• 1 Post
  • 14 Comments
Joined 1 year ago
cake
Cake day: June 16th, 2023

help-circle






  • I doubt he’s ignoring anything. And I know nothing but I think it’s a little unfair to bash him for this.

    Meta does not need the Fediverse to create a ready-populated instance all of its own. It doesn’t need to federate with anyone, it can probably kill Twitter and Reddit with a single stone (if it pours enough resource into moderating and siloing). Just stick a fediwidget in every logged in account page with some thoughtful seeding of content and it’s done.

    The danger of federating with Meta is much the same as not federating. It has such a massive userbase it will suck the lifeblood out of everywhere else whether or not it can see us.

    The possible silver lining is that there are other very large corporates which can do the same (some of which have said they plan to). We could all end up with multiple logins on corporate instances simply because we have accounts with them for other reasons. And that means a lot of very large instances with name recognition, and easy access, making it much harder for any of them to stop federation and keep their users to themselves.

    Being federated with one or more behemoths might well be hell. Some instances won’t do it. Moderation standards will be key for those that do. But multiple federated behemoths can hold each other hostage because their users can all jump ship to the competition so easily.

    This is much, much more complicated than just boycott or not. They cannot be trusted one tiny fraction of an inch but this is coming whether we like it or not. We need to work out how to protect ourselves and I’m starting to think that encouraging every site with a user login to make the fediverse a widget on their account pages might be the very best way to do it.




  • Thanks for that.

    Lack of blinding is a serious issue for subjective outcomes but blinding when treatment effects are obvious to both intervention and control groups is dishonest (Pharma does it all the time to make their trials look more credible than they are).

    Open label is the norm for cancer trials for exactly this reason. It is important to consider the biases that may arise, in subjective endpoints especially. But it is ludicrous to dismiss research on this basis alone. We can’t randomise 12 year olds to become lifetime smokers or not, let alone use placebo controls, but we do know that smoking kills. It’s just a bit more complicated to prove it when perfectly designed RCTs are not possible.