> it is hard to apply such a criteria across borders. In California, for instance, political affiliation is considered a protected class.
This is hard. But it's a different category of hard from the problem Google has chosen for itself.
Let's assume California has a protected class that New York doesn't. China has protected classes the U.K. doesn't. That could mean a comment visible in New York isn't in California. Or it could mean something else.
This isn't an easy problem. One needs to map comments to various protected classes and then measure "hatefulness" according to local constructs. One needs to decide what to do with content so flagged. But that's objectively easier than doing all of that plus coming up with the definitions for protected classes.
Viewed through this framework, deleting a comment, as Google is doing here, is almost always wrong.
> it essentially legalizes political persecution
The First Amendment is supposed to protect against this. Protected classes govern private actors.
This is hard. But it's a different category of hard from the problem Google has chosen for itself.
Let's assume California has a protected class that New York doesn't. China has protected classes the U.K. doesn't. That could mean a comment visible in New York isn't in California. Or it could mean something else.
This isn't an easy problem. One needs to map comments to various protected classes and then measure "hatefulness" according to local constructs. One needs to decide what to do with content so flagged. But that's objectively easier than doing all of that plus coming up with the definitions for protected classes.
Viewed through this framework, deleting a comment, as Google is doing here, is almost always wrong.
> it essentially legalizes political persecution
The First Amendment is supposed to protect against this. Protected classes govern private actors.