Evaluating Cultural Representation and Epistemic Bias in AI Generated Multimodal English Teaching Materials

Abstract

This paper assesses the cultural presentation and epistemic bias in AI generated multimodal English teaching resources. The paper states that the prevalence of Western cultural norms and epistemologies is high, and AI-generated content mostly represents the Western views and does not represent other cultures. These prejudices deprive students of the opportunity to see the world from different perspectives all over the world, which supports stereotypes and creates a very one-sided perception of the world. The study uses a mixed-methodology, as it examines 500 pieces of AI-generated educational content, such as interactive lessons, multimedia materials, and grammar, created by many different AI-driven systems. The results have indicated that although few materials contain global examples, most tend to present non-western cultures in a westernized way and reduce their authenticity. Epistemic bias also manifests in the same way when the knowledge in these resources is highly concentrated on Western scientific and education models and other forms of epistemology, including Indigenous knowledge systems and the Eastern philosophies are barely depicted. The imbalance has major effects on educational equity because students might not get a balanced, inclusive education that will be representative of the diversity of the global society. The paper suggests that AI developers should incorporate different cultural frameworks and knowledge systems into training data to overcome these problems so that the AI-created materials will be more representative. Moreover, the studies carried out in the future must consider the creation of AI systems to encourage cultural and epistemic diversity in education material. Longitudinal research would also be useful to understand the effect of these AI-generated materials in the long term on the cultural and epistemological perspective of students and, ultimately, make the learning process more inclusive and equitable.