Fiqh answers default to Salafi framing without asking madhhab

Category: Bugs Priority: P2 Severity: major Area: ai-answers Reports: 1 thread Status: Open

Summary

A user reports that when Qaf was asked a general fiqh question, the assistant presented a modern Salafi/hadith-centred application as the default answer — without first asking which madhhab the user follows, and without distinguishing that approach from the relied-upon (mu'tamad) positions of the four madhhabs. Only after being challenged did the assistant acknowledge the framing bias. The user warns this can confuse non-specialists and unintentionally move them away from their established madhhab without understanding the methodological shift, and that lay users should not be invited to pick between competing opinions based on what sounds persuasive. This is a methodology/trust issue for an Islamic answers product. Abdellatif replied that Qaf currently uses Shamela as the underlying corpus and that a madhhab can be specified via custom instructions in settings, acknowledging that this "is not very intuitive right now" — i.e., there is no first-class madhhab preference or answer labeling today.

What users say

pinkie.rasa1b@icloud.com — 2026-08-06, English, 19fd731248474335 — Abdellatif replied

Reports the default-framing bias on a general fiqh question and proposes concrete behavior: ask the user's madhhab first, give its relied-upon ruling simply, and clearly label comparative or modern independent conclusions. (Note: the email says "When Qatar was asked" — a typo/autocorrect for "Qaf".)

When Qatar was asked a general fiqh question, the assistant initially presented a modern Salafi/hadith-centred application as the default answer. It did not first ask which madhhab the user follows or distinguish that application from the relied-upon approaches of the four madhhabs. Only after being challenged did it acknowledge the framing bias.

And his requested fix:

For practical fiqh questions, the assistant should first ask the user's madhhab, provide its relied-upon ruling simply, and label comparative or modern independent conclusions clearly. It should not invite lay users to choose between competing opinions based merely on what sounds persuasive.

After Abdellatif's reply (Shamela corpus, madhhab via custom instructions in settings), the user responded warmly: "I see, I look forward to the future of the tool - it's very useful! Jazakallahu Khairan" (the reply arrived from d2018c@gmail.com, apparently the same person behind the iCloud relay address).

Details / repro

Threads