A couple issues with this approach, primarily how to you decide what is “correctly” identified (given that it’s a community ID that is never set in stone)?
The second is that these some of these users are prolific at both observing and identifying, so a volume metric wouldn’t exclude them. That volume is part of the issue - it’s not a big deal if a sporadic user adds a few wrong IDs here or there, but it can be a real issue if an overzealous identifier is adding 100s or 1000s of unsupported or even incorrect IDs.