ETHICAL CHALLENGES OF AI-MEDIATED CULTURAL REPRESENTATION: ALGORITHMIC NARRATIVES, REFUGEE VOICES, AND PARTICIPATORY GOVERNANCE IN TUNISIA

The rapid proliferation of generative AI has fundamentally altered how cultural narratives are produced, circulated, and contested, raising urgent questions about representational authority, cultural ownership, and algorithmic accountability. Nowhere are these tensions more consequential than in the portrayal of forcibly displaced populations, where AI-generated content increasingly shapes public discourse and policy imaginaries. This study examines the ethical implications of AI-mediated cultural representation of Sub-Saharan refugees in Tunisia, a critical yet underexplored nexus of AI ethics, migration studies, and cultural mediation scholarship. Employing a qualitative normative-empirical design that integrates systematic discourse analysis of AI-generated migration content with semi-structured interviews involving refugees, civil society activists, and media professionals, the study pursues three objectives: (1) to identify representational patterns and ethical tensions embedded in AIgenerated migration discourse, (2) to explore how diverse stakeholders perceive and negotiate cultural ownership and narrative authority in algorithmic environments, and (3) to propose a culturally grounded ethical framework for responsible AI mediation. The findings reveal three intersecting dilemmas, narrative appropriation that displaces refugee agency, persistent stereotyping that reinforces reductive tropes, and algorithmic opacity that obscures the logics governing representational choices. Together, these dilemmas expose a critical accountability gap: AI systems produce culturally consequential narratives without meaningful input from the communities they represent. The study contributes to the growing interdisciplinary literature on responsible AI by advancing a participatory governance model that foregrounds refugee perspectives within digital knowledge systems and offers actionable guidelines for culturally responsible algorithmic representation in sensitive sociopolitical contexts.