EUVIMEDEUVIMED entdecken
Crossref

A GOVERNANCE-AWARE AI PROTOTYPE FOR SHARIAH SCREENING IN ISLAMIC DIGITAL FINANCE: DESIGN, TESTING AND COMPARATIVE ABLATION

Mohd Nor Omar, Bakhtiar Affendi Rosdi

iBAF e-Proceedings · 29.10.2026 · 12 (1) · 473-492

DOI: 10.33102/qy5kx603 ↗

Kurz gefasst

Fachartikel in der Zeitschrift iBAF e-Proceedings, online erschienen am 29.10.2026. Verfasst von Mohd Nor Omar und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Artificial intelligence is already part of digital finance, yet Shariah-sensitive use raises a distinct governance problem. The relevant question is whether an AI-assisted system can help reviewers identify material issues while keeping juristic and institutional authority with qualified humans. This paper reports the design, initial functional demonstration and comparative ablation of a governance-aware prototype for preliminary Shariah screening in Islamic digital finance. Developed through Design Science Research, the artifact combines structured intake, issue mapping, source-category discipline, preliminary risk-state classification, audit records and human escalation. It supports early review and records uncertainty; it does not issue fatwa or certify products. A grey-box reporting approach discloses the design rationale, public architecture, testing protocol and aggregate results while withholding hidden prompts, proprietary orchestration rules, source-weighting logic and deployment details. In Core Run 1, Claude Fable 5 (claude-fable-5) matched the expected preliminary risk state in all twelve pre-specified scenarios and obtained 108 of 108 functional points. The same model, tested without the Hybrid Shariah Governance (HSG) layer, scored 78 of 108 points (72.2 percent). Issue recognition remained high at 91.7 percent, but the plain model produced no bounded risk-state taxonomy, stated its non-fatwa status in only 41.7 percent of outputs and often escalated generically. Because several rubric items reward behaviours the HSG layer is instructed to produce, the gap indicates that governance behaviour must be engineered rather than assumed; it is not a measure of the model's juristic capability. Because the reported runs were single-session, single-round and initially self-coded, the result establishes protocol workability and initial governance discipline only. Expert-validated Shariah correctness remains outside the claim of this conference paper.

Zitieren

APA

Omar, M. N., & Rosdi, B. A. (2026). A GOVERNANCE-AWARE AI PROTOTYPE FOR SHARIAH SCREENING IN ISLAMIC DIGITAL FINANCE: DESIGN, TESTING AND COMPARATIVE ABLATION. iBAF e-Proceedings, 12(1), 473-492. https://doi.org/10.33102/qy5kx603

Vancouver

Omar MN, Rosdi BA. A GOVERNANCE-AWARE AI PROTOTYPE FOR SHARIAH SCREENING IN ISLAMIC DIGITAL FINANCE: DESIGN, TESTING AND COMPARATIVE ABLATION. iBAF e-Proceedings. 2026;12(1):473-492. doi:10.33102/qy5kx603

BibTeX
@article{Omar2026,
  author = {Omar, Mohd Nor and Rosdi, Bakhtiar Affendi},
  title = {A GOVERNANCE-AWARE AI PROTOTYPE FOR SHARIAH SCREENING IN ISLAMIC DIGITAL FINANCE: DESIGN, TESTING AND COMPARATIVE ABLATION},
  journal = {iBAF e-Proceedings},
  year = {2026},
  volume = {12},
  number = {1},
  pages = {473--492},
  doi = {10.33102/qy5kx603}
}
Kostenlos starten →