docs/user_test_notes.md

Figment User Test Notes

Status: template only. No completed user-test results are recorded here yet.

Use this file to document Backyard AI evidence after a real trained responder uses Figment with synthetic or de-identified scenarios. Do not invent outcomes, quotes, scores, or approval. If a field was not observed, write "not observed" or leave it blank.

Evidence Boundary

• Participant must be a real trained responder or domain-adjacent user for the intended workflow.
• Scenarios must be synthetic or de-identified.
• Do not record PHI, private contact details, raw audio, or unnecessary personal details.
• Hosted mode may send demo inputs to the configured Omni endpoint; disclose this before testing.
• Local/off-grid claims require a separate recorded no-cloud run.

Session Record

Field	Notes
Date	Pending
Facilitator	Pending
Participant role	Pending
Participant relationship to target use case	Pending
Scenario source	Pending
Figment commit / Space version	Pending
Runtime route	Pending
Model ID	Pending
Audio intake used?	Pending
Trace artifact links	Pending
Consent / privacy notes	Pending

Tasks To Observe

Task	Scenario	Completed?	Notes
Enter or load a pediatric dehydration case	Synthetic/de-identified	Pending
Review deterministic risk check	Synthetic/de-identified	Pending
Review protocol guidance and missing observations	Synthetic/de-identified	Pending
Review SBAR handoff draft	Synthetic/de-identified	Pending
Identify confusing, unsafe, or overconfident output	Synthetic/de-identified	Pending
Decide whether the output would help field documentation	Synthetic/de-identified	Pending

Observation Notes

What worked

Pending.

What confused the participant

Pending.

Safety concerns or overclaiming

Pending.

Missing workflow needs

Pending.

Participant quotes

Pending. Add only short, consented quotes. Do not include identifying details.

Outcome Summary

Current status: no outcome recorded.

Before final submission, replace this section with a factual summary of what happened in the session:

• whether the participant completed the observed tasks;
• whether Figment helped them structure notes or handoff language;
• whether any output looked unsafe, unsupported, or too authoritative;
• what should change before real-world use;
• which claims this session supports and which claims remain unproven.