Guardado en:
| Autor principal: | |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | https://arxiv.org/abs/2504.19005 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866915261766959104 |
|---|---|
| author | Lee, Gyeonggeon |
| author_facet | Lee, Gyeonggeon |
| contents | How the general public perceives scientists has been of interest to science educators for decades. While there can be many factors of it, the impact of recent generative artificial intelligence (AI) models is noteworthy, as these are rapidly changing how people acquire information. This report presents the pilot study examining how modern generative AI represents images of scientist using the Draw-A-Scientist Test (DAST). As a data, 1,100 images of scientist were generated using Midjourney v 6.1. One hundred of these images were analyzed by a science education scholar using a DAST scoring rubric. Using the data, the researcher went through prompt engineering to instruct gpt-4.1-mini to automatically analyze the remaining 1,000 images. The results show that generative AI represents stereotypical images of scientists, such as lab coat (97%), eyeglasses (97%), male gender (81%), and Caucasian (85%) in the 100 images analyzed by the researcher. However, gpt-4.1-mini could also detect those stereotypes in the accuracy of 79% in the same 100 images. gpt-4.1-mini also analyzed the remaining 1,000 images and found stereotypical features in the images (lab coat: 97%, eyeglasses: 95%, male gender: 82%, Caucasian: 67%). Discussions on the biases residing in today's generative AI and their implications on science education were made. The researcher plans to conduct a more comprehensive future study with an expanded methodology. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_19005 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Artificial Intelligence Generates Stereotypical Images of Scientists but Can Also Detect Them: A Pilot Study Using the Draw-A-Scientist Test Lee, Gyeonggeon Physics Education How the general public perceives scientists has been of interest to science educators for decades. While there can be many factors of it, the impact of recent generative artificial intelligence (AI) models is noteworthy, as these are rapidly changing how people acquire information. This report presents the pilot study examining how modern generative AI represents images of scientist using the Draw-A-Scientist Test (DAST). As a data, 1,100 images of scientist were generated using Midjourney v 6.1. One hundred of these images were analyzed by a science education scholar using a DAST scoring rubric. Using the data, the researcher went through prompt engineering to instruct gpt-4.1-mini to automatically analyze the remaining 1,000 images. The results show that generative AI represents stereotypical images of scientists, such as lab coat (97%), eyeglasses (97%), male gender (81%), and Caucasian (85%) in the 100 images analyzed by the researcher. However, gpt-4.1-mini could also detect those stereotypes in the accuracy of 79% in the same 100 images. gpt-4.1-mini also analyzed the remaining 1,000 images and found stereotypical features in the images (lab coat: 97%, eyeglasses: 95%, male gender: 82%, Caucasian: 67%). Discussions on the biases residing in today's generative AI and their implications on science education were made. The researcher plans to conduct a more comprehensive future study with an expanded methodology. |
| title | Artificial Intelligence Generates Stereotypical Images of Scientists but Can Also Detect Them: A Pilot Study Using the Draw-A-Scientist Test |
| topic | Physics Education |
| url | https://arxiv.org/abs/2504.19005 |