The right kind of fake data is better than real data for training people to successfully predict who will win an event such as a baseball game, a new study has found.
University of Montreal psychologist Gyslain Giguère and Bradley Love at University College London trained 84 U.S. college students to predict the outcome of Major League Baseball games between two teams by providing them with either the real outcomes of games or fictional outcomes where the higher ranked team always won.
The study, published this week in Proceedings of the National Academy of Sciences, found that when the two groups were later tested and asked to predict the outcome of other matches, the students trained with the fake outcomes performed better. That is the opposite of what happens when computers are trained using real versus fake or "idealized" data.
CBC Radio's national science columnist Torah Kachur explains why humans learned better with the fake data and discusses the broader implications for training human professionals in other walks of life.