Latest News

A Challenge to the Ad Industry: It’s Time We Mitigate Bias in Advertising

By Sheri Bachstein | General Manager, IBM Watson Advertising
January 19, 2022

Since the programmatic revolution swept through the advertising industry more than a decade ago, marketers have been using algorithms to determine which demographics to target and how to segment...

Since the programmatic revolution swept through the advertising industry more than a decade ago, marketers have been using algorithms to determine which demographics to target and how to segment audiences in their campaigns.

But today, we’re at a crossroads.

Among growing concerns over walled gardens, equity, transparency and consumer privacy, the industry has another obligation to embrace – working together to swiftly re-architect the future of ad-tech.

Advanced technologies are driving greater efficiency and accuracy, but the potential for unintended consequences — bias, in particular — still exits. Bias is often unintentional, a consequence of human assumptions and judgments. The danger is technology’s ability to capture and scale that bias, ultimately impacting and alienating the very audiences advertisers seek to connect with. That’s why we at IBM believe every brand and agency should take steps to understand bias and its impact on campaigns and consumer trust.

In June 2021, IBM Watson Advertising announced a research initiative to use open-source AI to gauge how prevalent unwanted bias is in digital advertising and to explore how to reduce it in ad campaigns. We started with a hypothesis: Bias exists in digital advertising today, and we can use AI to help identify and mitigate it. Working closely with IBM Research and using established approaches such as the AI Fairness 360 (AIF360) toolkit, we analyzed the factors leading to unwanted bias throughout the advertising funnel and how AI can help.

For the study, we partnered with The Ad Council to understand the possible presence of unintended bias in algorithms and data from a creative message testing effort in their COVID-19 awareness campaign, “It’s Up to You,” the largest U.S. vaccination education effort to date. The campaign ran in late spring 2021 and focused on helping drive awareness of the available vaccine options. 

Early research findings are promising and show that bias does exist in digital advertising and that the mitigation of bias is possible using AI tools and resources. This is a major moment and a testament to both research progress and the advancement of AI as we as an industry hope to build ad tech that will segment groups more equitably while also maintaining or in some cases improving ad campaign performance.

This is a start, but there is much more work to be done. We need the advertising industry to rally around this cause to ensure we are doing all we can.

IBM Watson Advertising is issuing a call to action and looking to collaborate with others to enhance open-source tools, data, and AI models with the goal of creating powerful open technologies that can help automate bias mitigation. We also need brands and agencies to join us in our ongoing research by contributing campaign data, log files and new models for exploration in an open industrywide initiative that will help usher in a new era.

You can read our initial research findings here and learn more in an upcoming discussion hosted by Bob Lord, IBM SVP Weather Company and Alliances, 4A’s CEO Marla Kaplowitz, and Robert Redmond, IBM Design Principal and Head of AI Product Design for Watson Advertising on Wednesday, January 19th here.

If brands, agencies, and publishers stand together with a commitment to advance and adopt best practices that can help mitigate bias in ad campaigns, we can address a problem that has existed for decades for the good of consumers and society as a whole.

Will you join us?

Please join us along with the 4A’s, The Ad Council and IAB in building the future of digital advertising, where unmitigated bias can be an artifact of the past. To learn more about how you can get involved, please visit:


Article Categories