Argumentation Layer

Distillary doesn’t just capture what authors claim — it captures why they believe it, how strong the evidence is, and what they argued against.

The Problem

A claim like “validated learning measures progress” is an assertion. But the author didn’t just assert it — they backed it with case studies, data, and logical argument. Without capturing that structure, you can’t answer “what’s the evidence?” and you can’t compare argument quality across sources.

The Solution: backing: Field

Every atom claim in the brain can have a backing: field in its frontmatter. Each entry captures one piece of evidence:

backing:
  - category: textual
    subtype: ayah
    ref: "النساء:115"
    snippet: "ومن يشاقق الرسول من بعد ما تبيّن له الهدى..."
    strength: definitive
    warrant: "الآية تحذر من اتباع غير سبيل المؤمنين مما يثبت حجية الإجماع"
  - category: transmitted
    subtype: hadith_sahih
    ref: "سنن ابن ماجه"
    snippet: "لا تجتمع أمتي على ضلالة"
    strength: strong
    warrant: "عصمة الأمة من الاجتماع على الخطأ تثبت أن إجماعها حجة"

The 5 Fields

category

One of 9 universal types that cover every domain:

CategoryWhat it meansIslamic exampleAcademic example
textualCitation from authoritative textآية قرآنيةPrimary source quote
transmittedReport through chain of peopleحديث نبويReported data
consensusCollective agreement of expertsإجماع العلماءScientific consensus
analogicalExtension from known to unknown caseقياس شرعيComparative study
empiricalDirect observation or measurementاستقراءExperiment, RCT
rationalLogical deduction or inductionدليل عقليFormal proof
experientialFirst-hand lived experienceCase study
authorityStatement from recognized expertأثر صحابيExpert opinion
silenceAbsence of evidence IS the evidenceلم يرد نصNo studies exist

These 9 categories are exhaustive and domain-agnostic. The same framework works for Islamic jurisprudence, academic research, legal texts, philosophy, and business books.

subtype

The domain-specific label. This is freeform — it captures the vocabulary natural to each field:

  • Islamic: ayah, hadith_sahih, hadith_hasan, hadith_daif, ijma, qiyas, athar
  • Academic: rct, meta_analysis, cohort_study, case_study
  • Legal: statute, binding_precedent, legislative_history
  • Business: anecdote, market_data, case_study

ref

Citation reference. Where to find the evidence:

  • "سورة النساء:115" or "صحيح البخاري 1117" or "Smith et al., 2020"

snippet

The first ~15 words of the actual evidence text. Enough to identify it:

  • "ومن يشاقق الرسول من بعد ما تبيّن له الهدى ويتبع غير سبيل المؤمنين..."

strength

How strong this evidence is, on a universal 5-level scale:

StrengthMeaningExamples
definitiveVirtually undisputedآية صريحة، حديث متواتر، meta-analysis
strongWidely accepted, minor debateحديث صحيح، إجماع، well-powered RCT
moderateReasonable but contestedحديث حسن، cohort study، strong analogy
weakSome value but unreliable aloneحديث ضعيف، anecdote، expert opinion
contestedActively disputedConflicting studies, disputed hadith

This is the author’s assessment (or the field’s standard), not ours.

warrant

The most important field. A single sentence answering: why does this evidence support THIS claim?

The same hadith can support completely different claims depending on the warrant:

Evidence: "لا تجتمع أمتي على ضلالة"

Warrant A: "عصمة الأمة تثبت أن إجماعها حجة"
  → Claim: الإجماع حجة ملزمة

Warrant B: "إذا كانت الأمة معصومة حين تجتمع، فتفرّقها عن الحق يسقط هذه العصمة"
  → Claim: ترك الإنكار يسقط الحماية الإلهية

Without the warrant, you can’t understand HOW the author connects evidence to conclusion.

Evidence Chains

Sometimes evidence builds on prior evidence. A hadith specifies a general ayah, then ijma confirms the understanding, then qiyas extends it. This isn’t 4 independent backings — it’s a chain.

Use depends_on to represent chains:

backing:
  - category: textual           # step 1
    subtype: ayah
    ref: "النساء:11"
    strength: definitive
    warrant: "النص يثبت أصل الميراث"
 
  - category: transmitted        # step 2 — builds on step 1
    subtype: hadith_sahih
    ref: "صحيح البخاري"
    strength: strong
    warrant: "الحديث يخصّص عموم الآية"
    depends_on: 0               # depends on backing[0]

Counter-Arguments (Rebuttals)

Authors — especially in Islamic jurisprudence — argue dialogically. “فإن قال قائل…” is half of al-Risala. Distillary captures these as separate claims:

tags:
  - role/rebuttal
  - rebuttal/defeated
rebuts: "[[الإجماع حجة ملزمة]]"

Tags:

  • role/rebuttal — this is a counter-argument
  • rebuttal/defeated — the author answered it successfully
  • rebuttal/acknowledged — the author concedes partially

Silence as Evidence

“No text addresses this” is itself evidence in some domains:

backing:
  - category: silence
    subtype: no_text
    scope: "لم يرد نص من القرآن أو السنة في هذه المسألة"
    strength: moderate
    warrant: "انعدام النص يبيح الاجتهاد بالقياس"

How the Extract Agent Captures This

The extract agent (v4.0) automatically identifies evidence patterns in text:

Text patternCategorySubtype
”قال الله تعالى”textualayah
”قال رسول الله” / “روى”transmittedhadith
”أجمع العلماء” / “لا خلاف”consensusijma
”قياساً على” / “بجامع العلة”analogicalqiyas
”قال ابن عباس” / “قال مالك”authorityathar
”studies show” / “N%“empiricalstudy
”لأن” / “therefore”rationalargument
”فإن قال قائل”→ separate rebuttal claim

Claims with no identifiable evidence get no backing: field — that’s fine. Not every claim needs formal evidence.

Why 9 Universal Categories Instead of Domain-Specific Profiles

Every domain has its own evidence vocabulary (ayah vs. statute vs. RCT), but there are only 9 ways humans argue. The universal category enables cross-domain queries and comparison, while the freeform subtype preserves domain-specific precision.

AspectWhat it enables
category”Show all claims backed by textual evidence” — works across ALL sources
subtype”Show all claims backed by hadith sahih” — domain-specific filter
strength”Show only definitive claims” — universal quality filter
warrant”How does Source A use this evidence vs Source B?” — cross-source comparison

Shared Evidence Hubs

When the same evidence (same hadith, same verse) is cited by multiple sources, it becomes a shared evidence hub. Bridge concepts in brain/shared/concepts/ already capture conceptual overlap — shared evidence adds a deeper dimension because sources share the same textual foundation.

What This Enables

QueryHow
”ما الدليل على حجية الإجماع؟“Find claims → return backing entries
”Show only definitive claims”Filter by strength/definitive
”Same evidence, different conclusions?”Shared evidence hubs with warrant comparison
”How does Source A argue vs Source B?”Compare backing category distributions
”What claims have no evidence?”Claims without backing: field

What It Looks Like in Practice

A typical source produces:

  • 60-80% of claims with backing (the rest are bare assertions)
  • 2-5 backing categories per source (most sources don’t use all 9)
  • 100% of backings have warrants (the extract agent always captures the reasoning)
  • A mix of definitive, strong, and moderate evidence — the distribution reveals the author’s argumentation style