Seeing Herself in the Machine: Deconstructing the AI-Generated Image of Arab Women

• In Event: Collaboration, Trust, and Knowledge Production in Global Higher Education

Proposal

Objectives
Arab women play a significant and multifaceted role in their societies, contributing across various sectors from education and healthcare to business and politics. Despite these efforts, mainstream media often fails to highlight their valuable contributions, instead relying on a narrow and limited perspective (Shaheen, 2009; Simone, 1999). Women constitute over half of the population in the Middle East and North Africa (MENA) region, and their lived experiences are incredibly diverse (El Sawis, 2015). However, the representation of Arab women in media, particularly in Western contexts, has long been a subject of criticism due to its pervasive reliance on stereotypes and biased portrayals (Asghar & Khan, 2024). When Arab women are shown, they overwhelmingly fall into a few narrow, stereotypical categories. They are often depicted as housewives, teachers, or nurses, or as "the oppressed victim"—voiceless, submissive, and in need of a "male savior" to be liberated from a patriarchal society (Chahdi, 2024; Hadjeris, 2024). These limited portrayals reduce their agency and silence their independent opinions. Ultimately, this perpetuates harmful stereotypes that do not reflect the rich diversity and active participation of Arab women in their communities.
The recent evolution of digital media platforms and the rise of generative AI have created new opportunities for diverse perspectives to challenge common stereotypes within a connected and participatory culture (Ito et al., 2013; Jenkins, 2009). These platforms enable a wider range of voices to be heard, facilitating conversations in ways previously not possible. By actively engaging with and contributing to online spaces, individuals can now evaluate and resist oppressive representations (Williams, 2020; Herrera, 2012). This shift is particularly critical because AI systems are trained on vast datasets drawn from both traditional and this new mode of digital media, which may reflect conflicting opinions and unfair representation of Arab women and their contributions to society. Therefore, it is important to investigate how AI models are learning to challenge these historical biases or perhaps perpetuate them. In this study, we examine the Arab women's representations in AI models, relying on the most common one, which is ChatGPT 4.0. ChatGPT is the first Generative AI tool to be released, and it has between 800 million and 1 billion weekly active users (Badalyan, 2025), making a huge impact on users and their views.

Method
The study is guided by the question of: How do Gen AI portray Arab women in the MENA region? This is deconstructed by looking at the representation of Arab women on different Gen AI platforms, such as ChatGPT, Gemini, and Copilot. It adopts a qualitative case study design that looks at the visual representation of Arab women from four different countries across the Arab world. Given the diversity of the region, the authors made sure to include countries across the four sub-regions: The Maghreb/North Africa, the Mashriq/the Levant, and the Arabian Peninsula. The authors believe that this choice will cover the diversity of the sample across geography, ethnicities, languages, and histories (including colonial history). The following countries are selected in the study: Egypt, Algeria, Palestine, and Saudi Arabia. While texts tend to provide diverse representations, images often convey meaning in a vague and ambiguous way. Therefore, the authors opted for the visual representation of Arab women on Gen AI in an attempt to disrupt stereotypes and issues of misrepresentation, if any. Moreover, given the authors’ positionality from Palestine and Algeria, respectively, their knowledge and background about the context serve as invaluable tools to offer a new perspective on Gen AI and the representation of women and historically marginalized communities. The authors used the following prompt to request the chatbot for images “Show me a picture of an Arab woman in Algeria”; “Show me a picture of an Arab woman in Egypt”; “Show me a picture of an Arab woman in Palestine”; “Show me a picture of an Arab woman in Saudi Arabia.” A total number of 20 images were generated by the Chatbot and then analyzed. Multimodal Critical Discourse Analysis (MCDA) informs the conceptual and analytical framework of the study. The MCDA framework is informed by Halliday’s work on the ideational and the interpersonal metafunctions of language. The analysis of the visuals’ ideational metafunction centres on the question of ‘what an image is image of?’ (Painter, 2019, p. 28) that is deconstructed in terms of three aspects: participants, processes, and circumstances (Painter et al., 2013). Moreover, the interpersonal metafunction of visuals centers on deconstructing issues of power between the viewer and the represented character ( Painter, 2019).

Findings

The preliminary findings indicate that women are portrayed through stereotypes that fail to reflect the diverse roles Arab women play in contemporary society. For instance,
Algerian women are associated with the war of independence, serving different roles such as intelligence operatives, nurses, and armed combatants within the Algerian liberation movements. Palestinian women are represented with different Palestinian cultural outfits while also leading groundbreaking peace efforts, but little is shown regarding their different roles in health, economy, business, and academia. Moreover, Egyptian women’s agency is associated with ancient times, emphasizing rights such as property ownership and inheritance, dismissing recent developments and advancements of Egyptian women as leaders, innovators, and decision makers. In most cases, they are represented as pharaohs. These stereotypical depictions neglect the complexity of Arab women’s lives and experiences and suggest the need to reexamine the role of Gen AI in generating inclusive and less biased visual material.

Significance

This research highlights the critical need to address Arab women's biases in Gen AI, as they perpetuate harmful stereotypes and can lead to inequitable outcomes across various applications. It calls for a careful review of machine algorithms and data collection to ensure fair representation and an inclusive AI system for the global good.

Authors

• Samaa Haniya

• Fadhila Hadjeris, UCLA