How can human-centered AI fight bias in machines and people?