One of the great promises of artificial intelligence (AI) is a world free of petty human biases. Hiring by algorithm would give men and women an equal chance at work, the thinking goes, and predicting criminal behavior with big data would sidestep racial prejudice in policing. But a new study shows that computers can be biased as well, especially when they learn from us. When algorithms glean the meaning of words by gobbling up lots of human-written text, they adopt stereotypes very similar to our own.

“Don’t think that AI is some fairy godmother,” says study co-author Joanna Bryson, a computer scientist at the University of Bath in the United Kingdom and Princeton University. “AI is just an extension of our existing culture.”

The work was inspired by a psychological tool called the implicit association test, or IAT. In the IAT, words flash on a computer screen, and the speed at which people react to them indicates subconscious associations. Both black and white Americans, for example, are faster at associating names like “Brad” and “Courtney” with words like “happy” and “sunrise,” and names like “Leroy” and “Latisha” with words like “hatred” and “vomit” than vice versa.

For more information, see: http://www.sciencemag.org/news/2017/04/even-artificial-intelligence-can-acquire-biases-against-race-and-gender