# Newcomer pilot — proposed, not run

The first question is practical: can a newcomer complete a session and understand
its output without mistaking the guide for an AI-consciousness verdict?

## First: an unaided usability check

Invite three to five consenting adult volunteers for an exploratory check.
Use fictional material rather than asking for private chats. Explain that the
tool is experimental, participation is voluntary, and they can stop at any time.
Do not collect names, accounts, or other personal details in the exported examples.

Give each person the website with one task: examine the fictional remembering
exchange, correct the claim if needed, prepare a question, and export a record.
Do not walk them through the controls. Record any help required and the exact
point of confusion. Agree before starting what notes, if any, may be retained.

Ask afterward:

1. What did the supplied record directly show?
2. What additional interpretation was proposed?
3. Which system or interval does the claim concern?
4. What observation would change the assessment?
5. What information would leave the browser if you used another AI or shared an export?

Before broader release, address any critical misconception: a badge interpreted
as a consciousness verdict; assumed source authentication; missing evidence
treated as absence; lost work during an ordinary correction/export; or unexpected
disclosure of full sources. Report completion and assistance counts as observed,
not as a population success rate. A small pilot finds problems; it does not
establish educational effectiveness.

## Later: a separate effectiveness question

If that workflow is understandable, consider a prospectively specified comparison
against an equally clear written explanation with the same teaching material and
comparable exposure time. Give participants a different transfer case afterward.
Include unsupported positive attribution, unsupported denial, and a documented
lower-level capability that should retain its credit.

Fix assignment, scoring, exclusions, sample rationale, exposure history, and the
success/failure criterion before collecting results. Have reviewers other than
the author score anonymized responses without condition labels where feasible;
retain disagreements. The outcome is accurate reasoning about a new case,
not agreement with a preferred position about AI.

Do not retrospectively describe the usability check as this comparative study.
No participants have been recruited, contacted, or tested for this release.