AI isn’t enough to protect social media communities from AI

AI’s biases Typical social media AI-based moderating systems use machine learning classifiers to analyze posts and identify and flag content that breaks platform rules. But it’s difficult for a machine to understand the nuances of sarcasm, satire, and slang. Further, some research (examples here, here, and here) suggests that marginalized groups can be disproportionately affected by AI moderation. Without human oversight, AI can end up penalizing the very communities most vulnerable to the hateful content the systems are designed to combat. Gilbert, who is also the research director of Cornell’s Citizens and Technology Lab, says that “marginalized and vulnerable populations are among those who experience the highest rates of moderation, and that typically this is a result of ‘false-positiv