The "random sample" part of the solution is good. The "trusted polls" part of the solution is not good, because who decides if a poll is trusted? There are cert

The "random sample" part of the solution is good. The "trusted polls" part of the solution is not good, because who decides if a poll is trusted? There are certainly a lot of polls I don't trust, because I suspect them of 1. cheating or being lazy with the sampling 2. Being a weasel with the phrasing to get the desired result 3. Being a push poll. Still, a "trusted" poll is slightly better than a freeform "community note", especially if it sticks solely to how prevalent an opinion is. Slashdot used random sampling in moderation 30-ish years ago. It worked OK, except that scores were used for very little (crucially they didn't even sort by them), and they had a more gameable non-randomized system to moderate the random system. And of course it was probably vulnerable to Sybil attacks. (By the way, I guessed 4% for the number of toxic users)