Chapter 5. What’s Wrong With This Picture? Japanese Students Explore Racial and Gender Bias in AI-Generated Images
Matthew Wiegand
Abstract
Along with realistic text-based output from large language models for L2 teaching, AI-generated imagery (AIGI) has become increasingly complex, detailed, and realistic. However, there are ethical concerns about these technologies. In some cases, the technology is used in misinformation and disinformation campaigns that target people of all ages across realms ranging from politics to science and education. While earlier image models produced strange images that were easily identifiable as AIGI, recent technology has become increasingly photo-realistic - sometimes nearly indistinguishable from photography- thus making it harder to identify as AI. Furthermore, research has shown that AIGI often images that exhibit racial and gender biases, the implications of which are concerning due to their potential to reinforce racial and gender-based stereotypes. However, there has been little research into the ability of English as a Second Language (L2) students to identify and articulate the biases present.
This chapter introduces research exploring L2 students' recognition and discussion of bias in AIGI. Students generated images based on prompts discussed and agreed upon in class, then considered the output. Prompts were based on nationality, occupation, and lack of income. The study suggests that students with a closer knowledge of the images being produced were better able to identify bias in the imagery. The author offers further suggestions for addressing the issue of helping students recognize bias in AIGI, including integrating media literacy into second-language learning programs.
About the Contributor
Matthew Wiegand holds BAs in Cultural Anthropology from UC Santa Cruz and Global Japanese Studies from Meiji University, and an MA from the Graduate School of International Culture and Communication Studies at Waseda University, focused on CALL. He is working towards post-graduate degrees from SOAS -Alphawood, University of London and Waseda University, Faculty of Sports Science, Sport Culture. He teaches English at a prestigious Japanese national high school and at several Japanese colleges and universities. He also teaches the history of Japanese art and design. He is dourly concerned with the use of generative AI, its inherent biases, and threats posed by generative AI to language learning, speaker agency and autonomy, critical thinking skills, intellectual property rights, and concerned with the sterilization of world Englishes, and LLM propaganda bots attacking democratic institutions. When not worrying about those things, he loves spending time with his daughter. He also loves to do Aikido.
Citation
Wiegand, M. (2026). What’s wrong with this picture? Japanese students explore racial and gender bias in AI-generated images. In R. Dykes, O. Edwards, D. Bollen, & T. S. W. Lin (Eds.), Artificial intelligence in Japan’s language learning classrooms (pp. 114–144). Candlin & Mynard. https://doi.org/10.47908/45/5
Along with realistic text-based output from large language models for L2 teaching, AI-generated imagery (AIGI) has become increasingly complex, detailed, and realistic. However, there are ethical concerns about these technologies. In some cases, the technology is used in misinformation and disinformation campaigns that target people of all ages across realms ranging from politics to science and education. While earlier image models produced strange images that were easily identifiable as AIGI, recent technology has become increasingly photo-realistic - sometimes nearly indistinguishable from photography- thus making it harder to identify as AI. Furthermore, research has shown that AIGI often images that exhibit racial and gender biases, the implications of which are concerning due to their potential to reinforce racial and gender-based stereotypes. However, there has been little research into the ability of English as a Second Language (L2) students to identify and articulate the biases present.
This chapter introduces research exploring L2 students' recognition and discussion of bias in AIGI. Students generated images based on prompts discussed and agreed upon in class, then considered the output. Prompts were based on nationality, occupation, and lack of income. The study suggests that students with a closer knowledge of the images being produced were better able to identify bias in the imagery. The author offers further suggestions for addressing the issue of helping students recognize bias in AIGI, including integrating media literacy into second-language learning programs.
About the Contributor
Matthew Wiegand holds BAs in Cultural Anthropology from UC Santa Cruz and Global Japanese Studies from Meiji University, and an MA from the Graduate School of International Culture and Communication Studies at Waseda University, focused on CALL. He is working towards post-graduate degrees from SOAS -Alphawood, University of London and Waseda University, Faculty of Sports Science, Sport Culture. He teaches English at a prestigious Japanese national high school and at several Japanese colleges and universities. He also teaches the history of Japanese art and design. He is dourly concerned with the use of generative AI, its inherent biases, and threats posed by generative AI to language learning, speaker agency and autonomy, critical thinking skills, intellectual property rights, and concerned with the sterilization of world Englishes, and LLM propaganda bots attacking democratic institutions. When not worrying about those things, he loves spending time with his daughter. He also loves to do Aikido.
Citation
Wiegand, M. (2026). What’s wrong with this picture? Japanese students explore racial and gender bias in AI-generated images. In R. Dykes, O. Edwards, D. Bollen, & T. S. W. Lin (Eds.), Artificial intelligence in Japan’s language learning classrooms (pp. 114–144). Candlin & Mynard. https://doi.org/10.47908/45/5
Information About the Book
|
Title: Artificial Intelligence in Japan’s Language Learning Classroom
Editors: Robert Dykes, Oliver Edwards, Dave Bollen, and Tina Shu-wen Lin Publication date: June 2026 Read more... |