Facebook executives were told that its algorithm contained biases against minority groups but chose not to overhaul its software over fears about backlash from "conservative partners" and more, according to The Washington Post. The paper obtained internal company documents outlining how researchers urged executives to aggressively alter the company's algorithm to remove hateful posts.
- To back their request, the researchers showed executives a post that they said represented the "worst of the worst" language on Facebook that was largely directed at minority groups.
- This included a post that used vulgar language to refer to four female Democratic lawmakers.
- Sources told the Post, however, that executives including Facebook VP Joel Kaplan resisted the plans because they were afraid a new system "would tilt the scales by protecting some vulnerable groups over others."
- In 2020, Facebook released the results of a civil rights audit that showed biases in its algorithm and proliferation of hate speech. So far, Facebook has only made some of the changes to comply with the audit recommendations — 65 of 177 have been implemented, and 42 are in progress or are ongoing.
- In September, Facebook (now Meta) was forced to apologize after its AI system mislabeled a video of Black men with a "primates" label.