Social Platform Trust & Safety

Incentives for productive speech: A speech community is likely to produce whatever kind of discourse it incentivizes. This means that users should have formal or informal incentives to produce speech that is recognized by other users as worthy.

A scholarly culture that reads and discusses enduringly great works incentivizes the production of such wo... See more

All of these ideas are free-speech friendly. They do not involve top-down censorship, but bottom-up user choice. Letting people police the content on their own pages and feeds is the natural next step for platforms that want to empower users rather than constantly surveil and censor them. Such features are also just common sense. No one has an inhe... See more

Centralised moderation isn't the only option that doesn't work. We also need to eliminate any option that doesn't have a credible mean of doing the work — and it's a lot of work, meaning that either someone needs to get paid or you need a lot of volunteers. And punting it to the users can never work.

For synthetic media to reach millions of views—as the Swift images did in just hours—they need massive, aggregated networks, which allow them to identify an initial audience and then spread. As the amount of available content grows with the broader use of generative AI, social media’s role as curator will become even more important.

A crucial element to lowering the stakes of any particular moderation decision is the knowledge that the user has the genuine option to go to another community or start his own. The same applies to moderators: The ultimate check on a moderator’s power is that if most users start to believe he is using it poorly, users can either protest until a new... See more