The problem is that somehow you wind up in long heated arguments with “centrists” which wander away from the topic and get personal

I’m not surprised I was identified by the bot, but it’s worth pointing out that ending up in heated arguments happens because people disagree. Those things are related. If someone is getting into lots of lengthy disagreements that are largely positive but devolve into the unwanted behavior, doesn’t that at least give legitimacy to the concern that dissenting opinions are being penalized simply because they attract a lot of impassioned disagreement? Even if both participants in that disagreement are penalized, that just means any disagreement that may already be present isn’t given opportunity to play out. Your community would just be lots of people politely agreeing not to disagree.

I have no problem with wanting to build a community around a particular set of acceptable behaviors -I don’t even take issue with trying to quantify that behavior and automating it. But we shouldn’t pretend as if doing so doesn’t have unintended polarizing consequences.

A community that allows for disagreement but limits argumentation isn’t neutral - it gives preferences to status-quo and consensus positions by limiting the types of dissent allowed. If users aren’t able to resolve conflicting perspectives through argumentation, then the consensus view ends up being left uncontested (at least not meaningfully). That isn’t a problem if the intent of the community is to enforce decorum so that contentious argumentation happens elsewhere, but if a majority of communities utilizes a similar moderation policy then of course it is going to result in siloing.

I might also point out that an argument that is drawn out over dozens of comments and ends in that ‘unwanted’ behavior you’re looking for isn’t all that visible to most users; if you’re someone who is trying to avoid ‘jerks’ then I would think the relative nested position/visibility of that activity should be important. I’m not sure how your bot weighs activity against that visibility, but I think even that doubt that brings into question the effectiveness of this as a strategy.

Again, not challenging the specific moderation choices the bot has made, just pointing out the problem of employing this type of moderation on a large scale. As it has been employed in this particular community is interesting.