Google boss Eric Schmidt has called for “spell-checkers for hate” on social media accounts. The call has created some confusion over exactly what forms of censorship, and by whom, Schmidt supports.

The comments come in an opinion piece in the New York Times in which Schmidt addresses the role of the Internet in general, and social media in particular, in recruiting efforts by people promoting violence and terror.

He begins by looking at the issue philosophically, saying tech tools can be used for good and harm, using the analogy “Ever since there’s been fire, there’s been arson.”

The argument then becomes somewhat confused. Schmidt rejects the argument of “authoritarian governments” that censorship is a necessity, countering that “stability and free expression go hand in hand.”

However, he then argues that “We should build tools to help de-escalate tensions on social media — sort of like spell-checkers, but for hate and harassment.” He also says that “we” should remove promotional videos from terrorist groups, but doesn’t clarify whether that “we” covers governments, tech companies, individual Internet users, or all three.

As well as being unclear about who should create the anti-harassment spellchecker, Schmidt offers no details on how it would work. For example, would it try to detect hate speech by flagging up keywords — and if so, would this really be accurate? Would it try to use sentiment analysis, and is that technology reliable enough yet?

There’s also the question of how such a tool would be deployed. Would it simply act as a reminder or warning to users, giving them the option to change their comments before posting, or would it literally block particular words and phrases from being posted? And if it’s purely an ‘advisory’ tool, would users be able to switch it off?