We developed the first automatic method to measure governance outcomes, by looking at how community members talk about their moderators.🗣️We collect all discussion for the top 10k subreddits over 5+ years💾 which we make public!

Super surprised and honored to receive the single Best Paper award at #ICWSM this year (out of 138 papers) for my work with @lleibm.bsky.social, @axz.bsky.social, @timalthoff.bsky.social on Reddit Rules!

bsky.app/profile/gale...

Amongst other things, we find that rules phrased prescriptively (“do this”) are received more positively than those phrased restrictively (“don’t do that”), and that the effect of adding new rules is initially positive, but wears off after ~ 6 months.

We associate these rules with key community outcomes: how do community members talk about their moderators? We classify comments discussing moderators according to their sentiment, and we make these datasets public.

We classify rules from 5.2k subreddits, capturing more than 2/3s of all activity on reddit, according to their tone (does the rule say what or what not do do), target (does the rule address content, formatting, or users) and topic (twelve specific topics).