Civics of Technology Announcements
Next Book Club: Hyperscale: The Ambition and Excess of Big Tech’s Data Empires by Paris Marx on Tuesday, December 15 at 8:00 PM Eastern. Register here.
New Podcast: Our Executive Directors joined the “What’s the Big Idea?” podcast for an episode titled, “‘Should we?’ A Conversation with Civics of Technology.” We discuss AI in Education hype, technoskepticism, and more. You can listen to the episode here or on podcast providers.
Call for Papers: Board member Dr. Stephanie Smith Budhai and colleagues are inviting abstracts for a TechTrends special issue titled "Critical Perspectives on Educational Technology in the Global South and Indigenous Communities." The issue will explore whose knowledge shapes educational technology with topics including algorithmic bias, data sovereignty, digital colonialism, Indigenous pedagogies, inclusive and multilingual design and more. At least one author must be from the Global South or Indigenous community being studied. Abstracts (400–600 words) are due January 15, 2027. Contact Stephanie at budhai@udel.edu with questions. Learn more and submit an abstract.
Dr. Matthew R. Deroo is an Associate Professor of TESOL and Bilingual education at Florida International University. His research interests include the social and cultural contexts of transnational immigrant youth, critical, digital media literacies, language education, and community engaged scholarship.
Recently, I published a new article in the Journal of Language, Identity, and Education Promoting Critical Perspectives of Language Learning Representations: Interrogating AI- Generated Images in a Pre-Service Teacher Education Course. I am happy to share the article with anyone who would like a copy (email: mderoo@fiu.edu) as this scholarship was not able to be published open access. The work featured in this article owes much of its conceptual framing to the Civics of Technology community, and I am very grateful for the opportunity to share key takeaways from this study with all of you. Specifically, this work is targeted toward teacher educators and others who work to support teachers’ professional development and learning.
In this piece, I reflect on an activity that I did in a pre-service teacher education course where I asked my students (who are preparing to become teachers) to use Artificial Intelligence image generators to compose an AI-generated image related to course ideas and concepts. The course was centered on how teachers can support bi/multilingual students’ language and content area learning in schools. I wanted my pre-service teachers to take up technoskeptical views (Pleasants et al., 2023 ) towards AI use, especially regarding the ways AI outputs depict bi/multilingual students, teachers, classrooms, and spaces of learning.
The learning task itself took place across three phases. First, pre-service teachers composed an AI-generated image from university supported models (i.e., Microsoft Copilot, Adobe Firefly). As they completed the image-generating process, I asked them to write down the various prompts or inputs they gave AI, and the number of iterations they went through to arrive at a desired output. Second, they brought their AI generated image to class to share with their peers. In that class session, we used a critical discussion protocol to support shared interrogation of the AI outputs. Finally, they reflected on their learning. In their final write ups, they traced their work from the initial image generation process to the in-class peer discussion, along with how they might reimagine their image using prompts from Kahn et al. (2022). They also noted how, if at all, they might use AI in their future teaching.
Pre-Service Teachers’ Responses to AI‑generated Outputs
My pre-service teachers found that while AI outputs could approximate images of students, teachers, classrooms, and other spaces of learning, the AI image models had trouble capturing language use, variation, and interaction among language users in meaningful ways. For example, the AI outputs produced images with features common to classroom settings or spaces of learning such as students seated at desks or tables, teachers and students looking over books, or students writing notes. However, even though the images depicted people interacting with one another in the classrooms, the models did not seem to authentically capture language among people engaging with one another.
My pre-service teachers expressed the disconnect they felt between their high-quality expectations about what language learning classrooms can and should look like with the low-quality AI outputs they received. For example, Nala (pseudonym) attempted to have her AI model show a dual-language bilingual classroom setting reflective of her field-based practicum experiences. However, this proved challenging. She stated, “I figured that technology would be advanced enough to create what I was imagining.” Yet Nala reported having to continuously change the prompts that she gave AI. After seven different attempts, Nala still had trouble getting an output that was truly reflective of bilingual spaces of learning. That is, the outputs she received were far less than what she hoped for or desired. AI has in many ways been promoted and advertised as having advanced or even super-natural capabilities, yet Nala’s experiences demonstrate the limitations of AI tools. For individuals who believe that AI can easily conjure or render what is in their minds, the shortcomings of AI often become quickly apparent. This is because AI is not sentient and there is no true transmission of “meaning” being taken up by the AI. Helping teachers to see this is an important part of the learning task.
By extension, the learning task further promoted criticality as pre-service teachers grappled with AI outputs that depicted students, teachers or classrooms in stereotypical or biased ways. Amanda (pseudonym) revealed how her prompts to show an ESL student repeatedly returned an Asian male learner. She critiqued how the outcome could be reflective of certain “stereotypes like the ‘model minority’ which affects how students are perceived in educational contexts.” She theorized that “expectations and assumptions that Asian students will excel academically may have been why the AI generated that specific image” suggesting the paradox between the AI inputs used to train image models and their outputs. Amanda was also troubled by the gendered representation in her AI output, sharing “most people believe that men are smarter than women, another baseless stereotype.” Beyond biases, other pre-service teachers noted how AI outputs could only show external representations. For example, Emily (pseudonym) reported wishing that her AI‑generated image went beyond “superficial visual features” to engage her senses and allow her “to hear and smell the classroom” which she believed would provide a lot more information about the space of learning beyond what could be seen.
Given the ways that AI is mediating and shaping our society and changing the ways that individuals experience the world, the outcomes of this study are important. In this study, pre-service teachers came to understand that while AI could take language from the prompts they gave their models to render outputs, the images that AI produced were not very accurate or reflective of what language teaching and learning is or should be. The learning task helped those pre-service teachers to better realize while AI can approximate teaching and learning, its outputs are incapable of fully capturing the deeply relational nature of teaching and learning where interactions are rooted in a shared humanity that a machine has trouble replicating. While each person must weigh the ethical and environmental costs associated with assigning learning tasks like the AI image generating activity with their own students, use of this task provided meaningful opportunities for critical interrogation of AI use and the advancement technoskeptical stances (Krutka et al., 2023) toward AI use. I hope and trust that others might modify aspects of my approach for use in their own contexts.
References
Deroo, M. R. (2026) Promoting critical perspectives of language learning representations: Interrogating AI generated images in a pre-service teacher education course. Journal of Language, Identity, and Education. 1-16.
Kahn, J. B., Peralta, L. M., Rubel, L. H., Lim, V. Y., Jiang, S., & Herbel-Eisenmann, B. (2022). Notice, wonder, feel, act, and reimagine as a path toward social justice in data science education. Educational Technology & Society, 25(4),80–92.
Krutka, D. G., Pleasants, J., & Nichols, T. P. (2023). Talking the technology talk. Phi Delta Kappan, 104(7), 42–46. https://doi.org/10.1177/00317217231168262
Pleasants, J., Krutka, D. G., & Nichols, T. P. (2023). What relationships do we want with technology? Toward technoskepticism in schools. Harvard Educational Review, 93(4), 486–515. https://doi.org/10.17763/1943-5045-93.4.486