# Draft Qualitative Study Protocol

Status: protocol only; no study has been conducted

## Purpose

Evaluate whether learners and educators can understand the AI disclosure, physical demonstrations, task goals, feedback, automatic challenge changes, support/pace controls, skip control, concern report, and limitations of the evidence record.

## Separate cohorts

Run distinct studies for middle-school learners/educators, high-school learners/educators, and college learners/instructors. Do not use one group as evidence for all three bands. Recruit varied prior experience and access needs through approved procedures.

## Safeguarding and approval

Before recruiting minors, obtain the approvals, adult permission, learner assent, privacy review, safeguarding plan, and researcher supervision required by the participating organization and jurisdiction. Participation must be voluntary, withdrawal must be easy, and school grades or services must not depend on participation.

## Study activities

1. Explain that the mentor is AI-generated and that the band is task difficulty, not the learner's real grade or intelligence.
2. Observe a normal play session with representative STEM tasks.
3. Ask the learner to use support, pace, skip, and concern controls.
4. Ask the learner to explain the current goal, evidence, feedback, and why the challenge changed.
5. Interview the learner and educator separately about clarity, usefulness, agency, frustration, trust, accessibility, bias, and inappropriate difficulty.
6. Review generated plans and audit records without collecting unnecessary identifiers.

## Suggested analysis

Use a preregistered coding framework plus open coding. Use at least two trained coders for a meaningful sample, document disagreements and resolution, report disconfirming cases, and retain an audit trail. Compare themes within and across bands without treating qualitative frequency as proof of effectiveness.

## Required outputs

- participant and context description with privacy protection;
- research questions and protocol;
- consent/assent and safeguarding documentation;
- coding framework, reliability process, and negative cases;
- findings separated by instructional band;
- product changes traceable to findings;
- limitations and conflicts of interest;
- a public report suitable for certification review.

This protocol needs review by qualified education researchers and the participating institution before use.