Behavioral scientists have long used implicit bias tests to measure hidden preferences, but research is now revealing significant methodological blind spots in these widely-used tools.