Hacker News new | past | comments | ask | show | jobs | submit login

With sufficiently complex systems to detect fraud, such as AI or that involve constraint optimisers, it is not always doable to isolate a single, clear parameter that is responsible for the data to be flagged. This is an issue also for them (source: been working on several complex rule based systems that exhibited this very problem).

As for the human powered appeal process, how to design it so that it is not spammed/abused as easily as the upload?

It's easy to explain everything away because of evil corporation, but those are also actual technical issues.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: