Skip to content
Getting startedNot reviewed

AI assistance, evidence, and privacy

Understand how contextual explanations are separated from statistical computation and controlled by configured privacy boundaries.

When to use this

Use this before requesting an AI explanation or writing draft.

Why it matters

Researchers need to know which evidence is included, which provider is configured, and that AI text is not a statistical result.

Step-by-step instructions

1. Open the AI Research Assistant for the intended project.
2. Review the configured provider and privacy notice.
3. Select an allowed task and evidence context.
4. Avoid including unnecessary respondent-level or identifying content.
5. Review the response against saved results and source evidence.
6. Edit and approve any text before using it in research writing.

Expected outputs

A contextual explanation or draft with provenance boundaries, not a new statistical calculation.

How to interpret the results

Treat AI output as assistive text that may contain errors. Verified numerical results remain those produced by the analysis engines.

Common mistakes

Do not paste confidential respondent data into prompts. Do not cite AI text as evidence or allow it to replace methodological judgement.

Was this article helpful?

Your response helps the methodology team improve this documentation.