3.3 – Trajectories of comment thread conversations Observations In analyzing sections of comment threads where two or more people engaged in a back and forth, we did not find a single example exhibiting a positive resolution. Such interactions ended when one poster stopped posting; we did not see any interactions where one poster acknowledged that their opinions had changed. Instead, the interaction often escalated. Similar to Observation 1, there also appeared to be a pattern in which counter speech comments under positive anchor posts were more likely to receive positive support from other commenters even when there was a hate-speech attack on them. Whereas counter-speech under posts with overwhelming negative comments was more often directly and personally targeted by more hate-speech comments. This sometimes turned into a form of bullying, more so if the speaker was a woman. When positive comments were attacked by hate-speech under positive anchor posts, users sometimes made reference to positivity either by referring to the anchor’s content or a plea to look with positive perspective10. In a few cases, counter speech under hate-speech posts also emphasized positive aspects of life by saying things such as ‘isn’t people loving each other better than hating each other?’. This happened when the comment had something to refer to in the original post; we noted it under posts about interfaith relationships and same-sex marriage. Finally, it was common to see comment threads that escalated when the counter-speaker argued with the original hate-speech speaker, and vice versa, and where the counter-speech often included abusive language. Stand-alone counter speech also sometimes contained strong language and there was no pattern as to when a speaker decided to speak in strong language, abusive language, or neutral language. Interpretations Trajectories of comment thread conversations that lack positive resolution suggest at least two possible interpretations. One possibility is that no-one is positively affected by counter-speech in comment threads; the counter-speaker may just get provoked in the process to use hate speech and people’s views can become entrenched. The second possibility is that the original hatespeech poster, the one posting the counter-speech, or other observers could be affected by the counter-speech. This effect could result from the individual comment thread interaction, or it could be the aggregate result of multiple such interactions; it could happen immediately or later on, but the effect is not acknowledged via a post in the comment thread. Thus while we did not find clear evidence for the positive effects of comment thread interactions in those same threads, the absence of evidence is not evidence of absence. It may be that positive 10 Annex 7 11