Currently, we dichotomize accounts into bots and non-bots based on a (previously-validated) threshold on the bot-score. The bot-score is continuous on [0, 1]. We then report user-level and conversation-level average levels of botiness. We should instead report
the user-level average as the average of the bot-scores for the unique users, and
the conversation-level average as the mean of the bot-scores weighted by the number of tweets by each account
Currently, we dichotomize accounts into bots and non-bots based on a (previously-validated) threshold on the bot-score. The bot-score is continuous on [0, 1]. We then report user-level and conversation-level average levels of botiness. We should instead report