IBM Explores AI Tools to Identify and Reduce Bias in Targeting Online Ads
June 24 (Reuters) – IBM Corp (IBM.N) is developing tools that would ensure that online advertising algorithms do not unfairly serve ads to specific groups, such as most men or wealthy people, in an attempt to address issues of discrimination that have attracted the attention of the entire industry.
The company told Reuters on Thursday that a 14-person team would seek “fairness” in advertisements over the next six months, exploring ways to detect and mitigate unintended bias, including in the public and messages themselves.
Academic researchers and civil rights groups have discovered for a decade that certain audiences, including blacks and women, may be excluded from viewing jobs, housing and other ads due to potentially choice. illegal acts made by advertisers or the automated systems they use. Read more
Following complaints from US anti-discrimination regulators and activists, Facebook Inc (FB.O) and Alphabet Inc (GOOGL.O) Google, which are the world’s largest digital ad sellers, have adopted some changes.
But problems remain, as a greater concern over data privacy has already begun to reshape internet marketing.
“The foundations of advertising are crumbling and we have to rebuild the house,” said Bob Lord, senior vice president of IBM. “While we’re at it, let’s make sure that fairness is in the plan.”
IBM’s Watson Advertising unit sells advertisements on its The Weather Company properties and also provides adware to other companies. Customers could potentially be presented with anti-bias offers.
IBM’s initial analysis of its own ad purchases suggests that ads can run fairly evenly to all groups without affecting metrics like the percentage of users who click on an ad, said Robert Redmond. , responsible for the design of AI advertising products.
Redmond gave as a hypothetical example a system said by an advertiser to direct an ad to men across the United States. It can start showing the ad only to those who are in big cities and whites, as this group clicks more. IBM is investigating whether artificial intelligence programs can detect this and then redirect ads more evenly.
The company then plans to analyze data from the non-profit Advertising Council’s public service announcements on COVID-19 vaccines.
“Systemic biases are unfortunately ingrained in almost every corner of the advertising industry,” said Lisa Sherman, President and CEO of the Board.
“We hope this research endeavor will help validate the steps we take to mitigate unintended bias (and) identify areas that we and others need to continue to refine.”
Reporting by Paresh Dave in Oakland, California; edited by Richard Pullin
Our standards: Thomson Reuters Trust Principles.