AI has a gender bias and robots are sexist – we must push harder for equality

Last week’s International Women’s Day made me cringe a bit. It’s wild really, that one day a year we ‘celebrate’ women; giving them permission for some humble bragging whilst offering organisations a platform to talk up their equality efforts. A car crash of an occasion where stats around the increase in femicide, news of the worsening gender pay gap, and promo spa vouchers collide.

AI prejudice

We chose to lift the lid on AI and its portrayal of gender. We’ve been playing around with AI for a while, but it’s clearly now mainstream, as proven by the Siddiqui family on Gogglebox recently. We asked Midjourney to generate images based on job title prompts. First, our own… the meta-world version of googling ourselves. Turns out I should have a beard and tattoos.

Then we asked Midjourney to visualise the top 20 highest earning professions in the UK. You can probably guess how it played out… 88% of the images were of men. Most of the time, they presented with exaggerated masculine features. Steel jaws, overtly muscular… decidedly ‘hunky’, as my mum would say. Meanwhile, women were only presented in stereotypical job roles. And even in professional situations, they were commonly sexualised with larger-than-life boobs, unbuttoned tops and sexy pouts.

Reality prejudice

As an agency founder, an ECD and a woman, I still struggle to find that many others that tick all three of those boxes. And there lies the problem - it underscores why the outputs of our experiments are what they are. These tools scrape imagery and data points from up to five billion sources across the web for each search. They work with what’s out there in the ether - they directly reflect the inequalities of society. Not to mention our obsession with fillers, botox and the Love Island aesthetic. Making matters worse, the racial bias of these platforms is just as evident as the gender bias.

So, what can we do? There is no quick fix solution, it is up to society to push harder for equality and employers to deliver the goods when it comes to diversity ambitions. Until then, AI will continue to mirror societal trends. In the short term, we need to be cognisant of machine bias - in a creative setting, we’ll become increasingly reliant on its speed and sophistication, especially to knock out copy and imagery for fast turn-around asks. But will it be pitch perfect any time soon? Very unlikely. Bias is perhaps an unavoidable human trait, but I expect better from the machines and algorithms in charge.

Written by Kat Thomas, founder and global executive creative director of full service earned media agency One Green Bean