Based on emotional contagion theory, we longitudinally examine the transfer of state anxiety in computer-mediated communication (CMC) in a sample of 277 leaders and 1,649 followers across 87 companies. To test the proposed relationship, we use a machine learning (ML) approach to detect state and trait anxiety in 5,025,171 tweets of leaders and followers, resulting in a total of 88,946 daily dyadic leader-follower observations. We account for the dynamic changes in state anxiety, and control for leader and follower personality traits, demographics, and organization size. We also investigate the role of leader trait anxiety as a moderator of the relationship between leader- and follower state anxiety. In addition, we compare the relationship between leader and follower anxiety before and after the onset of the COVID-19 pandemic, over a total period of 316 days. We find a positive relationship between leader state- and trait anxiety and follower state anxiety even when considering a multi-day lagged analysis; no significant changes between pre and post-onset of the COVID-19 pandemic were observed. This study contributes to the literature on emotional contagion via computer-mediated communication using machine learning algorithms, allowing us to overcome several limitations outlined in previous research, and providing new insights into the complex relationship between leader and follower anxiety.