Why we need more diverse STEM role models

When you ask cutting-edge generative AI to produce images of people in various science, technology, engineering, and math (STEM) occupations, the result is overwhelmingly white and male.

Figure 1: Images generated with Stability AI’s “stable-diffusion-xl-base-1.0” using the prompts “scientist” through “information security analyst”.

According to 2021 statistics from the National Center for Science and Engineering Statistics (NCSES), women make up 35% of the U.S. STEM workforce. When comparing statistics for computer science majors, the percentage is even lower. There is no clear answer as to why this is the case. Some might argue that women are simply less interested in STEM than men. However, in an article in Scientific American, the authors argue that while this claim may be true, the vast majority of male role models may be discouraging the women and girls who actually want to pursue STEM, further widening the gap.

The images above were created using Stable Diffusion XL 1.0-Base by StabilityAI, which had over 3 million downloads as of February 2024 alone.

The popularity and widespread use of models like this one begs the question: what image of who participates in STEM do we want to promote and thus reinforce?

The initiative IF/THEN, funded by Lyda Hill Philanthropies, made it its mission to encourage young women to pursue careers in science, technology, engineering, and math by highlighting and supporting female trailblazers in STEM.

Their exhibition, #IfThenSheCan, features 120 life-size, 3-D printed statues of contemporary female innovators. Each statue is bright orange and includes a QR code that visitors can scan to learn more. All of the statues were displayed together at NorthPark Mall in Dallas in 2021 and at the Smithsonian Institution in Washington DC in 2022. Since then, #IfThenSheCan has appeared in pop-up exhibitions across the country.

In February 2024, the pop-up exhibition came to the Stanford University campus and featured five women in STEM: biologist and science communicator Jenny Briggs, roboticist and dancer Catie Cuan, engineer and toy maker Debbie Sterling, microbiologist and immunologist Dorothy Tovar, and polymer chemist Helen Tran.

Figure 2: The five IF/THEN statutes from the IF/THEN pop-up exhibition
at Stanford University in February 2024.

As society continues to rapidly digitize, we must ensure that the technology being developed considers all of humanity. A diverse STEM workforce is the first step in ensuring that more first-hand perspectives are brought into research questions, codebases, applications, and thinking from the beginning, not as an afterthought.

Initiatives and clubs could help more women realize that there is a place for them in STEM careers. And perhaps we will reach a point where the proverbial scientist or computer engineer conjures up a diverse set of images — not just one.

Appendix

Following is the code used to generate the images for Figure 1. Each prompt was used in succession.

from diffusers import DiffusionPipeline
import torch

pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16, use_safetensors=True, variant="fp16")

pipe.to("cuda")

prompts = [
"computer programmer",
"rocket scientist",
"information security analyst",
"microbiologist",
"scientist",
"chemist",
"mathematician"
]

images = []
for prompt in prompts:
images.append(pipe(prompt=prompt).images[0])

# display images below...

References

Cheryan, Sapna, Master, Allison, and Meltzoff, Andrew. “There Are Too Few Women in Computer Science and Engineering.” Scientific American, 27 Jul. 2022, https://www.scientificamerican.com/article/there-are-too-few-women-in-computer-science-and-engineering/. Accessed 14 Mar. 2024.

Wander, Robin. “Breaking the mold: Vibrant statues highlight women in STEM.” Stanford Report, 14 Feb. 2024,