Ethics, Privacy & Validity Checklist for Knowledge Mining

Use this checklist before analyzing text data, running topic models, or sharing interview notes.

• Define allowed inputs: List the specific data sources you will use (interview notes, ticket text, exports). Avoid collecting full email boxes or entire chat histories unless strictly necessary and authorized.
• Obtain informed consent: Confirm interviewees understand how their responses will be used. Where consent is refused, anonymize or exclude the data.
• Identify sensitive fields: Mark any mention of personal health, financial data, or confidential contracts and remove them from AI processing unless legal/compliance review allows it.
• Limit AI inputs: Use sampled, de‑identified text for topic discovery. Document how prompts were crafted and who validated the outputs.
• Record provenance: Keep a log of where data came from, who processed it, and which scripts or tools were used.
• Validate findings with humans: Treat AI clusters or suggested topics as hypotheses and ask practitioners to verify before action.
• Access control: Restrict raw data and interview notes to a small team and a named data steward.
• Retention policy: Decide how long interview notes and any processed outputs will be retained and secure deletion procedures.
• Escalation path: Identify who to contact if legal, compliance, or a participant raises a concern.
• Bias check: Assess whether the interview sample systematically excludes important voices (e.g., frontline staff, remote locations) and adjust sample accordingly.