Source-Constrained Comparative Presentation: A Methodology for Employing Artificial Intelligence in Comparative Tafsir and An Experimental Application on Juz Amma

Abstract

This paper addresses the problem of employing artificial intelligence in comparative tafsir, amid the proliferation of digital Qur'anic applications and rising concern that the tool might usurp the authority to adjudicate and interpret, despite lacking any competence to determine Sharia rulings. The study aims to establish a methodological principle governing such use, to translate it into a practical mechanism, and to test it experimentally; it adopts a descriptive-analytical method in the theoretical grounding and discussion, and an applied-experimental method in building and testing the mechanism. The study proposes a principle it terms “source-constrained comparative presentation”: that the tool present the mufassirs' statements as verbatim, documented quotations and identify the points of agreement, variation, and contradiction among them, without issuing any ruling that substitutes for the judgment of the specialist scholar. On this principle a two-stage pipeline was built: an automated first stage that segments the tafsir text ayah by ayah, hashes it into digital fingerprints, and computes quantitative indicators (isnad density in al-Ṭabarī; juristic issues and variant readings in al-Qurṭubī; and the overlap of shared wording); and an analytical second stage, undertaken by the researcher, at the loci the triage flags as meriting attention. The methodology was applied experimentally to Juz ʿAmma (564 verses, 37 surahs), comparing al-Ṭabarī and al-Qurṭubī: fifty-nine comparative segments were analysed, distributed across twenty-eight loci of agreement (47%), thirty loci of variance-type difference (51%), and a single locus of contradiction (2%) in Sūrat al-Tīn. Agreement and variance together account for ninety-eight percent of the segments, showing that what appears to be disagreement between the two mufassirs is mostly a variance that enriches the meaning rather than contradicting it. The paper concludes by proposing a shift from exhaustive verbatim coverage to directed triage as the realistic path to scaling the method to the whole Qur'an.