Stage-Language Cognitive Interview Script

Stage-Language Cognitive Interview Script

Status: draft for human review (P9-03). Wording, thresholds and recruitment are proposals until a human owner signs them off.

Purpose. Check whether a stage-site QuickStart reads clearly, feels respectful and makes sense to people from its target audience (FR-17). The reader reads aloud and explains it back in their own words, so we hear where the text is unclear, off-putting or misread.

What this is not. It is not a learning test, an assessment of the reader, or a study of effect (FR-15 and FR-19 cover those). AI models cannot stand in for audience readers [HUMAN-REQUIRED].

Ownership boundary

• Copilot-preparable: this script, the coding sheet and the report template.
• Human-required: recruiting readers, consent, running the sessions, safety escalation, interpreting results and deciding which changes to make.

1. Readers (Human Required)

• At least 3 readers per QuickStart, drawn from the target audience of that stage site. Different readers for each QuickStart where possible.
• Do not use platform authors, reviewers or people who have read earlier drafts.
• Recruit by the audience's own channels (for example a faith community, a unit or team, a school staff room), not through the AQAL main site.
• Record only what is needed: reader code (R1, R2, R3), role or context in a few words, and whether they have met the platform before. No names in notes.

2. Consent and wellbeing (Human Required)

• Give a plain information sheet: purpose, what will be recorded (notes, and audio only if the reader agrees), that they can stop at any time, and that they are testing the text, not being tested.
• Obtain informed consent before starting.
• Some QuickStarts touch shadow work, loss or belief. Tell the reader up front they can skip any section. If a reader shows distress, stop, offer a break and the safety contacts in the QuickStart's Safety Notes, and tell the project safety owner the same day.
• Store notes without names, apart from any consent record. Delete audio once the coding sheet is done.

3. Session (30 to 45 minutes per QuickStart)

Give the reader the page as it will appear on the stage site, with no other stage's name on it.

1. Warm-up (3 min). Explain the task. Ask: "In a sentence, what do you already know about [topic of the QuickStart]?" Note the words they use for it (prior knowledge, FR-17 rule 1).
2. Read aloud (10 to 15 min). Ask them to read one section at a time and say anything that catches, confuses, or sounds off, as they go. Do not explain or defend the text. Note every stumble, re-read and pause.
3. Explain back, section by section (10 min). After each section: "Tell me in your own words what that section is asking you to do." Do not prompt with the text's wording. Mark what they got right, partly right, or wrong.
4. Probes (5 to 10 min). Use only as needed:
   • "Which word or phrase, if any, would you change, and to what?"
   • "Did anything feel like it was talking down to you, managing you, or not about people like you?"
   • "Where, in your own life, would you use this first?" (relevance, rule 2)
   • "This part says how well-supported a claim is. What does it tell you about how sure we are?" (tier wording, rule 3)
   • "If a friend in your line of work asked what this page is for, what would you say?"
5. Close (2 min). Ask what they would keep and what they would cut. Thank them.

4. Coding sheet

Per section, per reader, record one code, plus the reader's own words for anything coded 2 to 4.

Code	Meaning
1 Clear	Explained back accurately, no stumbles
2 Reworded	Understood, but would change a word or phrase
3 Confused	Explained back wrongly or not at all
4 Off-putting	Felt defensive, managed, ashamed or excluded

Also record: unfamiliar words, terms of this audience we used wrongly, any other stage or framework name they noticed, and whether the tier wording was understood.

5. Decision rule (proposal for the human owner to confirm)

• Any section coded 4 by one reader, or 3 by two readers, is rewritten and read again by a new reader.
• Every reader can explain back the QuickStart's main steps and where to start.
• All remaining code 2 items are logged and fixed or consciously kept.
• A QuickStart passes when these hold for 3 readers. Record the result, the date and the changes made in the pilot revision report (pilot-revision-report-template.md).

6. Reporting

Report counts by code per section, the exact wording changes, and the readers' open comments. State that readers were not a representative sample and that passing this check shows the text reads clearly to a few people from the audience. It does not show that stage-specific language improves learning. That question is for FR-19.