Nov 21 2018   7:05AM GMT

Beware of unwitting bias in AI algorithms, experts say

Nicole Laskowski Nicole Laskowski Profile: Nicole Laskowski


Companies need to know how bias can unwittingly creep into recommendations from AI algorithms, according to panelists at the recent AI and the Future of Work event at MIT.

Sophie Vandebroek, vice president of emerging technology partnerships at IBM, suggested companies start with scrutinizing their training data. “AI algorithms are only as good as the data used to train the system,” she said.

For example, an AI algorithm trained on data from a company that has more men than women software engineers might conclude that men are better software engineers than women, “which of course we know has nothing to do with the job and is irrelevant to the hiring decision,” Vandebroek said.

Checking the training data for bias is necessary but not sufficient for ridding AI algorithms of bias, said Gabi Zijderveld, chief marketing officer and head of product strategy at Affectiva.

Bias can also be subconsciously coded into an algorithm by developers. “We build what we know,” Zijderveld said. She recommended that companies strive for diversity when putting together their teams, as they can collectively act as an anti-bias failsafe. Besides, she said, diversity also “fosters creativity and innovation.”

CIOs can also help set the tone by investing in a strong culture, according to Nichole Jordan, managing partner of markets, clients and industry at advisory and accounting firm Grant Thornton LLP. “You’re bringing together a lot of individuals with different backgrounds — behavioral scientists, data scientists, sociologists — to work together,” she said. “You’ve got to be clear on your culture.” It’s important that a company’s ethics, values and acceptable behaviors are spelled out to employees.

1  Comment on this Post

There was an error processing your information. Please try again later.
Thanks. We'll let you know when a new response is added.
Send me notifications when other members comment.
  • jigman
    How was said: "Bias can also be subconsciously coded into an algorithm by developers", but I think that the meaning of the word "bias" could be in risk by it common use by the AI market. "Bias" is a concept obligatory to AI generalization, inside the algorithms! Why don't create a new buzzword that mean specifically all the controversy of AI's applications results?! Because, IMHO, this "bias" coded subconsciously, is a lack of data preparation in the right manner and, a lot of uncertainty to what analyse.
    0 pointsBadges:

Forgot Password

No problem! Submit your e-mail address below. We'll send you an e-mail containing your password.

Your password has been sent to:

Share this item with your network: