Three Insult Rule

Content Moderation as typically conceived of is circumscribed by the "four corners of the content" - it tries to apply a set of rules to a piece of content, and test whether or not (or to what degree) the content violates those rules.

This approach is limited in its capacity to capture patterns of behavior that constitute abrogation of community standards, but fail to do so in the context of a single piece of reportable content. Particularly in the context of harassment + cyber-bullying, this poses a headache for platforms trying to formalize and standardize their treatment of content, while respecting and serving user expectations. 

One approach to solving this problem is the so-called "Three Insult Rule" that Facebook has used. From that Lawfare Article:

"The rule required the presence of multiple mild insults in a post (or across several posts) before Twitter would step in. The rationale behind the rule was to avoid arbitrating schoolyard disputes between users; the company’s top-down content moderation procedures would only kick in past a certain point of hostility. The three-insult rule was one of hundreds of granular guidelines arbitrating the finer points of online speech—a set of detailed guidelines that exist at nearly every tech company but only rarely see the light of day."

While it is certainly not worth going into the myriad of small rules like this one that platforms can implement, it's an important category of intervention to call out one facet of this kind of pattern. Signals along one dimension (content) can be aggregated into signals along other dimensions (users) that help achieve better outcomes. This isn't unidirectional, and it isn't limited to these two classes of signals.