Bob Pearson, co-author of Countering Hate and CIO of W2O Group says that AI is able to identify hate speech today.
All human beings follow patterns online. You can see what language, content, channel, and people matter to them. You can see which words trigger information seeking, which language is most associated with hate topics or sites, which people are the most important influencers and you can see a range of behavioral characteristics.
Except that humans can’t define “hate speech;” we can’t even define “hate.” All that can be done is for each of us, individually, to identify what it seems like that to us—not what it is to anyone else. A Justice’s remark about pornography—that he couldn’t define it, but he knew it when he saw it—is worse than wholly inadequate when Government tries to regulate speech, or when private enterprises try to regulate speech in the public spaces they create as their business models or offshoots of them. Such regulation is a threat to individual liberty.
Artificial Intelligence certainly is not up to this task of discriminating unacceptable speech from acceptable. The definitions of those terms will be programed by a select group of humans—who can’t define the terms except as they personally apply the terms to themselves. This frangibility is demonstrated by those cases that go to jury trials and the differing outcomes differing juries reach on substantially similar cases. All AIs can do is reflect the prejudices of their programmers.
How, indeed, is “hate speech” to be discriminated from the merely rude or offensive or uncivil speech? To narrow the thing a bit, think about civility—simple courtesy. Why should a New Yorker’s version of civility be forced to take a back seat to a Midwesterner’s or a Californian’s version of civility? Why should one of the other two be forced to take a back seat to a New Yorker’s version? On what basis would any definer pick one or identify a middle ground that doesn’t wind up being a more insidious censorship?
Consider Marshall McLuhan’s the medium is the message. Sometimes the overt rudeness—even “hate”—is a demonstration of a point. Is burning our national flag an act of hatred or political speech? Is posting a picture of Mohammed…? Prove it: it’s a hard line to draw between that and rudeness on the one hand or actual hate on the other.
Not even civility is “simple.” How is something as complex as hate speech to be handled?
Yet the whole argument over identifying, and then censoring, hate speech is irrelevant other than the threat to freedom that is that effort to identify and censor.
The problem he [Pearson] says, is that Facebook and other companies have not taken up the charge to make the battle against hate speech a major priority.
No, Pearson is utterly wrong on this. Facebook, et al., must not take up the battle, any more than can Government be allowed to. Any effort along these lines can only be the despicable bigotry and the equally shameful cowardice of censorship. Bigotry because all such efforts amount to is one man, or a small group of them, imposing his prejudice of acceptability on the speech of all others.
Hackneyed as it’s become, Justice Louis Brandeis still is right: the answer to bad speech—however defined—is more, and better, speech—however defined.
Engage in the discourse, don’t cower away from it.