Power Law in Popular Media (2018)

Related: Zipf's Law (https://en.wikipedia.org/wiki/Zipf%27s_law), Sturgeon's Law (https://en.wikipedia.org/wiki/Sturgeon%27s_law), and endless "most of ${X} on ${SOCIAL_MEDIA_PLATFORM} has no / few (views|prevalence|interaction).

An interesting corollary is that this makes moderation / management somewhat more tractable if one focuses not on all content but that which is gaining some prevalence. I've suggested multiple times, though am aware of no actual moderation practices, in which all content above a specified prevalence threshold receives some human moderation. This can be used both reactively (when an item / post reaches the threshold it receives moderation) or prescriptively (when a post reaches the threshold it is held for moderation and won't be recirculated / promoted until this occurs). Given the inverse-log relationship, it's reasonably straightforward even on enourmous platforms to moderate most items being presented to members with a relatively small staff.

That formula could of course be modified with automated risk assessment looking at content characteristics, e.g., has media (NSFW/NSFL/CSAM risk), keyword, Bayesian, and these days LLM/AI models, photo content analysis, distribution patterns, behavioural patterns, etc.), to focus attention on items more likely to be problematic at lower prevalence thresholds, e.g., when they've only been shared/viewed/presented a smaller number of times.