Validating daily social media macroscopes of emotions

Abstract

To study emotions at the macroscopic level, affective scientists have made extensive use of sentiment analysis on social media text.However, this approach can suffer from a series of methodological issues with respect to sampling biases and measurement error. To date, it has not been validated if social media sentiment can measure the

Exposures and Behavioral Responses to Wildfire Smoke

“…Regarding our sentiment analysis, roughly a quarter of American adults are estimated to use Twitter, with limited differences in use by race (slightly higher use by non-White populations) but higher use among younger Americans and higher-income Americans (71). Nevertheless, temporal variation in Twitter-derived sentiment measures appears to validate well against measures from independent survey data (25) and to correlate well with survey-derived measures of behavior (72). Finally, previous work on climate stress demonstrates that Twitter-derived sentiment measures responds similarly to such stress as compared with responses measured in representative vital statistics or survey data (73,74).…”

“…Following this earlier work, we analyze ∼1.7 billion georeferenced Twitter updates ("tweets") posted since 2016 using natural language processing algorithms that extract information on the sentiment revealed in each tweet (28) (Methods). This approach has been validated at population scale against self-reported measures of emotional state (25), and complements earlier work that used Twitter to directly measure wildfire activity (29) and infer smoke concentrations (30).…”

“…Such preferences underlie standard theoretical models of choice behavior but are hard to observe directly and at scale. Past work has shown that social media behavior can be a sensitive and accurate tool for understanding individuals' preferences and sentiments toward what is happening around them (25), including a changing environment (26,27). Following this earlier work, we analyze ∼1.7 billion georeferenced Twitter updates ("tweets") posted since 2016 using natural language processing algorithms that extract information on the sentiment revealed in each tweet (28) (Methods).…”