Significant changes are afoot in the advertising industry. In the last month alone, Netflix announced it may enter the ad business, lawmakers introduced bipartisan bills to throttle Google’s digital ad dominance and Facebook rolled out changes to help advertisers achieve more precision in their targeting. As major players prepare, advertisers have an opportunity to manage these changes in a way that optimizes ad spending and addresses the problem of bias in ad technology.
Bias is a well-known issue for the ad industry, and the programmatic technologies the companies have adopted to supercharge marketing campaigns may not be improving matters. Nearly $1 trillion of digital media flows through programmatic engines that segment and target specific audiences, sometimes missing large consumer groups in the process. Not only can that contribute to improper bias, but it’s also an inefficient way to spend your ad dollars.
It’s why, as an industry, we must tap into AI and leverage the powerful tools at our disposal to help mitigate the bias problem.
As AI algorithms come to dominate in the industry’s efforts to find audiences and serve ads, we must integrate mitigation tools to avoid reinforcing biased thinking. That is, rather than letting AI exacerbate the problem, we must make the technology part of the solution. Doing this can help bring fairness by adapting ad buying behavior to reach more diverse audiences. By embedding fairness metrics and AI algorithms into the core of marketing processes, we can deliver a more effective value exchange between consumers and brands and potentially generate improved ROI on media dollars spent.
The technology needed to mitigate bias in ads already exists, and companies in finance, human capital management, healthcare, education and many other industries are testing open-source toolkits that build bias mitigation into their marketing processes. It’s time for the advertising industry to make a concerted effort to build fairness into our marketing technology as well.
AI bias occurs when the machine learning process used to create AI models places certain privileged groups at a systematic advantage and certain unprivileged groups at a systematic disadvantage. Such bias could impact a financial institution’s ability to fairly assign credit scores or issue mortgages, or it could affect an insurance company’s ability to accurately predict medical expenditures for different clients.
In advertising, bias can prevent consumers from being exposed to certain brands and information based on flawed algorithmic analysis. Often, this does harm to both the consumers and the brands. Embedding fairness metrics and AI algorithms into the marketing processes could enable the technology to, for example, automatically — and at scale — generate anomaly reports when something doesn’t look right with the data indexing as media plans are executing.
If such a fairness solution can be applied to the core of how we do marketing today, we could not only help reduce bias, but also potentially help brands get a better return on their media spending.