Facebook outlines its work on preventing bias in AI

At its F8 developer conference held on 1–2 May in California, Facebook outlined its work on trying to avoid bias in how its artificial intelligence (AI) systems are functioning. Among other things, the company announced that it has a special team focused on ethics in AI, and has developed solutions to avoid bias in AI-based decisions. One such solution is Fairness Flow, a tool which can determine potential prejudices for or against particular groups of people. According to research scientist Isabel Kloumann, cited by Cnet, Fairness Flow was initially used to ensure that job recommendations were not biased against some groups over others. But the company has now integrated the tool into its machine learning platform, and is working on applying it, ‘to evaluate the personal and societal implications of every product’ it builds. Facebook also described, in a blog post by Guy Rosen (Vice-president of Product Management), how it is using AI to detect and remove ‘bad content, like terrorist videos, hate speech, porn or violence’ from the platform.