Chief Bias Officer position may be necessary to combat algorithmic bias

While many companies continue to debate how to shape their executive teams in response to increasing digitization, and whether they need to create a Chief Data Protection Officer position, some may need to consider creating a “Chief Bias Officer” position, according to an editorial published by Forbes.

Ben Reuveni, who co-founded AI career development company Gloat, argues that the historical problems of bias, made clear by a recent Royal Historical Society review that found bias towards men, and therefore against women, in 87 percent of 108 different economic sectors, could easily be transferred to artificial intelligence algorithms. Reuveni also cites the work of Joy Buolamwini, who has repeatedly found evidence of algorithmic bias in biometric systems, and notes that an IBM research paper has defined and classified 180 different human biases. By basing algorithmic training on past examples, some of the biases are bound to creep into AI systems, he contends.

In response, Reuveni says that companies must consider reforming not only their datasets, but also the processes and personnel involved with AI. Dedicated teams will be required, and possible even a new executive position with the responsibility for scrutinizing data and applying rigorous ethical standards to combat bias.

“A “Chief Bias Officer” could, for instance, affect hiring in engineering teams by pushing for diversity, then regularly checking the data inputs the engineers choose to give the AI algorithms,” Reuveni writes. “This is an example of “personnel is policy.” Recognizing that AI is not infallible is essential: bad inputs mean bad outputs. A Chief Bias Officer would conduct regular performance reviews with controlled tests of algorithms examining outputs, with continuous fine-tuning.”

The problems of bias in AI, and tech more generally, have received increasing attention, including from a U.S. House subcommittee.

Article Topics

algorithms  |  artificial intelligence  |  best practices  |  biometric bias  |  biometrics  |  ethics  |  facial recognition  |  Joy Buolamwini