Table 1.
26 types of pro-mask hashtags ranked in token frequency.
Table 2.
Nine types of anti-mask hashtags ranked in token frequency.
Table 3.
Functional subcategories of pro-mask hashtags ranked by relative frequency.
Table 4.
Functional subcategories of anti-mask hashtags ranked by relative frequency.
Fig 1.
Trends of pro-mask hashtag uses on Twitter.
This figure plots the 7-day moving averages of the daily rates of nine most popular pro-mask hashtags from early March to early August 2020. Along the timeline are shown news headlines in four categories: mask-related guidelines made by public health authority CDC, administrative actions compliant with the CDC guidelines, administrative violation of the CDC guidelines, and mixed messages from White House on masks.
Fig 2.
Trends of anti-mask hashtag uses on Twitter.
This figure shows the 7-day moving averages of the daily rates of all anti-mask hashtags from early March to early August 2020. As Fig 1, along the timeline are shown media coverages on CDC recommendations and administrative actions related to masks.
Fig 3.
Trends of top 3 anti- and pro-mask hashtags on Twitter.
This figure shows the 7-day moving averages of the three most popular anti-mask hashtags (#NoMask(s), #Mask(s)Off, #MasksDontWork) and pro-mask hashtags #WearAMask, #WearADamnMask, #MaskUp) in the period of interest in a stacked area chart.
Fig 4.
Exponential trends of mask-related hashtags from March 1 to August 1 2020.
Fits (a-d) are best-fit exponentials with growth rate λ over a given interval; for (a) λ = 0.122/day from March 2 to March 26 with R2 = 0.993, for (b) λ = 0.025/day from April 8 to June 1 with R2 = 0.998, for (c) λ = 0.124/day from March 2 to March 22 with R2 = 0.988, and for (d) λ = 0.049/day from March 23 to June 1 with R2 = 0.998.
Fig 5.
(A) Scatter plot of each user’s replies to anti- and pro-mask tweets, with each red circle representing an anti-mask user and each teal circle representing a pro-mask user. The horizontal and vertical axes indicate the number of a user’s replies to pro-mask tweets and to anti-mask tweets, respectively. (B) Distributions of replies to pro-mask tweets where the 0-reply teal bar is minimal in contrast to the 0-reply red bar representing over 20% of anti-mask users. (C) Distributions of replies to anti-mask tweets where the 0-reply teal bar representing 99% of pro-mask users sticks out on the horizontal axis, towering over the 0-reply red bar.
Fig 6.
The distribution of the response bias B (Eq 4) for both pro-mask and anti-mask users.
The vast majority of users fall near the extremes of possible values of the response bias (i.e. within 0.2 of B = +1 or B = −1), with the percent of users in a given group indicated for each extreme bin.
Fig 7.
Correlation matrix of the time series of pro-mask hashtags, anti-mask hashtags, and daily confirmed COVID-19 cases.
On the diagonal are shown the density plots, which visualize the distributions of the three variables over continuous intervals, with hashtag/case counts on the horizontal axis and number of days (normalized) using kernel probability estimation on the vertical axis.
Fig 8.
(A) Mask-related hashtag counts and daily confirmed COVID-19 case counts in the U.S. from early March to the end of July 2020. The figure plots the trends of mask-related hashtags on the left y-axis and the trajectory of daily confirmed COVID-19 cases on the right y-axis using seven-day moving averages for the two data series. (B) Three temporal trends from early March to the end of July 2020: (1) monthly statewide mask mandates, (2) monthly Google Trends searches for “face mask,” and (3) monthly high-profile mask-related news headlines. The volume of each temporal trend is rescaled from 0 to 1 for the ease of comparison.