RETELLING ALGERIA BY ALGORITHM:

Abstract

Generative AI is increasingly becoming a mediator through which cultural narratives are retold, adapted, and circulated, raising questions about how culturally specific forms of meaning survive algorithmic reconstruction. This study examines AI-generated retellings of three Algerian folktales representing distinct narrative traditions: Lalla Loundja, Bechkerker et l'Ogre, and Uccen d Insi. The tales are approached as culturally embedded narratives whose distinctive meanings emerge through different configurations of female agency, grotesque survival, animal symbolism, and pragmatic morality. Using qualitative comparative textual analysis, documented versions of the three tales are compared with retellings generated by ChatGPT through a deliberately minimal prompting protocol. The analysis examines transformations in narrative structure, character agency, moral logic, humor, symbolism, cultural attribution, and oral storytelling features. The findings reveal four recurring forms of transformation: narrative domestication, in which culturally specific conflicts are reshaped according to familiar narrative conventions; narrative compression, in which episodic and cumulative oral structures are substantially reduced; narrative substitution, in which a documented tale is replaced by a substantially different narrative; and cultural misattribution, in which regional identity is inaccurately assigned. Taken together, these transformations suggest that generative AI tends to privilege familiar plot structures, coherent moral frameworks, and accessible narrative conventions over culturally specific forms of storytelling. The study argues that AI-generated retellings constitute acts of cultural mediation rather than neutral reproduction, contributing to a process of algorithmic narrative normalization through which underrepresented oral traditions may be reshaped as they enter digital circulation.