Olives @olives@qoto.org

Thoughts on COSL. The website (they point out it's a WIP) talks of projects being independent, but then, there's language implying they're not so independent. One's a project which probably should be independent, even if they offer consulting services (as they have suggested they do) to them.

Looking at COSL's annual report. There seems to be a bunch of therapists (and a lawyer). I would have thought there'd be more tech expertise (not replacing an existing person but in addition).

https://reason.com/2024/06/21/the-supreme-court-again-strengthens-the-right-to-a-jury-trial-in-criminal-sentencing/ An old one but steep sentencing enhancements now require a trial by jury. A boon for due process.

https://reason.com/2024/04/18/u-s-sentencing-commission-restricts-federal-judges-ability-to-use-acquitted-conduct-at-sentencing/ An old one but judges can no longer boost someone's sentence on conduct a jury acquitted them of. A boon for due process.

I see someone asking why a European instance prohibits communism. It might have something to do with countries like Germany having banned the Communist Party (this was many decades ago).

Should I try out Bluesky?

Does anyone here use Threads? How is it? Is it active? Not talking about opinions from pundits and such.

I'm also not a fan of the term Trust & Safety. I prefer the terms content moderation and content policy.

I've historically not been a fan of Bluesky's moderation practices myself (I've heard that they are puritanical in some ways). However, we should be wary of making these kinds of insinuations.

This should go without saying but suggesting that Bluesky's Trust & Safety Head is somehow a child predator because he won't moderate a specific user who apparently holds hateful views is not okay.

https://en.wikipedia.org/wiki/Scunthorpe_problem
https://tvtropes.org/pmwiki/pmwiki.php/Main/ScunthorpeProblem
Remember the Scunthorpe Problem (algorithmic moderation). #FreeSpeech

However, if someone takes issue with cases like that, it's strange to call for less review.

Without someone manually reviewing it, it also means automatically accusing someone of being a criminal to the police.

A while ago, there was a bit of a scandal over someone's Google account being terminated over photos taken for medical purposes. They could never get it back. It's not clear it was an algorithm, a human moderator is unlikely to consider that either.

And think about how many services one company might operate, someone might lose access to a lot of things based on an algorithm accusing them of something.

"Someone shouldn't review hits for CSAM, they should let the algorithm do it's thing."
Someone said something like that. Except, there are issues like false positives, questions about the curation of the data (yes, there are cases where something is erroneously included), and so forth.

"Not all AI moderation is bad, there's sentiment analy-"
No, no, no, no, no. That would be hideously inaccurate.

"AI" moderation is pretty bad.