Auteurs : Ben Hammouda Saleh Al-mahdi .
This study advances an original theoretical and empirical argument concerning the application of artificial intelligence (AI) to the corpus of Islamic manuscripts. It posits that prevailing AI digitization frameworks constitute an "algorithmic veil" — a technically sophisticated but epistemologically incomplete layer of mediation that selectively renders accessible those manuscript traditions whose script types are well-resourced in computational training data, while rendering effectively invisible the traditions whose scripts are not. Through a systematic literature review (n = 91 studies, 2010–2024), a comparative analysis of OCR/HTR performance across seven Arabic script types on four major platforms, a structured survey of 87 active digitization projects (2015–2023), and direct field engagement with the Kairouan Manuscript Laboratory, this study demonstrates three original findings. First, there exists a statistically significant inverse correlation between the historical importance of a manuscript repository and the degree of its AI integration, a pattern designated here as the "capability inversion paradox." Second, current AI systems achieve character accuracy rates of 94.7% on standard Naskh but collapse to 14.3–43.8% on Maghribi, Kufic, and Andalusi scripts — precisely the scripts that dominate the oldest and most historically significant North African and Andalusian collections. Third, the dominant paradigm of AI adoption in Islamic manuscript studies systematically undervalues what this study theorizes as "civilizational epistemology" — the ensemble of philological, jurisprudential, historical, and aesthetic competencies required for the full comprehension of Islamic manuscript materials. In response, the study proposes the Philological-Technological Interface (PTI) Framework, an original model that reconceptualizes the relationship between AI and Islamic philology as structurally complementary rather than substitutive. The framework is operationalized through seven concrete implementation protocols for the Kairouan collection and analogous North African repositories, offering a transferable model for AI integration that is both technically rigorous and civilizationally responsible.
Algorithmic Veil; Islamic Manuscripts; Epistemic Asymmetry; Kairouan; Civilizational Epistemology; PTI Framework; Script Diversity; Digital Humanities