Generative AI, The Authority of Religious Knowledge, and The Reorientation of Islamic Religious Education Teachers: An Islamic Epistemological Perspective
Kata Kunci:
Generative AI, Authority of Religious Knowledge, Islamic Epistemology, Reorientation of Islamic Religious Education (PAI) Teachers, Digital TabayyunAbstrak
The acceleration of Generative Artificial Intelligence (Generative AI) based on Large Language Models (LLM) has shifted the search for religious knowledge from the turats and talaqqī traditions towards instant algorithmic synthesis. This study aims to critically-philosophically analyze the epistemic limits of religious output generated by AI, map shifts in religious authority, and formulate a pedagogical reorientation of Islamic Religious Education (IRE) Teachers. This qualitative research uses a philosophical - analytical research library with content analysis techniques and philosophical hermeneutics. Empirical Data on the performance of ChatGPT, Claude, and Gemini were extracted from IslamicLegalBench, FiqhQA, and QIAS 2025 Shared Task, then diagnosed using tripartite Islamic epistemology (Bayani, Burhani, Irfani). The results showed three layers of limitations that are structural, not technical. First, in Bayani’s reasoning, AI excels at compiling textual data and mapping fatwas, but is devoid of isnād, naqd al-isnād, and Shari'a responsibilities; the hallucination rate of 55% in some models and 21% in the best models, is not a statistical anomaly, but rather a manifestation of the absence of a living transmission relationship. Second, in Irfan's reasoning, AI experiences a total epistemic void: decreased performance in Arabic indicates AI is working through statistical patterns, not understanding the meaning of life, so it is not able to touch the tazkiyatun nafs, khasyyah, transmission, and barakah that require the presence of a conscious qalb and direct encounter (talaqqī). Third, on Burhani's reasoning, AI is able to construct comparative analyses of the 'illat law and sociology schools, but operates probabilistically without awareness of 'aql; sycophancy above 40% on 6 of the 9 models as well as the FiqhQA paradox GPT-4.0 excels in accuracy but Gemini is better at abstaining, proving the AI's output is only a crude hypothesis, not aql. Consequently, religious authority shifted from relational transmission to computational visibility giving rise to pseudo-authority. This study concluded that AI is based solely as a wasīlah (Ta'līm instrument), not a ghāyah or religious authority holder. Therefore, the role of the PAI teacher must be reoriented from mudarris to mu'addib and murabbi through Digital Tabayyun, planting Adab al-'Ilm, and moral example (uswah hasanah).
