LLM_Kazakh_Philosophy_Dataset_Final.docx

This supplementary dataset presents a comparative multilingual analysis of ChatGPT-generated responses to texts from Kazakh ethical and philosophical traditions. The materials were produced under controlled prompting conditions using the publicly accessible ChatGPT interface. The primary multilingual interaction sessions corresponding to the main thirteen analytical cases were conducted during February 2026. Additional repeated-prompt robustness analyses and comparative control sessions introduced during the revision stage were conducted in May 2026. All prompts were identical across English-, Russian-, and Kazakh-language sessions. No system prompts, role prompts, or persona-based instructions were used. The dataset is organised into analytical cases (Texts I–XIII). Each case includes: Original source text ChatGPT-generated responses (English / Russian / Kazakh) Comparative analytical table Analytical findings and interpretive observations The revised supplementary materials additionally include repeated-prompt robustness analyses for selected cases, as well as a comparative Kantian control case introduced in response to reviewer comments concerning interpretive stability and multilingual consistency. The dataset is intended to support methodological transparency and comparative interpretive analysis within the study “AI and Cultural Asymmetry in Generative AI Systems through the Lens of Kazakh Ethical Philosophy.” Note on Terminology and Multilingual Materials The primary analytical cases included in this supplementary dataset were originally produced during February 2026 prior to the peer-review revision stage. Consequently, some analytical tables and findings retain the terminology employed during the initial phase of the study, including references to “LLMs,” “ontological asymmetry,” and related conceptual formulations. In the revised manuscript, terminology has been refined in response to reviewer comments in order to distinguish more precisely between foundation models, publicly accessible ChatGPT application environments, and generative AI systems more broadly. These later terminological clarifications do not alter the underlying empirical observations presented in the original analytical materials. The dataset intentionally preserves multilingual analytical materials in English, Russian, and Kazakh. This includes original source texts, prompts, ChatGPT-generated responses, comparative tables, and portions of the analytical findings. Since the study examines multilingual interpretive behaviour and cultural-semantic variation across languages, the preservation of original-language materials constitutes an essential methodological component of the research design rather than a translation inconsistency.