Hemphill: Effective hate speech moderating must go beyond civility

Moderating online content for hate speech usually is an exercise in content-matching: profanity and incivility is flagged. Associate professor of Information, Libby Hemphill, argues that “Big Tech” needs to move beyond basic moderation to truly end extremist speech. 

In an article for VentureBeat, Hemphill says that current content moderation overlooks how hate speech is used to punish, humiliate and control marginalized groups. She says that prioritizing civility online has not only allowed civil but hateful speech to thrive, it has even normalized white supremacy.

Hemphill’s research, including during her time as an Anti-Defamation League Belfer Fellow, found that white supremacists tend to avoid profane language and weaponize civility against marginalized groups. She also found that these extremists often platform-hop, avoiding detection and working around bans. 

Her research shows that machine learning can be used to identify patterns in white supremacist speach that can be used to improve the detection and moderation systems in tech platforms. Using training data from a far-right, nationalist website, Hemphill’s team trained algorithms to detect sentence structures, specific phrases, and recurring themes. 

“Ultimately, research shows that social media platforms would do well to focus less on politeness and more on justice and equity,” says Hemphill. “Civility be damned.”

RELATED

Read “To truly target hate speech, moderation must extend beyond civility,” on venturebeat.com.

Read more about Libby Hemphill’s work as a Anti-Defamation League Belfer Fellow and her report “Very Fine People: What Social media Platforms Miss About White Supremacist Speech.”

Read more about Dr. Hemphill’s research on how machine learning technology can be used to moderate hate speech.