[mrtksn]

> when we manually verified an account to be a scammer

This makes all the difference with other services that block out users only to let them guess why they were blocked.

If an automated system did that, I would have said it's evil. Yet, I hope you have a communication channel in case there was a human error.

[jjoonathan]

Yes, although I would add an attention threshold too, as it's not entirely unknown for hired manual review to just spam the "guilty" button so they can get to lunch. In any case: your false positive rate needs to be massively low if you want to be a massive asshole to the people it flags -- or else you are just an asshole.

If you can afford to get the FPR down, sure, have fun, but if not, please have the decency to not pretend.

[leshenka]

You can implement a jury trial system - have a pool of moderators, select a few at random and have them look at the account, only flagging it if there is a consensus that it’s a scam account

[SQueeeeeL]

That scale doesn't really work for small startups

[vanviegen]

Admittedly, there is the occasional false positive. For such cases, we display an email address right underneath the error message. Scammers rarely dare to complain, and when they do, they are usually not very convincing.

[jjoonathan]

These humane touches make all the difference. Thanks for taking the time.