3 ms·
The notion that views or presentations of material is a useful (and quite arguably the most appropriate) metric for abuse, and most especially contrasted with i
by dredmorbius 6y ago
The notion that views or presentations of material is a useful (and quite arguably the most appropriate) metric for abuse, and most especially contrasted with items posted, first surfaced that I'm aware from Facebook's VP Integrity, Guy Rosen, as "prevalence", in an 11 Dec 2020 tweet: https://nitter.cc/guyro/status/1337493574246535168?s=20 https://nitter.cc/guyro/status/1337493574246535168?s=20 Rosen's tweet disclosed a content-violation prevalence of 0.10--0.11% of views, and falling as of 11 February to 0.07--0.08%, 7-8 views per 10,000 on the platform.
There are several other articles addressing YouTube's new "violative view rate" metric. Variety had the best mix of clear headline and minimal paywall, but other substantive coverage is at:
"Tracking Viral Misinformation" (NYTimes): https://www.nytimes.com/live/2020/2020-election-misinformation-distortions/youtube-discloses-the-percentage-of-views-going-to-videos-that-break-its-rules https://www.nytimes.com/live/2020/2020-election-misinformati...
"YouTube says it’s better at removing videos that violate its rules, but those rules are in flux" (recode/Vox): https://www.vox.com/recode/2021/4/6/22368809/youtube-violative-view-rate-content-moderation-guidelines-spam-hate-speech https://www.vox.com/recode/2021/4/6/22368809/youtube-violati...
"YouTube claims it’s getting better at enforcing its own moderation rules" (The Verge): https://www.theverge.com/2021/4/6/22368505/youtube-violative-view-rate-transparency-stat https://www.theverge.com/2021/4/6/22368505/youtube-violative...
Whether you call it "prevalence", "views", or "violative view rate", the notion that you want to look at some function of items of content times reception of that content seems apt. Severity or risk metrics for both content and audience may also be useful --- this is starting to sound an awful lot like an advertising campaign, but one for which the aim is to mimimise impacts based on the material, exposures, and receiving audience.
Such a measure also, at least partially, side-steps some concerns over free speech, and resembles in part characteristics of other public-carriage / public-service functions, in which general access is permitted, but harmful or violative access remains constrained, regulated, and specifically discouraged. The notion that an entity might publish to a platform, but see limited prevalence (say, exposure to only direct or close followers), that identified clusters which are known to amplify disinformation have their capability to do so curtailed, or the like, address at least one line of current concerns with social media's socially-corrosive disruptive capabilities.
(The fact that one person's terrorist is often another's freedom fighter remains, and is not easily resolved. But at least the overall problem is now reduced in size.)