Fig. 3: Participation of users is not dependent on toxicity.
From: Persistent interaction patterns across social media platforms and over time
a, Examples of a typical trend in averaged user participation (top) and toxicity (bottom) versus the normalized position of comment intervals in the threads (Twitter news dataset). b, Box plot distributions of toxicity (n = 25, minimum = −0.012, maximum = 0.015, lower whisker = −0.012, quartile 1 (Q1) = − 0.004, Q2 = 0.002, Q3 = 0.008, upper whisker = 0.015) and participation (n = 25, minimum = −0.198, maximum = −0.022, lower whisker = −0.198, Q1 = − 0.109, Q2 = − 0.071, Q3 = − 0.049, upper whisker = −0.022) trend slopes for all datasets, as resulting from linear regression. c, An example of user participation in toxic and non-toxic thread sets (Twitter news dataset). d, Pearson’s correlation coefficients between user participation and toxicity trends for each dataset. e, Pearson’s correlation coefficients between user participation in toxic and non-toxic threads for each dataset.
