Hi Richard,
Thank you for weighing in. I’m not entirely convinced that usage is small on Discourse. We have no way to verify installations’ customized code etc. without first building and sharing tools for users to review those systems’ behavior themselves. Even if usage were small, it’s still important to take a public stance because usage elsewhere started small and is now widespread. I’ll have to disagree with your comment here,
So why is that adoption of shadowbanning within the Discourse platform so low? It’s not because shadowbans are bad (which they are), it has another reason. From a technical point of view, shadowbans are very easy to detect on Discourse. Simply visit the forum without logging in or using a different account and spot the differences.
While that is technically possible, most users won’t think to do it. After all, Reddit’s comment removals work the same way, which you can see by commenting in r/CantSayAnything. 99% of Redditors have no idea it works like this. I had an account on Reddit for seven years before I discovered its widespread use of shadow moderation. I had also visited as a reader for probably three years prior to that. And even now that Reveddit has been out for four years, I would estimate that less than 1% of Redditors understand how comment removals work. Finally, even after building Reveddit it wasn’t until four years later, this summer, that I really understood that shadow moderation tools existed and were widely used on all major platforms.
I agree with your point that manipulation of algorithmic feeds are harder to detect. And as you say, non-algorithmically filtered content is reviewable. However, date-ordered comment sections always fit this category, and I believe every major social media site has them. Those are therefore reviewable, provided someone builds tools to do that, because again users don’t think to check this themselves. Even if you do check a comment’s status once, it’s possible for a moderator to quietly remove your comment an hour later, thereby invalidating your check! This is why I built Reveddit Real-Time1, an extension that polls the system for shadow removals and notifies you when that happens.
Such tools should be built for all platforms until there is sufficient trust between users and social media sites that shadow moderation is in the rearview mirror. Shadow moderation is so widespread that I think everyone should assume it’s being used by whatever system they’re using to communicate, unless the system expressly denies the practice. And even in that case you need to watch their wording, such as Twitter’s choice to publicly use the word “rank”1 rather than internal descriptors like “de-amplify”1. YouTube, on the other hand, more openly acknowledges it will “reduce”1 content.
transparent moderation can still be subjective.
Can you expand on this?
Besides my work in the Discourse ecosystem I’m also founder of a company that combats online disinformation and hate speech.
Interesting, can you say which one? And are the talks you mention available online? I agree that YouTube’s bans are also concerning. I suspect their moat, like others’, was built using shadow moderation tools that “reduce” or “shadow remove”. We should therefore focus on that first.
I think the real censorship of our era is shadow moderation. Transparent moderation can be censorship if it is not “viewpoint neutral”, but it is more often akin to editing. A final case against real censorship comes from Bob Corn-Revere, a leading first amendment lawyer: “the censor never has the moral high ground.”1