Ethics and Epistemology of Generative AI in Local-Wisdom-Based Religious Moderation Research: Reflections from the JOGLO Model

Abstract

Generative artificial intelligence (AI) is increasingly used in religious studies research, yet its ethical and epistemological risks are rarely examined from users’ direct experience, even though language models can produce fluent text with no grounding in real data. This study aims to document the benefits and the problematic incidents of using generative AI in developing JOGLO (Jagongan Slow), a participatory dialogue model for religious moderation in Tegal Regency, Indonesia, built on Javanese local wisdom and the song Tegal Toleransi, and to formulate an ethical protocol for AI use in Islamic studies research. The study employed an autoethnographic approach supported by document analysis of the author’s interaction trail with AI assistants across the project. The results show that generative AI accelerated literature synthesis, conceptual framing, instrument design, and form automation, but also produced two patterns of problematic incident: the wholesale fabrication of findings, including statistics and quotations, for a forum that had never been held, and the presentation of social media reach as evidence of impact. Both were detected through manual cross-verification and simple conceptual questioning rather than automated checks. Read through maqāṣid al-sharī‘ah, particularly ḥifẓ al-‘aql and scholarly trustworthiness as part of ḥifẓ al-dīn, these incidents threaten both sound reasoning and religious integrity. The study concludes with an eight-principle protocol, including mandatory human verification, a prohibition on writing findings before data exist, the separation of reach from impact, and transparent disclosure of AI involvement, offered as a reflective framework rather than a generalizable claim.