Olives @olives@social.cleverlibre.org

@Hikki What sort of anime do you like?

@rose Uh, are you getting my messages or is the federation borked?

@skyflare What sort of anime do you like?

@lispi314 @frankiesaxx The report was shoddy, and I think the guy behind it is troublesome.

First off, he doesn't seem to understand how federation works, or how common inter-instance censorship is. It's hard to say how representative any particular statistic is of the fediverse.

He ignored that the worst instance, at the time, had a reputation of not being well moderated (at an earlier point, I figured they just made a few mistakes in their moderation), and tried to paint other instances in that *country* with a broad brush. Btw, the owner of the new company who own the instance is an American living there.

He also seemed very confused about whether it was their content policy, or their level of moderation. His aimless speculation here didn't add anything useful. That said, there are instances with permissive content policies, and who don't seem to end up like that.

There are also other possible explanations, such as language barriers. What particularly struck me is that he didn't seem to have spoken to anyone at all, he just got out a pen, and wrote out lines of speculation.

There were a lot of assertions, speculation (especially, of a racist kind), spooky posts with older timestamps, and "potential hits".

He used incendiary language such as "the most hits ever encountered" (it didn't seem much different from a previous project on Twitter when their scanning tool broke, maybe less so, he otherwise relied a lot on a far more experimental nudity detector).

He didn't look at the images, for legal reasons, although probabilistically speaking, it's very likely it hit something in the database for Microsoft's algorithm. This matters more when you remember that running this scanning tool results in an automatic report being made to the police. Microsoft explicitly markets this fact.

The report seemed curated for an audience and without raw data and statistics.

I don't expect every raw post (minus links / images), but it would be nice to know statistical trends, especially as there have been large influxes of users into the fediverse lately (which coincide with what were very recent anecdotal reports relating to one particular instance, there was also an anecdotal report of a certain large site tightening their moderation around that time).

That data was conspicuously absent. We just know some things happened there. Simply commenting on this is difficult.

At a later point, he went on about how an "image might be cropped" so as to be impossible to distinguish as abuse or that someone "might have difficulty telling the age". It was a facially transparent attempt to try to sell Microsoft's scanning solution (and also undermined some of his previous assertions).

He even seemed to encourage greater federation but with that scanning tool. He was suddenly concerned with fragmentation. The solution to everything, it seems, is for Microsoft to inspect your messages.

He also not only conflated art with abuse, he also got very confused by suggesting art is "illegal in Japan", and simultaneously suggested it was not. If you're wondering, it's not.

He chose language which would appear the most incendiary, damned of whether it was even internally consistent.

And perhaps there is some instance out there, if they are indeed supportive of discussions of grooming, that is troublesome and action should be taken. That said, it's not clear it is that, and if it was, his proposals seemed particularly strange for addressing that (and, frankly, shouldn't involve racial rhetoric).

Also, it's not unheard of for an instance (or mainstream social media) to get dodgy spam attached to hashtags, and to deal with that, you well, moderate it.

He tried to shock and awe with numbers which could fit in a headline, but when it came to providing a lot of concrete data, he came up short.

That is why it was bad.

Perhaps, moderation on some instances could have been improved, and there may have been problems. Still, this report was abjectly awful.

https://en.wikipedia.org/wiki/Publish_or_perish