Online advertising’s racism mess - Salon.com
Harvard privacy researcher Latanya Sweeney found that Google's AdSense algorithm was showing arrest-record ads 25 percent more frequently when searching for names perceived as Black (like Latanya or Rasheed) compared to white-sounding names. The algorithm wasn't programmed to discriminate; rather, it was mechanically learning from aggregate user behavior—racist employers clicking arrest ads more often for Black-identified names taught the system to associate those names with criminal records. This research illustrates how automated systems can amplify and perpetuate societal racism without any intentional bias in the code itself.