Facebook's Algorithms Accidentally Flagged The Declaration Of Independence

36shares

Facebook automatically flagged and took down passages from the Declaration of Independendence last week. The reason? Hate speech.

Liberty County Vindicator, a newspaper in Texas last week posts excerpts from the Declaration of Independence to celebrate Independence Day on July 4. No harm done, right? Well, not really. Apparently, Facebook‘s algorithm took offence to some passages, mistaking them for… hate speech. While it may be hard to believe, the algorithm actually did its job very well.

Facebook mistakenly flagged excerpts from the Declaration of Independence last week for hate speech. But was it actually justified in doing so?

Click To Tweet

You see, paragraphs 27-31 of the Declaration (the ones in the post that was flagged) contain a the sentence “the merciless Indian Savages.” Referring to Native American Indians in this way is indeed what one may call “hate speech.” While that is true, it just shows how much Facebook’s AI lacks context and the ability to recognise recognise historical context.

Facebook’s AI is pretty good at catching hate speech in posts, but it seems to lack historical context.

Click To Tweet

The post was restored of course, and Facebook’s VP of public policy, Richard Allan, published a Hard Questions blog post addressing the mix-up. “We’re a long way from being able to rely on machine learning and AI to handle the complexity involved in assessing hate speech,” he explained.