Bots are extremely good at mimic human interactions but you can still spot them especially if you have a bit of experience managing communities. SN can try to automate it but it must be something that you can train in order to polish the detection and decrease the amount of false positives to zero, or close to it. Also a neat feature could be some sort of "flood alert": When SN is under a spam attack it could automatically raise all costs to post/comment to an absurd amount and reward people that "fight" during the attack (with reports and such)