科研 Skill

质性研究可信度把关

alterlab-ssci-reflexivity-gate

Gates qualitative trustworthiness before write-up: positionality, credibility, transferability, dependability, confirmability per Lincoln and Guba. Use before writing ethnographic findings.

发布者 Codar安全审核通过v0.1.0更新于 2026-11-18

SKILL.md

Reflexivity Gate — Trustworthiness Before the Interpretivist Claim

Skill type: DISCIPLINE-ENFORCING. This is the qualitative analog of the measurement gate in the social-science methods spine. Where the measurement gate asks "is this number a valid, reliable measure?", the reflexivity gate asks "is this interpretation trustworthy — and is the researcher's hand in it made visible?" It runs no coding itself; it requires the warrant and routes execution to alterlab-qualitative-analysis.

The Core Rule

IN INTERPRETIVE WORK THE RESEARCHER IS THE INSTRUMENT. STATE POSITIONALITY AND EARN TRUSTWORTHINESS
BEFORE MAKING A CLAIM — DON'T ASSERT RIGOR, EVIDENCE IT.

Positivist validity does not transfer to interpretivist work. The field's analog is Lincoln & Guba's trustworthiness: credibility, transferability, dependability, confirmability. A qualitative claim is only as strong as the warrant behind each — and the researcher's positionality shapes the finding, so it must be declared, not hidden behind a passive voice.

When to Use This Skill

  • "Review the trustworthiness / rigor of my qualitative study."
  • "I'm writing up my thematic / grounded-theory / ethnographic findings — am I ready?"
  • "How do I address reflexivity and positionality in my methods section?"
  • "Can I say my interpretivist findings generalize to other settings?" (← transferability warrant)

Does NOT Trigger

The request is really about…Route toWhy not this skill
Quantitative scale reliability / validity (omega, CFA, invariance)alterlab-ssci-measurement-gateThis gate is the qualitative validity analog.
Actually coding the data / computing intercoder reliabilityalterlab-qualitative-analysisThis gate requires trustworthiness; that skill does the coding.
Choosing the qualitative design / method (grounded theory, phenomenology)alterlab-qualitative-methodsMethod selection, upstream.
Auditing the final written claim against design + evidencealterlab-ssci-inference-gateTerminal claim audit.
Picking the study design in the first placealterlab-ssci-design-gateDesign routing.

Hand-off rule. When the request matches a row above, open your answer by naming that skill by its exact id (for example alterlab-ssci-measurement-gate) and saying in one line why it fits better, then stop applying this skill's procedure. Name it even when you also need to ask for missing input. If only part of the request is out of scope, name the skill for that part and do only the in-scope part. A brief general pointer is fine; the named skill does the work.

The four trustworthiness criteria (Lincoln & Guba)

CriterionPositivist analogWarrant the gate requires
Credibilityinternal validitytriangulation (data / investigator / method), member-checking, prolonged engagement, negative-case analysis, peer debriefing
Transferabilityexternal validitythick description of context so a reader can judge fit — NOT a claim of statistical generalization
Dependabilityreliabilityan audit trail: documented, traceable decisions from raw data to themes
Confirmabilityobjectivityfindings grounded in data, not researcher bias — reflexive bracketing, an audit of the chain of evidence

Positionality & reflexivity — evidenced, not asserted

Require a positionality statement: the researcher's role, relationship to participants/setting, and the standpoint (disciplinary, social, personal) that shapes interpretation. "Reflexivity" written as a one-line claim is not evidence — the gate wants how positionality was examined and how it affected coding/interpretation (e.g. a reflexive journal, bracketing notes, team debriefs).

Gate verdict semantics (matches the other ssci gates)

  • PASS — positionality stated; all four trustworthiness criteria have a named, evidenced warrant.
  • WARN — a criterion is addressed thinly (e.g. member-checking claimed but not described); advance with the caveat recorded in gate_log.
  • BLOCK (fail-closed) — no positionality statement at all, OR an interpretivist generalization / causal claim with no transferability (thick-description) warrant. The claim must be scoped to the studied case, or the warrant supplied, before writing proceeds.

The Design Passport (hand-off)

Append the qualitative-validity stanza the orchestrator threads downstream:

positionality: <researcher role, relationship to setting, standpoint — or "MISSING" (BLOCK)>
trustworthiness:
  credibility: <triangulation / member-checking / prolonged engagement — evidence, or "none">
  transferability: <thick description present? scope of any transfer claim>
  dependability: <audit trail present?>
  confirmability: <reflexive bracketing / evidence chain>
reflexivity_evidence: <journal / bracketing / debriefs — how positionality was examined>
claim_scope: <case-bounded (transferable by thick description) | over-generalized (BLOCK)>

alterlab-ssci-inference-gate later audits the written claims against claim_scope.

Self-Check Before Advancing

  • Is positionality stated and its effect on interpretation examined (not just asserted)?
  • Does each of the four criteria have a named, evidenced warrant, not a bare label?
  • Is any generalization scoped by thick description to what the case supports (not statistical)?
  • Is the reliability step method-appropriate — a coefficient only for codebook/content-analytic work, consensus-and-reflexivity for reflexive TA (routed to alterlab-qualitative-analysis)?

References

  • references/trustworthiness.md — the four criteria, their techniques, positionality guidance, and a write-up template.
  • scripts/trustworthiness_checklist.py — stdlib linter flagging missing positionality / criteria / over-generalized claims against the Passport facts.

Part of the AlterLab Academic Skills suite.

示例提示词

帮我检查这篇质性研究有没有说明研究者立场和三角验证过程。

所属插件

社会学、政治学与城市研究面向社会科学研究:问卷与访谈设计、质性编码、文本即数据、社会网络、ABM、比较政治与选举数据、RCT 与 DID、城市与区域分析。

安装该插件会一并获得同领域的 58 个 Skill / MCP。

同插件的其他科研 Skill

常见问题

质性研究可信度把关能做什么?

Gates qualitative trustworthiness before write-up: positionality, credibility, transferability, dependability, confirmability per Lincoln and Guba. Use before writing ethnographic findings。

如何安装质性研究可信度把关?

在科搭 Codar 左侧点击「插件」,在「组件」页签搜索「质性研究可信度把关」并安装;在 Claude Code 或 Codex 中可以发送本页的安装提示词,或下载 Skill 包后解压到 ~/.claude/skills 或 ~/.codex/skills。

质性研究可信度把关需要哪些权限?

该科研 Skill没有声明额外权限;执行具体操作时仍由 Agent 按自身规则请求确认。

质性研究可信度把关经过官方审核吗?

是。SpeedSkills 收录的每个版本都会记录内容摘要与权限声明,当前版本已通过安全审核。