Facebook’s Political Ad Detection Reportedly Showed High and Geographically Uneven Error Rates

Researchers from KU Leuven in Belgium and New York University conducted a comprehensive audit of Facebook's political ad detection and policy enforcement system, examining 33.8 million Facebook ads that ran between July 2020 and February 2021. The study found that Facebook's AI system made wrong decisions for 83% of ads that required enforcement decisions regarding political classification. Specifically, the system missed 116,963 political ads (62% of the problematic ads) that should have been labeled as political, while incorrectly flagging 40,191 non-political ads (21%) as political. The detection performance varied significantly by country, with Malaysia showing the worst performance where 45% of political ads from obviously political pages were missed. During the U.S. 2020 election moratorium on political ads, Facebook allowed more than 70,000 political ads to run from over 1,000 advertisers who had previously run exclusively political ads. The researchers estimated that advertisers spent $12.2-20.7 million on ads that violated Facebook's policy but were later detected, and an additional $4.6-9.2 million on violating ads that went undetected. The study covered elections in both the United States and Brazil during this timeframe.