Table 3.

Coding categories, category prevalence, and coder agreement per category

Dataset	Coding categories	% of texts coded for (min–max)a	Fleiss’ kappa (min–max)b	PABAK (min–max)b
Feedback data	Reach out to the seller, Share objective facts, Express feelings, Help the seller, Avoid harming the seller, Help other buyers, Reward or punish the seller	3.70–69.80	.25–.53	.42–.90
Election data	Care, Harm, Fairness, Cheating, Loyalty, Betrayal, Authority, Subversion, Purity, Degradation, Non-moral	2.11–56.24	.18–.45	.73–91
Reddit data	Anger, Disgust, Enjoyment, Fear, Sadness, Surprise	1.34–40.08	.27–.53	.54–.95
Hate speech data	Hate speech, Offensive language, Neither	5.77–77.43	.55	.72

^a The % of texts coded for the least and most frequently occurring coding categories, respectively

^b The lowest and highest values, respectively, of Fleiss’ kappa/PABAK across the categories in the coding scheme