How Distrust Shaped COVID-19 Skepticism

Predictive modeling and survey data

Priniski, J. H., & Holyoak, K. J. (2022). A darkening spring: How preexisting distrust shaped COVID-19 skepticism. PLOS ONE, 17(1), e0263191.

GitHub — Data & Analysis

Tracing Skepticism Before and During the First Wave

People rarely hold beliefs in isolation; instead, they hold systems of beliefs that mutually cohere. During the COVID-19 pandemic, many Americans underestimated the virus’s risks despite widespread communication from health experts. Through preregistered surveys and Bayesian modeling, we examined how preexisting distrust in Democratic politicians and medical scientists shaped this skepticism. As the first wave unfolded, skeptical attitudes became more deeply embedded within a complex network of political, economic, and medical beliefs.

Surveying Beliefs as the Pandemic Unfolded

We surveyed Americans on March 13, April 17, and May 19, 2020. Analyses included 628 participants in March and 1,610 across April and May.

Survey dates on March 13, April 17, and May 19, 2020, shown against the seven-day rolling average of reported U.S. COVID-19 cases. — **Figure 1. Survey timing.** The March survey preceded the first U.S. wave; the April and May surveys captured attitudes during it. Cases are shown as a seven-day rolling average, visualized with Our World in Data using Johns Hopkins University data.

We developed nine scales, each composed of three to five items, to measure perceived COVID-19 severity, scientific and medical distrust, and political considerations surrounding the pandemic response.

Published survey table listing nine belief scales, example survey items, and whether each scale was administered before or during the first wave. — **Table 1. Survey scales and example items.** Seven scales were included before the first wave and eight during it. Media distrust was measured only in March; vaccine effectiveness and reopening the economy were added in April–May. The statements shown are items participants rated on a seven-point agreement scale.

Early Skepticism Centered on Political Distrust and Vaccine Fears

In March, distrust of Democratic politicians and fears of a future COVID-19 vaccine predicted taking the virus less seriously. Political ideology added no reliable predictive effect after accounting for the other beliefs.

Leave-one-out cross-validation favored a compact model containing political distrust and vaccine fears: its predictive accuracy was within one standard error of the full model. The model using political ideology alone performed worst (Figure 5).

ASurvey Data

March survey responses on perceived COVID-19 severity plotted against political distrust, vaccine fears, and foreign-threat concerns, with fitted conditional effects. — Points show participants’ average scale responses; blue and red identify socially liberal and conservative respondents. Density curves summarize response distributions. Black lines and shaded bands show conditional model estimates and 95% credible intervals.

BBayesian Predictors

March Bayesian model estimates for all belief predictors and political ideology. Political distrust and vaccine fears predict lower perceived severity. — Posterior estimates compare predictors while accounting for the other beliefs and political ideology. Dots mark posterior means; thick and thin bars show 95% and 99% credible intervals. Negative values predict lower perceived severity, or greater skepticism.

Skepticism Became Linked to a Broader Web of Beliefs

During April and May, support for reopening the economy and distrust of large medical organizations joined political distrust and vaccine fears as strong predictors of skepticism. Uncertainty about the virus’s origins predicted taking it more seriously.

A five-predictor model without political ideology was the simplest model within one standard error of the best predictive accuracy (Figure 9). Comparing the surveys suggests that skepticism became embedded in a wider system of political, economic, and medical beliefs.

ASurvey Data

April–May survey responses on perceived COVID-19 severity plotted against five belief scales, with fitted conditional effects. — Participants’ average scale responses show how skepticism relates to political distrust, reopening the economy, vaccine fears, medical distrust, and beliefs about the virus’s origins. Colors identify social ideology; lines and bands show conditional model estimates and 95% credible intervals. Low, Unsure, and High summarize the seven-point scales.

BBayesian Predictors

April–May Bayesian model estimates showing political distrust, reopening the economy, vaccine fears, and medical distrust predicting lower perceived severity, and uncertainty about the virus’s origins predicting higher severity. — Posterior estimates compare predictors while accounting for the other beliefs and political ideology. Dots and bars summarize the estimates and their uncertainty. Negative values predict lower perceived severity. A broader set of beliefs now predicts skepticism.

Understanding Health Attitudes as Systems of Beliefs

The findings connect health judgments to the broader beliefs people use to interpret new information. They inform my research on narrative agency and educational interventions: understanding resistance to a message requires understanding how it fits into a person’s existing view of the world.

The surveys show how vaccine attitudes fit within broader systems of beliefs. To examine the causal relationships people express in their own words, I tested my AI pipeline for modeling human beliefs on tweets about COVID-19 vaccines. The pipeline maps claims linking vaccination to anticipated benefits and harms, extending this work from survey responses to beliefs expressed in online discussion.