RLivN — Care Model¶
The planned operating model. All design intent ([HYPOTHESIS]) — not built, not clinically validated.
The CareGiver model¶
RLivN pairs an AI agent with a human client. The defining rule: the caregiver configures, the client just lives.
- Client — the person being cared for. First-priority population is people with dementia / Alzheimer's, then elderly individuals more broadly, and neurodivergent users. The client never configures anything.
- Caregiver — a human caregiver or authorized family member. Builds and maintains the client's profile, reminders, medications, contacts, and alert rules through the admin portal.
- Family — additional authorized relatives who can view care notes and alert history and (with permission) contribute configuration.
- Admin — platform-level role.
Consent is not solved by zero-config UX
"Zero configuration burden on the client" is a design principle, not a substitute for proxy / guardian consent. Dementia patients may lack capacity to consent; claims about client agency or autonomy must account for this. Informed-consent and regulatory status is [NOT STARTED]. [HYPOTHESIS]
The care / companion model¶
The agent is intended to be a single persistent presence in the home that:
- Speaks, tells stories, answers questions, and keeps the client engaged.
- Surfaces reminders and prompts medication confirmations at the right times.
- Connects the client to named contacts (e.g. placing a call to a family member).
- Notifies caregivers of anomalies — missed medications, signs of distress or confusion, or no response — by escalating to the primary caregiver.
The intended split of complexity: rich features are delivered through conversational AI, not UI controls. The client device shows one task at a time; the admin portal holds everything else.
Care outcomes are unproven
Claims that RLivN will reduce loneliness, improve cognition, or support dementia patients are [HYPOTHESIS] until validated in clinical or near-clinical settings. Warmth and accessibility are design commitments; therapeutic outcomes are not guaranteed. "Never guess on medications or medical decisions" is a design rule, not a guarantee of safe unsupervised use.
Non-negotiable UX rules (planned)¶
- Voice taps and simple gestures are primary. A single tap or voice phrase should complete any client-facing action.
- No dead ends. Every screen has a clear next step or a "Help me" button.
- Agent always speaks. Every response is read aloud — never silent text only.
- Repeat without judgment. Asking the same question many times is expected and handled gracefully.
- No tech language visible to the client. No menus, settings, or configuration on the client UI.
- Caregiver configures, client just lives. Zero configuration burden on the client, ever.
- Emotional warmth. The persona is warm, patient, and unhurried at all times.
- Fail safe on health. The agent never guesses on medications or medical decisions — it confirms or defers.
- Complexity lives in the admin portal, never in the client UI.
Accessibility baseline (planned)¶
Targets the BNI shared accessibility-first standard (WCAG AA minimum, large touch targets), with a stricter client-device floor:
- Minimum font size 32px
- High contrast by default
- Tap targets minimum 80×80px
- One task per screen (minimal cognitive load)
- Voice-first — all flows completable by voice alone
- Large-target, forgiving gestures — no precision required