Algorithmic representation of Iranian cultural values: A comparative semiotic analysis of AI image generation platforms

Background: In the age of artificial intelligence, AI image generation platforms have become active agents in cultural representation. This study examines how three core Iranian cultural values— belief in God, patriotism, and the idealized woman— are represented on platforms such as Gemini, ChatGPT, and Copilot. The main problem is the presence of cultural biases in training data and algorithmic architectures, which may lead to stereotypical, inaccurate, or Western-centric portrayals of non-Western cultural identities.Aims: This research aims to analyze the mechanisms through which these three fundamental values are represented in AI-generated images and to determine the extent of bias in these platforms compared to indigenous and religious models within Iranian society.Methodology: Employing a semiotic analysis based on Roland Barthes's five-code framework and Stuart Hall's theory of representation, this study qualitatively analyzed ۱۸ images generated by three platforms (Gemini, ChatGPT, and Copilot).Findings: The findings indicate that these platforms do not neutrally reflect cultural reality but actively participate in meaning construction and ideological representation. Three key findings emerged: (۱) Inaccurate and generalized representation of specific Shiite rituals, with a tendency toward generic Sunni Islamic imagery; (۲) A significant shift in the representation of patriotism, centering women as primary agents of national identity, in contrast to traditional masculine and military iconography; and (۳) A contradictory, algorithm-driven representation of the ideal woman, where her agency and success are defined not by traditional roles (such as motherhood) but by her ability to navigate the tension between tradition and modernity and conform to modern professional standards.Conclusion: AI systems function not as neutral tools but as powerful cultural actors. Algorithms reproduce algorithmic majoritarianism, whereby dominant cultural patterns in training data marginalize minority traditions (such as Shiism) and redefine values like womanhood based on modern, algorithmically preferred criteria. This research underscores the necessity of critically revising AI's representational politics and involving represented communities in the design of these systems to achieve fairer and more accurate cultural representations.

کلیدواژه ها:

algorithmic bias ، Artificial Intelligence ، Cultural representation ، gender representation ، patriotism representation ، Religious identity