Nature Podcast · Springer Nature Limited

How stereotypes shape AI – and what that means for the future of hiring

·36 min·3 clips

ChatGPT gives higher scores to older men in hiring, perpetuating stereotypes learned from internet images.

Reporter Sharmini Bandel interviews Stanford Graduate School of Business researcher Douglas Gilbeau about a new study published in Nature. The research analyzed a massive dataset of images from Google Image Search, Wikipedia, and the Internet Movie Database (IMDb), systematically documenting patterns in how different genders are represented. The IMDb component was particularly valuable for research purposes because it contains verified biographical data: researchers could confirm exact ages of people at the time images were taken. The consistent finding across sources was that women in these images tend to be represented at significantly younger ages than men. This is not random noise — it applies across job categories, with certain roles showing clear age-gender stereotypes even when the underlying workforce reality is different. The episode explores two levels of harm. First, the representation itself shapes how people perceive who belongs in various roles — impressionable young people seeing consistent age-gender patterns in images of doctors, executives, or scientists internalize those patterns as normal. Second, these biased image datasets are the training data for AI systems now used in hiring and screening contexts. AI that learned from biased images will reproduce and amplify those biases in recommendations. The researchers discuss what changes to internet image culture, image tagging practices, or AI training methodologies could interrupt this feedback loop. The episode also covers astronomers' favorite exoplanet discoveries and the 2025 Nobel Prizes.

As heard by us

A brisk, broad Nature Podcast episode that moves from internet images to exoplanets, Nobels, and the show's 20th anniversary.

The Nature Podcast episode works best when it starts with stereotypes in internet images and keeps the tone plainspoken and direct. From there, it shifts into exoplanets, with a clear stop at Proxima Centauri and the first planets found there, then widens to the 2025 Nobel…

Read the full review in PlayNext →

Why you'd press play

You want a clear look at how stereotypes can lurk in images.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Nature Podcast