Just a reminder. A research project involving Cornell and Stanford looked at ways to automate the classification of particular categories of poster: Antisocial Behavior in Online Discussion Communities
cs.stanford.edu/people/jure/pubs/trolls-icwsm15.pdf
There's a section that Germaine Bunbury might find relevant at present, it's almost as if there can be 'sleepers' who post for a while, matching the tone of other posts and then try and move that tone to a more hostile and aggressive one. One that might almost make it seem like there are "extremists on both sides". Or they adopt the faux-naif, just wondering how to deal with this, golly it's such a conundrum approach:
Antisocial behavior, which includes trolling, flaming, and griefing, has been widely discussed in past literature. For instance, a troll has been defined as a person that engages in “negatively marked online behav- ior” (Hardaker 2010), or a user who initially pretends to be a legitimate participant but later attempts to disrupt the community (Donath 1999).
Empirically, we find that many of these banned users exhibit such behavior. Apart from insults and profan- ity, these include repeated attempts to bait users (“Ouch, ya got me. What’s Google?”), provoke arguments (“Liberalism truly breeds violence...this is evidence of that FACT”), or de- rail discussions (“All I want to know is...was there a broom involved in any shape or form?”).
...
How do (community disruptors) generate activity around themselves? Do [community disruptors] purposefully try to create discussions, or opportunis- tically respond to an on-going discussion? Here, we find that their behavior differs depending on the community (Figure 2a). On Breitbart and IGN, trolls are more likely to reply to others’ posts, but on CNN, they are more likely to start new discussions (t>9.1, p0.25). Still, across all communities, (community disruptors) appear to be effective at luring other users into potentially fruitless discussions...