By now, many people are aware of the concept of “cognitive bias”.
Because humans are humans, it is something that can manifest in advertising creative.
Now IBM’s machine learning engine Watson wants to find evidence of biased advertising, so that it can build a solution.
Uncovering bias
IBM Watson Advertising announces new research initiative to explore AI's role in detecting & mitigating bias in advertising. Learn more: https://t.co/DFcMNsKEIo pic.twitter.com/WrzoymuIYV
— IBM News (@IBMNews) June 24, 2021
Back in June, IBM Watson Advertising announced it had begun a research project to apply open-source AI technology to better understand how prevalent unwanted bias is in advertising.
In this video interview with Beet.TV, Bob Lord, IBM’s Senior Vice President, Worldwide Ecosystems, explains how the company is working with the Ad Council, studying data from their recent “It’s Up to You” campaign to explore whether or not bias can be detected and what kinds of performance outcomes mitigation leads to
“We’re finding that there are biases underneath even the segmentation analysis that are there. These proxy characteristics are coming out and it’s buying decision-makings. And as those algorithms learn more, the machine only perpetuates itself, so that the campaign becomes more biased over time.”
The bane of bias
The IBM Watson Advertising initiative, which is being carried out with IBM Research, has three areas:
- Finding evidence of bias in advertising data and algorithms
- Uncovering the extent to which contextual targeting signals drives that bias.
- Exploring whether or not AI solutions can be developed to mitigate that bias
At launch, it cited a study by Geena Davis Institute on Gender in Media which found male characters appeared in advertisements 12% more than females, despite women-led and gender-balanced videos yielding 30% more views.
The new research strands are also interesting as they come at a time when more people are understanding how programmatic and AI algorithms can themselves be subject to in-built biases.
“I’m actually inviting people to share their data, to see if they have bias in their advertising, prove me right or wrong, and then help me figure out what kind of tools we need to build,” Lord says.
Honored to speak on stage at @advertisingweek today covering a topic I’m passionate about: removing unwanted bias from advertising with #AI. Tune in here at 4:30pm ET #AWNewYork: https://t.co/ImShd7rYmm
— Bob Lord (@rwlord) October 20, 2021
The quantum revolution
It is another way in which IBM is applying its Watson AI toolset to the advertising and media business.
But Lord thinks the next era in computing is going to bring a revolution.
“The merging of classical computing and quantum computing coming together is going to help us solve some of the major unsolvable issues that we have in society,” he says.
“It’s not that far away. We’re running simulations now with the likes of ExxonMobil to figure out how energy is made with Mercedes-Benz, how they can basically make a battery more efficient.
“”There are things that we’re working on now to figure out how those molecular structures work that – you can’t do (that) in classical computing, but you can do with quantum computing.
“I’m pretty excited about where that’s going to go. I think it’s going to be a game changer for us as a human society.”