نوع مقاله : مقاله علمی پژوهشی
نویسندگان
1 پژوهشگر همکار، گروه فلسفه دین، پژوهشکده بینالمللی امام رضا (ع)، جامعه المصطفی العالمیه، مشهد، ایران.
2 پژوهشگر مدعو، گروه فلسفه دین، پژوهشکده بینالمللی امام رضا (ع)، جامعه المصطفی العالمیه، مشهد، ایران.
چکیده
کلیدواژه‌ها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Although the discussion of AI errors is a frequent topic in computer science, the analysis of its consequences for the distortion of religious concepts has yet to receive serious attention. Aiming to address this gap, the present article offers a theoretical framework for identifying the roots of and managing AI errors in this domain. According to this framework, such errors are the product of the interaction of three interrelated factors: (1) bias in data: the structural marginalization of Shīʿī sources in training data; (2) bias in algorithms: technical deficiencies and value‑orientations of model designers; and (3) the complexity of religious texts: the resistance of metaphorical language and multi‑source textual structures to mechanical comprehension. In line with this framework, the article emphasizes the necessity of employing a systemic set of technical and strategic solutions, including the creation of benchmark corpora and knowledge graphs at the data layer; the use of fine‑tuning, retrieval‑augmented generation, and human feedback at the algorithmic layer; and a paradigm shift toward the “illuminative tool” at the textual layer. Finally, by distinguishing functions into three categories—content‑based (high‑risk), formal (safe), and dual (conditional)—the article demonstrates that stringent technical solutions are reserved for the content‑based domain, whereas in other domains, focusing on enhancing digital literacy and skills such as prompt engineering suffices.
کلیدواژه‌ها [English]
)