Here's an example of where metrics go wrong and create bad incentives. What Twitter (and virtually everyone else, to be fair) are doing is optimizing for the n

Here's an example of where metrics go wrong and create bad incentives. What Twitter (and virtually everyone else, to be fair) are doing is optimizing for the number of issues taht are dealt with automatically by some ML system or whatever. This is a bad idea because it incentivizes false positives and the only recourse is kicking up a big enough stink on social media such that an actual human looks at it. Even "appeals" get automated in this universe. Here's what you should optimize for: the number of cases a human support person can deal with (per unit time). Appeals must be manually reviewed so will drop the cases/hour. This disincentivizes false positives. If you optimize for this kin dof system you will use automated systems to funnel cases to a human for review. I know of several people who have gotten suspended or banned off these platforms for completely stupid reasons that any human looking at it would recognize. I'm wary of bad regulation here and certainly don't think people have a "right" to be on Twitter (for eample) but I do believe people should have a right to human review for any ban and you should able to challenge that ban in court if you have to. There should be no "we can ban you for any reason" out. Twitter in particular is so bad that the usual avenue of getting something to an employee is met with limited success because this had become so rampant (ie necessary) that Twitter was internally cutting down on that. Lastly, any such ban should be segmented. Google is the biggest violator of this principle. You should never, for example, lose access to your Gmail because an automated system falsely detected a questionable photo you uploaded or another automated system decided your name wasn't "real" (the latter happened in the Google+ days).