How culturally aware is AI? Students explore bias and representation through AI‑generated Barbies and reflect on ethical and responsible use of AI.
- Individual image creation task
- Group discussion
- In class or online
- AI image generation and critical evaluation of bias
- All disciplines
- Basic
- approx. 70 min / 1-2 session
- 15-40 / small groups of 3-5
- GenAI image tools
- Flexible classroom setting
- Shared platform
Short description
In this activity, students are tasked with using an AI tool to generate a Barbie or an action figure image that represents their culture. This can relate to nationality, ethnicity, locality or particular interests or activities that they identify with and know well. They then evaluate the quality and accuracy of the AI-generated representation, noting any potential bias, and discuss this in small groups. The images and prompts are compiled in a shared platform and key points emerging from the small group discussions are shared in the wider group. The activity concludes by connecting these insights to broader questions of evaluating AI‑generated outputs in academic contexts, with particular attention to bias, representation, and limitations of generative AI.
Competence domain of the Didactic Framework: Foundational AI Knowledge
By the end of this activity, students can…
- discuss strengths and limitations of GenAI. (FLAIR Didactic Framework: LO4)
- define bias, quality of output, representation, fairness and explain how they are related to common AI limitations. (FLAIR Didactic Framework: LO5)
- develop strategies for evaluating the quality and appropriateness of AI‑generated outputs. (FLAIR Didactic Framework: LO6)
- recognise common GenAI failures, including biased or inaccurate representations. (FLAIR Didactic Framework: LO7)
Instructions
Provide context for the activity by introducing the AI-generated Barbie trend and highlighting examples from a (since-deleted) BuzzFeed article on AI Barbies of the World that caused controversy as a result of evident stereotypes in the outputs (see References section).
Briefly introduce the generative AI tool that will be used for image creation and the platform for sharing results (e.g. Padlet can be used for both image creation and to collect/share/comment on responses).
Ask students to use a GenAI tool to produce a Barbie or action figure image that represents their culture (i.e, nationality, ethnicity, local area, activity or interests they identify with). Students should record their prompt(s) and may iterate to see how changes affect the output. All prompts and generated images are saved for later sharing with the class.
The class is divided into small groups of 3-5 students to share their results and discuss the quality and accuracy of their outputs, paying particular attention to bias, stereotypes, and omissions in the AI‑generated images.
Students upload their AI-generated image(s) along with prompt(s) used and key points of evaluation to a shared class platform (i.e., Padlet, discussion board in LMS, etc.) so that the full class can review. The collection of examples serves as a common reference point for the full class discussion.
Bring the full class back together and have all examples from the shared platform up on the screen to scroll through. Invite groups to share anything that particularly stood out to them and use these observations to elicit a plenary discussion about what this means in terms of evaluating other forms of AI-generated content in academic contexts. Guide the discussion toward identifying common limitations of generative AI, such as bias, stereotyping, inaccuracies, or lack of contextual understanding. Where appropriate, broaden the conversation to other related ethical considerations, including environmental costs and data‑privacy risks associated with some AI image‑generation trends (e.g. AI doll trend).
Sum up the class discussion and provide a brief explanation of training data for GenAI models and limitations of GenAI outputs, including bias and hallucination. Finally, discuss critical appraisal and share necessary steps for evaluating GenAI outputs, including assessing the accuracy, objectivity, and relevance of the output. Conclude with an invitation for students to reflect on the broader implications of unchecked bias in GenAI outputs.
Assessment
This activity is best suited for formative assessment. Learning can be evidenced through students’ participation in discussion and their ability to critically evaluate AI‑generated outputs with regard to bias, representation, and quality.
As a follow‑up, students may be invited to submit a short reflective contribution (e.g. a post in the course LMS forum) responding to prompts such as the broader implications of unchecked bias in generative AI outputs. Such reflections can be used to assess depth of understanding, ethical awareness, and students’ ability to connect the activity to academic and societal contexts.
Possible challenges
- Some students may feel uncomfortable or object to using generative AI for ethical or personal reasons.
- AI‑generated cultural representations may be experienced as stereotypical, biased, or upsetting, particularly when they relate to students’ own backgrounds.
- Discussions may focus narrowly on a small set of obvious stereotypes, limiting the depth of analysis.
How to adress them
- Ensure that students who choose not to generate their own AI output can still participate fully by evaluating and discussing others’ examples.
- Address the possibility of biased or uncomfortable representations at the start of the activity and emphasise that the purpose is critical analysis, not endorsement. Allow students to shift to less personal topics if needed (see Variants section).
- During small‑group or plenary discussions, use a loose guiding framework (e.g. representation, accuracy, omissions, stereotypes, and quality) to broaden the range of features students consider without prescribing a fixed checklist.
The Barbie/action figure topic was chosen for this activity as it is internationally recognisable and often associated with stereotype, making it straightforward to evaluate for all groups. However, this activity can also be adapted to refer to other common stereotypes or to course-specific topics by changing the focus of the AI-generated output to a topic covered in class (requiring specific content knowledge/expertise). In this case, the teacher should set a discipline-specific research question, topic, or task relevant to the current level of the students and ask them to evaluate that output.
Cheung, J. (2023). “How AI Image Generators Make Bias Worse”. The London Interdisciplinary School. 21 August. Available at: https://www.lis.ac.uk/stories/how-ai-image-generators-make-bias-worse
Hellmann, O. (2025). “Historical images made with AI recycle colonial stereotypes and bias – new research”. The Conversation. 23 October. Available at: https://theconversation.com/historical-images-made-with-ai-recycle-colonial-stereotypes-and-bias-new-research-268070
Koh, R. (2023). “A list of AI-generated Barbies from ‘every country’ gets blasted on Twitter for blatant racism and endless cultural inaccuracies”. Business Insider. 11 July. Available at: https://www.businessinsider.com/ai-generated-barbie-every-country-criticism-internet-midjourney-racism-2023-7
Lever, T. and Sradarov, S. (2025). “’Australiana’ images made by AI are racist and full of tired cliches, new study shows”. The Conversation. 14 August. Available at: https://theconversation.com/australiana-images-made-by-ai-are-racist-and-full-of-tired-cliches-new-study-shows-263117
McMahone, L. and Rahman-Jones, I. (2025). “Everyone’s jumping on the AI doll trend – but what are the concerns?”. BBC. 12 April. Available at: https://www.bbc.com/news/articles/c5yg690e9eno
Messingschlager, T. V., & Appel, M. (2026). Algorithmic bias in image-generating artificial intelligence: prevalence and user perceptions. Information, Communication & Society, 29(5), 1656–1678. https://doi.org/10.1080/1369118X.2025.2584146
Using this resource
This resource is licensed under Creative Commons BY-NC-SA 4.0 license. Suggested citation: Flair Collaboration. (2025). FLAIR Toolkit. Teaching GenAI Competencies.
