The Moral Imperatives of Self-Quarantining

• In Event: COVID Communication and Politics

Abstract

This paper is a frame analysis of the moral dimensions underlying citizens' attitudes toward stay-at-home orders mandated by local officials during the Covid-19 pandemic. The thesis of the study is that social media elicit morally inflected responses that map to partisan affiliation and thereby predict citizen cooperation. Such responses improve upon conventional measures of public opinion for their spontaneity, granularity, and capacity to be tracked in real-time over a long period of time.

The influence of moral judgment is amplified on social media because citizens seek approval ("likes") that often depend on being seen as the right kind of person. Thus, news stories containing a moral frame are more likely to be shared on social media, especially so on emotionally charged issues. For example, Facebook discussions among opponents of vaccination are supported by stories of government and media oppression.

The study is based on a theoretical framework derived from Moral Foundations Theory. We add a liberty/oppression foundation to the standard five foundation pairs that make up the core moral appeals (i.e., care/harm, fairness/injustice, loyalty/betrayal, authority/subversion, purity/degradation). Preliminary analysis of our data shows that both liberty and oppression are significant predictors of citizen resistance (by contrast, care is the best predictor of compliance). In an example of the consequence of such resistance, South Dakota Governor Kristi Noem advised citizens to exercise “personal responsibility,” imposed almost no restrictions, and was indifferent to mask wearing. As a result, South Dakota’s Covid-19 death rate is one of the worst in the nation, but the Governor’s approval ratings remain high.

The data were drawn from approximately 7M tweets collected during the first wave of the pandemic, from March 1 through June 30 2020. Each of these tweets has both keywords related to COVID-19 as well as those related to stay at home (home, lockdown, quarantine, open, inside). Out of these tweets, around 35k tweets were geotagged, i.e., contained geocoordinates, in the United States. Because keywords may not be sufficient to determine whether a tweet is about stay-at-home attitudes and may yield false positives, the data were further refined in two stages. The first stage was a hand-coded random sample drawn from the geotagged tweets which was used to build a machine learning classifier for tweet relevancy. The second stage is composed of the portion of the geotagged tweets which were predicted as relevant by the classifier (7k in total). This dataset was further coded for relevance, compliance, underlying moral frame, and vividness.

The geolocation coordinates permit the mapping of sentiment to the county level and thereby to data on the rate of infections and political preferences (2016 presidential vote). These variables help us test the hypothesis that compliance is correlated with those foundations most highly associated with liberal partisanship. The data also permit us to test the hypothesis that elite rhetoric at the local level leads to greater engagement on moral judgment on the part of citizens. That is, we expect more social media activity in areas where local officials have taken a strong stand on the issue (Clifford et al., 2015).

The dataset includes measures of virality (retweets) that can be mapped to the moral frames that underlie the text of the messages, as well as their vividness. The latter refers to personal heartfelt examples used by citizens to illustrate the urgency of their messages. A range of research in perceptions of risk shows that such appeals are much more persuasive than generic appeals or those illustrated with arid statistics (Fetherstonhaugh et al., 1997; Napier & Luguri, 2013). We test the hypothesis that such tweets are more likely to be spread. Preliminary analysis shows that such vivid appeals are most likely to accompany the vice component of a foundation pair, an illustration of the “bad is stronger than good” finding in psychological research (Baumeister et al., 2001). We also test the hypothesis that anger is most likely to accompany such messages. The emotion is associated with risk-seeking behavior and political activism (Kosmidis & Theocharis, 2020).

The paper adds to our understanding of the influence of social media on governance and citizen responsiveness to elite messages on issues of the highest importance for the literal and figurative health of the nation.

Authors

• Andrew Rojecki, University of Illinois, Chicago

• Elena Zheleva, UIC

• Lauren Levine, UIC