advisor.ethics

Alpha (0.1.0a1), and a mock: it does not yet do the Research Ethics Adviser's real work.

The Research Ethics Adviser is one of the personas of the agent roster (Agents). It serves two steps of the Blueprint's process flow:

Step	What the mock does
governance.sensitive-data-identify (identify sensitive data)	Flags each field whose name suggests personal or otherwise sensitive data, and asks the researcher to confirm consent and ethics approval when the data involve people.
governance.embargo (confirm whether an embargo is required)	Asks whether an embargo is needed, or asks the researcher to check an intended one against their policies.

It flags and stops there. How sensitive data is treated, governance.sensitive-data-handle, is a person's decision, and no agent serves it (Agents, C2). The adviser never suggests anonymisation.

Input and output

SensitivityContext gives the dataset's title, description and fields (field or column names), and whether it involves_people and whether an embargo is intended (embargo_intended). Each is optional, but a context with neither fields nor involves_people abstains with INSUFFICIENT_INPUT.

SensitivityFlags holds:

• flags: one SensitivityFlag per field whose name matched, with the category (direct identifier, indirect identifier, special category, location) and the keyword matched;
• questions: what the researcher must answer or confirm, and with whom;
• limitations: what the mock does not do.

How it decides

Each field name is split into words and matched against keywords.yaml. The adviser reads no data values, consults no source and calls no model, so its grounding mode is none and it cites only its input.

What it does not do

• It reads names, not values: a field it does not flag may still hold sensitive data.
• It does not check free text, images or audio, or combinations of fields that together identify people.
• It does not know the participants' consent, the ethics approval or the applicable law.

The real adviser is TODO(#355).