Tweeting under uncertainty: The relationship between uncertain language and negative emotions in the wild.

Abstract

Despite decades of research characterizing the relationship between uncertainty and emotion, little is known about how these constructs interact in the wild. Using naturalistic, large-scale language produced on Twitter, we ask whether increases in environmental uncertainty and associated aversive emotional reactions can be captured by the millions of digital traces of people sharing their thoughts online. Analyzing more than 20 million tweets from more than 7.5 million unique users, we find that uncertainty expressions peak when environmental uncertainty is high. This effect, however, is modulated by the type of trigger that increases uncertainty. Pandemics (COVID-19 in 2020) and national U.S. elections (2021) exhibit an increase in uncertainty language and negative sentiment in the real world, illustrating the well-documented relationship between uncertainty and aversive emotional reactions acting in lockstep. In contrast, when uncertain events involve a moral violation (i.e., the 2021 U.S. Capitol attack), specific negative emotions (i.e., anger, fear, and moral outrage) sharply increase, while uncertainty language abruptly decreases. This reveals that in the real world, uncertainty and emotion have a more complex relationship than originally assumed.