Failure Modes for Large Language Models in Islamic Legal Reasoning

Muhammad Aurangzeb Ahmad

Keywords: fiqh, islamic legal reasoning, religious AI, pluralistic alignment, provenance grounding, homogenization, large language models, hallucination

TL;DR: Taxonomy of LLMs for Islamic legal guidance fails

Abstract: Millions of Muslims consult large language models (LLMs) for Islamic legal guidance, yet the specific failure modes of LLMs in this domain remain uncharacteristic. The Islamic jurisprudence fiqh is characterized by authenticated transmission chains isnad, codified schools of law, formal certainty grades qat./zann, and institutionalized scholarly authority. One can map open problems at the intersection of AI and Islamic law into open problems in LLM reliability, calibration, and alignment. We present a six-category failure mode taxonomy grounding each failure at the intersection of a classical jurisprudential concept and a specific ML failure class, derive evaluation criteria for each mode, and propose a research agenda connecting us. ul al-fiqh epistemology to ML methodology. We conclude that fiqh is a productive and underexplored test-bed for formal legal AI.

Track: Track 1: ML Research Addressing Challenges Faced by Muslim Communities

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Non Archival Confirmation: I understand that submissions to MusIML are non-archival and can be submitted to other venues.

Submission Number: 28