What Does AI Think a Chemist Looks Like? An Analysis of Diversity in Generative AI
Presenter(s)
Meagan M. Kaufenberg-Lashua
Abstract
With the advent of generative AI technologies has come myriad concerns, one of which is the possibility of biases in terms of race and gender. Testing this in a focused way presents many challenges. We have sought to explore the realities of these biases within the microcosm of our field, chemistry. Using several well-known AI image generators, we have prompted for images of chemists with a generic prompt. Students' perceptions of the gender and racial categorizations of each AI-generated chemist were surveyed to provide pseudo demographics of the generated population. This data has been analyzed alongside students' own perceptions of what a chemist looks like to provide comparisons and to baseline the surveyed student population. We have identified biases in the images generated by all four AI image generators used in this study.
College
College of Science & Engineering
Department
Chemistry
Campus
Winona
First Advisor/Mentor
Joseph West
Location
Ballroom - Kryzsko Commons
Start Date
4-18-2024 1:00 PM
End Date
4-18-2024 2:00 PM
Presentation Type
Poster Session
Format of Presentation or Performance
In-Person
Session
2a=1pm-2pm
Poster Number
20
What Does AI Think a Chemist Looks Like? An Analysis of Diversity in Generative AI
Ballroom - Kryzsko Commons
With the advent of generative AI technologies has come myriad concerns, one of which is the possibility of biases in terms of race and gender. Testing this in a focused way presents many challenges. We have sought to explore the realities of these biases within the microcosm of our field, chemistry. Using several well-known AI image generators, we have prompted for images of chemists with a generic prompt. Students' perceptions of the gender and racial categorizations of each AI-generated chemist were surveyed to provide pseudo demographics of the generated population. This data has been analyzed alongside students' own perceptions of what a chemist looks like to provide comparisons and to baseline the surveyed student population. We have identified biases in the images generated by all four AI image generators used in this study.
