Removing bias in advertising will become a major challenge for marketers and brands in 2022, so a group within the Interactive Advertising Bureau is working to ease the burden for brands and publishers. On Wednesday, the IAB standards working group published a paper to help companies understand bias in artificial intelligence.

Understanding Bias in AI for Marketing: A Comprehensive Guide to Avoiding Negative Consequences with Artificial Intelligence maps out artificial intelligence and machine-learning practices.

Dave Olesnevich, head of product at IBM Watson Advertising and a lead for the working group, recognizes that eliminating and reducing bias is not an easy task and can be quite an undertaking, but the company doesn’t suggest that it can “singlehandedly fix the problem, nor are we attempting to define discrimination or place liability on brands or platforms.”

IBM Watson Advertising executives do believe the companies in a unique position to help based on its support for the open web, rich scientific history, global AI ethics leadership and pioneering work in AI for advertising and other industries.

“So, moving forward, we know we have a sizable problem, and it is going to take a group, like the one we’re co-chairing with the IAB, to come together with shared commitment to attack it,” he said. “That work is hard and will not be solved overnight. Together we must identify what environments are most ripe for bias, what is causing this to happen and what technology needs to do to minimize or, better yet, eliminate it.”

Olesnevich believes it’s important for companies to establish a core team to meet frequently that includes business decision makers, technologists such as developers, engineers, and legal and compliance to work together to establish business requirements and needs. This team should access, test, monitor and assess all levels of bias, establish a process to either eliminate and reduce each bias at a time, and ensure that they have some sort of feedback look from their customers. 

The biggest challenge is a lack of understanding around how AI is used in marketing. There is the perception that AI is inherently bad. The industry needs to do a better job explaining why bias exists and where it comes from, according to Angelina Eng, vice president of measurement and attribution at the IAB.

AI is powerful, processing large data sets more accurately and quickly than humans, but it requires a level of human involvement. It will never be a set it and forget solution. AI advertising and marketing technology requires a team of people from various disciplines and experiences to develop and manage it on-going, Eng said.

When asked to describe the starting point to fix the problem of years of bias, Eng said, “Companies need to recognize that their platforms and systems may have existing, unwanted, and unintended biases. Everyone has a role. Everyone has a responsibility.

Companies need to build processes and frameworks to recognize where bias exists.  There are different types of biases, with different types of outcomes. The guide can help each company to determine where and how.

The guide provides a list of some of those biases and explain what needs to happen at each stage of development, as well as their responsibilities on an on-going basis. 

“In the AI Standards Working Group, we acknowledged that one company’s understanding of where bias is occurring may not be the same for another company,” Eng said. “Collectively as a team, the business decision makers, technologists, and legal and compliance teams need to assess the prevalence of each bias, have a process in place to address them, and decide what to do if they exist. In addition, companies need to be up-to-speed and assess if they are compliant with regulations, internationally, nationally, and locally.”