Skip to main content Skip to navigation

Artificial Intelligence Generates Stereotypical Images of Scientists but Can Also Detect Them: A Pilot Study Using the Draw-A-Scientist Test

Project Overview

The document explores the application of generative AI in education, particularly through the lens of the Draw-A-Scientist Test (DAST), which assesses perceptions of scientists. A pilot study utilizing the Midjourney AI tool generated 1,100 images of scientists, allowing for an evaluation of stereotypical portrayals in the representation of scientists. The findings reveal that generative AI can be leveraged not only to create images but also to analyze and identify stereotypes within these representations. This dual capability of generative AI underscores its significance in science education, highlighting the necessity for educators and students to critically engage with AI tools. By addressing and understanding these stereotypes, the study emphasizes the potential of generative AI to foster more inclusive and accurate representations in educational contexts, ultimately encouraging a more nuanced discourse about the role of technology in shaping perceptions and narratives in science.

Key Applications

Midjourney for image generation and gpt-4.1-mini for analysis

Context: Analyzing public perceptions of scientists through AI-generated images, targeted at educators and researchers in science education.

Implementation: Images were generated using Midjourney with a specific prompt, and analyzed using a scoring rubric (DAST) by both a researcher and gpt-4.1-mini.

Outcomes: The study found that generative AI represents stereotypical images of scientists and can detect these stereotypes with a reasonable accuracy (79%).

Challenges: Issues related to bias in AI-generated content and the need for careful interpretation of the results.

Implementation Barriers

Bias

Generative AI models may perpetuate existing stereotypes about scientists, impacting public perceptions and STEM education.

Proposed Solutions: The need for corrective measures in AI representations and further research to explore the characteristics of AI-generated images.

Project Team

Gyeonggeon Lee

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Gyeonggeon Lee

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies