Are widespread stereotypes accurate or socially distorted? This continuing debate is limited by the lack of large-scale multimodal data on stereotypical associations and the inability to compare these to ground truth indicators. In this talk, I will present our recent work in which we address this challenge in the analysis of age-related gender bias, for which age provides an objective anchor for evaluating stereotype accuracy. D

espite there being no systematic age differences between women and men in the workforce according to the US Census, we find that women are represented as younger than men across occupations and social roles in nearly 1.4 million images and videos from Google, Wikipedia, IMDb, Flickr and YouTube, as well as in nine language models trained on billions of words from the internet. 

This age gap is starkest for content depicting occupations with higher status and earnings. We further show how mainstream algorithms amplify this bias. A nationally representative pre-registered experiment (n = 459) finds that Googling images of occupations amplifies age-related gender bias in participants’ beliefs and hiring preferences. 

We additionally show that when generating and evaluating resumes, ChatGPT assumes that women are younger and less experienced, rating older male applicants as higher quality. I conclude by discussing ongoing work that builds on this computational paradigm to show how we can leverage large-scale social data and artificial intelligence to discover novel dimensions of stereotypes that are predictive of human psychology.