Reminiscence Therapy Device: Prompt Library Design Without Clinical Claims

Content Scope Is a Regulatory Boundary
A reminiscence therapy device is defined by what it does not do. It does not assess, diagnose, monitor or intervene in memory function. Its prompts — whether asking about first bicycles, radio brands or school uniforms — exist solely as voluntary conversation starters. The distinction is structural: every prompt must be skippable, repeatable and decoupled from any output that suggests progression, scoring or longitudinal tracking. This is why WingPhoenix treats prompt libraries as firmware modules, not AI models — each revision undergoes separate CE RED (EN 301 489) and FCC Part 15 B/C retesting if audio assets change playback timing or amplitude. For full details on how firmware updates affect compliance scope, see the CE RED document checklist.Era Questions Must Anchor to Public, Non-Contested Memory
“Where were you when the Berlin Wall fell?” risks emotional variance and political interpretation. Better prompts name tangible, widely distributed cultural artefacts: “What colour was your first transistor radio?” or “Which cereal mascot appeared on your lunchbox in 1972?” These rely on shared material culture, not contested history. All era questions are preloaded — no internet-dependent retrieval — and remain fully functional offline. This aligns with the companion robot’s offline-first architecture and supports importers needing predictable RF behaviour under EN 300 328.Repetition Is a Feature, Not a Limitation
Older users often request the same question three times in one session. The system honours this without logging, flagging or altering subsequent output. Repetition is handled at the interaction layer — not the model layer — meaning no embedded memory of prior answers is retained or used to infer engagement. This avoids unintended data processing under GDPR Article 4(1) and simplifies audit trails. It also means no cloud sync is required, reinforcing the smart health ring and health smartwatch design principle: local execution where possible.Caregiver Notes Are Structured Metadata, Not Clinical Records
Caregivers may add short text notes — e.g., “Spoke warmly about seaside holidays” — but these are stored locally, encrypted at rest, and never transmitted unless explicitly exported via USB. No timestamped behavioural log is generated. No inference engine reads these notes to adjust prompts. This keeps the device outside MDR classification and ensures alignment with the AI wearables line’s documented risk assessment framework.What This Platform Cannot Do — And Why Importers Must Enforce It
It cannot claim to improve recall, slow memory loss or support dementia care pathways. No report exists — nor will WingPhoenix commission one — linking prompt usage to cognitive metrics. Any third-party study citing “therapeutic benefit” refers to external research, not product validation. Importers must reject supplier language suggesting clinical utility, especially where distributors insert their own marketing copy. This boundary is non-negotiable: crossing it triggers MDR, HIPAA or UK MHRA obligations that WingPhoenix does not assume. See how to vet a supplier who respects this line in how to vet a supplier.Procurement Checklist: Firmware, Content and Compliance
| Decision Point | WingPhoenix Standard | Importer Action Required |
|---|---|---|
| Prompt library update | Firmware module, tested per revision | Rebook CE RED (EN 301 489) and FCC Part 15 B/C if audio assets change playback profile |
| Offline operation | All era questions preloaded; no cloud dependency | Verify no background telemetry in test build via network capture |
| Caregiver note export | USB-only, no cloud sync, no auto-upload | Confirm encryption-at-rest implementation matches declared algorithm |
| Mood context use | Only for prompt pacing — no biometric input or inference | Audit firmware to confirm absence of ambient audio analysis or voice stress detection |
| Age grading | Defined by packaging label; toy scope booked per build | Supply EN 71-1/-2/-3 test reports if age grade ≤14 years |
| Battery compliance | UN38.3 + SDS supplied; watt-hour printed on unit | Validate print legibility and SDS version against shipment batch |
| Skin-contact items | EN 1811 nickel release applied to rings and wristbands | Request test certificate referencing exact plating process used |
WingPhoenix builds for clarity of function — not feature density. A reminiscence therapy device works because its limits are explicit, its content is anchored in public memory, and its compliance path is traceable per hardware and firmware revision. To begin scoping a prompt library for your market, review the OEM/ODM services page, then book a technical briefing using the MOQ, sampling and lead times guide to align on firmware revision cycles and test scheduling.
Questions buyers ask
What makes a reminiscence prompt library safe for CE RED and UKCA when targeting older users?
It avoids diagnostic language, clinical outcomes, or therapeutic intent in both UI text and backend logic — framing all output as voluntary reflection, not treatment. Compliance rests on how prompts are scoped, not just how they’re phrased.
How does WingPhoenix treat mood-related outputs in a reminiscence therapy device?
Mood is addressed only as surface-level context for prompt sequencing — e.g., selecting gentler era questions after prolonged silence — never as a measured state or intervention target. No biometric inference is performed or claimed.
Why did our distributor’s sample fail EMC testing when adding custom voice prompts?
Because untested audio assets triggered unexpected RF emissions during playback — a reminder that firmware revisions carrying new prompt sets must rebook CE RED (EN 301 489) and FCC Part 15 B/C scopes with an accredited lab before shipment.
What is the one claim no supplier should make about reminiscence content — and why does it matter?
That prompts “support cognitive decline mitigation”. This implies a medical device function under EU MDR and UK MHRA rules. WingPhoenix does not develop, validate or claim any effect on neurocognitive conditions — full stop.
Related: more on ai companions · buyer guides · customisation tiers