Who Answers for the Fatwā? Machine Mediation and the Structure of Religious Advice

Abstract

The institution of iftāʾ is, structurally, one of the most carefully built answerability mechanisms any civilisation has produced. A questioner brings a situation. A named scholar receives it, asks back, establishes the facts, characterises the case, locates a ruling, weighs the questioner's circumstances, issues an answer under their own name with reasoning attached, remains available to be asked about it afterwards, and leaves it non-binding so that the questioner retains their own share of the decision. Every one of those features is a design choice, and several exist for no reason other than to keep the answer attached to a person who can be asked why. This article argues that machine mediation of religious guidance removes precisely those features and leaves the others. It decomposes the advisory act into eight components — the Muftī Function — and argues that current systems perform the middle components moderately, the first two poorly, and the last three not at all. The last three are attribution, continuing availability, and non-bindingness, and they are the components the tradition built most deliberately. Four constructs are developed. Question Fidelity is the degree to which an answer addresses the actual questioner's actual situation rather than a flattened version of it, and Istiftāʾ Flattening is what happens to a situation on its way to becoming a query. Answer Detachment is the separation of a ruling from the conditions under which it was issued — a problem the tradition has always known and has always treated as a defect, which machine mediation converts from an exception into the default. Authority Without Attribution names the resulting condition: an answer that carries weight with no party behind it, occupying a category the tradition never produced and therefore never named. The article is structural throughout. It makes no theological claim, issues no ruling, takes no position on any matter of substance, and expresses no view on any school, scholar, institution, or service. No scriptural text is reproduced and no real case appears. Its subject is institutional design: what an advisory institution is for, which of its features are load-bearing, and what happens when a technology supplies the answer while removing the structure that made an answer safe to give. The second half is practical and draws on the tradition's own safeguards. A decision grid sets out what a system should do by question type, from answering freely on settled general matters to routing personal situations on contested questions to a named scholar. An eight-test battery lets an institution establish in an afternoon how its own service behaves, with each test written so that a failure is recognisable without adjudication. A remediation ladder orders seven interventions by cost, beginning with one that is free and almost always done wrongly: placing the disclaimer before the answer rather than after it. A table of actions by role covers institutions, scholars, developers, questioners, funders, and researchers. A worked deployment scenario shows an institution with genuine scholarly standing building a service from its own scholars' answers, making eight individually defensible decisions, and removing six of the eight components — while observing a service that appears to work, because no error signal and no complaint signal exist. It also shows the highest-value and easiest change available: a service built on a corpus of named scholars' answers already holds attribution in its own data and is discarding it at presentation. Paper 8 of 10 in The Answerability Series. Manuscript ID IFTA-MEDIA-2026-08. 40 pages, 9 figures, 22 tables, five appendices including institutional worksheets, a component reference card, and a worked scenario — all released under CC BY 4.0. A framework contribution citing no external sources and stating its refutation conditions.