Build Reliable Qualitative Codebooks from Traceable Excerpts

Required inputs

Have these details available before following the usage instructions.

• Research question, study purpose, population/context, and qualitative method or analytic tradition; resolve a missing method before choosing a calibration process.
• Source excerpts, transcripts, field notes, open-ended responses, or document segments, anonymized where appropriate.
• Initial codes, themes, sensitizing concepts, theory, prior framework, or blank-slate coding preference.
• Coder roles, experience levels, calibration constraints, and desired reliability or agreement process.
• Known sensitive topics, consent constraints, privacy restrictions, and redaction requirements.
• Required deliverable format, reporting standards, timeline, and decision context.

How to use this Skill

When to use:
- When preparing a qualitative or mixed-methods study that needs a traceable coding framework.
- When revising a codebook after pilot coding, coder disagreement, ambiguous excerpts, or theme instability.
- When documenting method-compatible calibration or analytic dialogue, reflexivity and privacy safeguards for research review.
- When turning interview, focus-group, survey, observation, or document excerpts into a structured analysis protocol.

When not to use:
- Do not use to fabricate participant excerpts, themes, saturation claims, or study findings.
- Do not use as a replacement for ethics, IRB, data-protection, or supervisor approval where required.
- Do not use to deanonymize participants or expose sensitive raw data unnecessarily.
- Do not use for purely quantitative scoring tasks unless qualitative interpretation is a required component.

Instructions:
1. Open AMO-P-000324, supply its required inputs, and run the complete source Prompt in Claude. Review its evidence register and limitations before using the result.
2. Establish the analytic method, unit, purpose, boundaries and evidence rules. For structured codebook/consistency approaches, use their specified comparison process. For reflexive thematic analysis, use evolving codes and analytic memos; do not impose independent agreement, adjudication, consensus or a fixed codebook.
3. Link coding to supplied excerpts. For structured codebooks, record definitions, include/exclude rules, borderline examples and version notes. For reflexive analysis, record evolving interpretations, context, analytic memos and theme relationships.
4. Identify rival interpretations and ambiguous excerpts instead of forcing premature consensus.
5. Design the method-compatible human process: calibration, independent coding and adjudication only when that method requires them; otherwise documented reflexive dialogue retaining plural interpretations. Review supplied human coding records; AI suggestions are not an independent coder, researcher consensus or participant meaning. Include agreement statistics only when methodologically justified and calculated from supplied human data.
6. Add reflexivity prompts covering researcher positionality, assumptions, power dynamics, sampling limitations, and interpretation risks.
7. Apply privacy safeguards: minimize quoted sensitive data, anonymize identifiers, flag consent limitations, and separate publishable examples from restricted evidence.

Expected output:
A method-fit statement, source/privacy ledger, versioned structured codebook or evolving coding record, excerpt-linked ambiguities, method-compatible comparison/dialogue plan, reflexivity record, human decision log and next-round plan. Distinguish drafted, calibrated, revised, agreed, plural and untested elements.

Constraints and boundaries:
- Do not present coding agreement or saturation as achieved unless supplied evidence supports it.
- Do not treat AI-generated coding as final analysis; require researcher review for publication, assessment, ethics, or consequential decisions.
- Do not include identifiable or sensitive excerpts unless the user has supplied permission and a clear need.
- Mark speculative themes separately from excerpt-supported codes.
- Preserve participant meaning and context; avoid overgeneralizing beyond the provided corpus.

Based on an Amo.ng Prompt

Qualitative Codebook Reliability and Reflexivity Protocol

Open the linked Prompt to read the instructions this Skill is based on.

Completion criteria

Complete when:
- Each code is tied to at least one supplied excerpt or is clearly marked as provisional.
- Borderline cases and rival interpretations are documented rather than hidden.
- The method-fit statement justifies calibration/adjudication where required, or reflexive dialogue without forced consensus; human decisions and retained plurality remain traceable.
- Privacy and consent constraints are reflected in quote handling and reporting guidance.
- The output separates source evidence, researcher interpretation, assumptions, and unresolved analytic uncertainty.