How Adobe's Ethics Committee Helps Manage AI Bias

By Jared Council 

Review boards can help companies mitigate some of the risks associated with using artificial intelligence, according to Adobe Inc. executive Dana Rao.

Mr. Rao, Adobe's general counsel, said one of the top risks in using AI systems is that the technology can perpetuate harmful bias against certain demographics, based on what it learns from data. Ethics committees can be one way of managing those risks and putting organizational values into practice.

Adobe's AI ethics committee, launched two years ago, has been able to review new features for potential bias before those features are deployed, Mr. Rao said Wednesday at The Wall Street Journal's Risk & Compliance Forum. The committee is made up of employees of various ethnicities and genders from different parts of the company, including legal, government relations and marketing.

"It takes a lot of people across your company to help figure this out," he said. "Sometimes we might look at it and say there's not an issue here," he said, but getting a diverse group of people together can help identify issues product developers might miss.

A feature designed to detect unauthorized purchases of Abode software, for instance, could inadvertently learn to block customers of a certain demographic, he said.

Mr. Rao said the AI ethics committee recently reviewed a fraud-detection feature that could potentially discriminate against certain groups.

"[AI] can learn from last names and geographies and make a connection that...there's a lot more credit card fraud coming from Brazil," Mr. Rao said. "And it may not just stop people from Brazil coming in, which would be bad. It might stop people with Brazilian names from [purchasing] the software."

He said the committee directed the team proposing the feature to conduct further testing.

Mr. Rao said an AI ethics assessment is one mechanism for identifying which features Adobe's ethics committee should review. This is a form product developers fill out that spells out what the AI feature will do and how it works.

If the assessment review shows no major ethical AI risks, such as an AI tool that recommends text font, then the team behind it doesn't need to present the feature to the board for review, he said.

It's also important for the review board to be made up of diverse voices, he said.

"We've had examples where some African-American members on our committee have spotted issues in an imaging AI filter that no one else did, because it affected only people with Black skin or their hair specifically and no one else would have gotten the issue."

Write to Jared Council at [e-mail]

(END) Dow Jones Newswires

May 05, 2021 21:01 ET (01:01 GMT)

Copyright (c) 2021 Dow Jones & Company, Inc.