"our recommendations algorithms will start pushing you down a rabbit hole of more egregious content."

Miki Rothschild: "[W]ith the recent PR fire TikTok has seen around inappropriate content for teens (see here), we were chatting within Youth XFN about our own deficits when it comes to discoverability, namely: (1) it's still very easy to find borderline and sometimes violating content/accounts in search, e.g. try typing 'drug' or 'love' (2) if you start following borderline accounts or interacting with such content, our recommendations algorithms will start pushing you down a rabbit hole of more egregious content." "3. How are we thinking about non-rec/violating Rabbit holes in recommendations? Do we have any specific approaches to this?" ████: "On #3 we've done some early experimentation with the Drebbel team on additional enforcement for users who we detect to be in a rabbithole on Feed Recs. TL;DR we did see a meaningful drop in exposure with targeted demotion, but it came with a clear engagement cost and the effects did not last after intervention."