Rules that guide us to protect our patients
Vital Mosaic was created by a complex illness patient and enterprise technology strategist to transform fragmented health records into actionable clarity. But deep medical data is a double-edged sword: without deliberate ethical boundaries, artificial intelligence can amplify panic, isolate vulnerable individuals at 2:00 AM, or damage doctor-patient relationships. Here are the unyielding rules that govern everything we build.
“Vital Mosaic should help people understand what their health information may mean, what it cannot establish, and what they can responsibly do next.”
The Four Core Rules
These four non-negotiable rules govern every feature design, line of code, assistant prompt, and clinical communication in Vital Mosaic.
Official Brand Name & Spelling
We always write “Vital Mosaic” with an intentional space between words across all user interfaces, clinical documentation, and patient communications. It represents the careful, patient assembly of fragmented tiles into a unified, coherent picture of health.
Multi-System & Multi-Faceted Language
We do not fixate on or declare specific diagnostic labels in our general platform wording, headers, or default framing. Chronic multisystem illness involves interconnected physiological axes across neurology, immunology, cardiology, and gastroenterology simultaneously.
Respectful Clinician Partnership
Respectful clinician partnership. We do not disparage, attack, or foster adversarial dynamics with doctors. Physicians, nurse practitioners, and clinical staff are operating under crushing documentation demands and severe 15-minute appointment constraints.
Emotional & Anxiety Safeguards
Our design must protect against emotional triggers, health anxiety, hypochondria, OCD/rumination, and medical trauma. We present trends in a calm, grounded manner—never with alarming red banners. Lab variations are contextualized as normal biological fluctuations to be reviewed collaboratively with a physician.
The Three Ethical Risks We Actively Combat
Artificial intelligence in healthcare can empower agency, but without safeguards it can harm. We design explicitly to eliminate three failure modes:
1. Information Without Clarity
Dumping raw pathology codes, uninterpreted lab flags, and complex PDFs onto a patient sitting alone at 2:00 AM causes acute panic. Our Guided and Easy Modes translate complex terminology into plain language, contextualize fluctuations, and provide reassuring, grounded perspective.
2. Data Without Context
Raw genetic variants and isolated out-of-range lab markers can be terrifying when viewed in a clinical vacuum. We tether every datapoint to the patient’s longitudinal baseline, actual physical symptoms, and bounded clinical probabilities—never abstract worst-case fears.
3. Protecting the Right User
For a chronic patient whose illness has been invisible, granular data is liberation. For someone trapped in somatic anxiety, infinite trackers become fuel for panic. We build mindful friction, rate-limit anomaly loops, and adapt question cadences to nurture psychological safety.
The 7-Question Design Decision Checklist
Before any new screen, assistant script, or notification is released, our engineering and clinical teams must answer these 7 questions:
| # | Design Question | Our Required Standard & Behavioral Guardrail |
|---|---|---|
| 1 | Does this add context or merely produce an answer? | Answers in a vacuum create anxiety. Every insight must be anchored to the patient's lifestyle, longitudinal baseline, and physical history. |
| 2 | Could a patient mistake this for a diagnosis? | Circuit breakers intercept diagnostic declarations (“You have...” → “Your records show a pattern your doctor may investigate...”). AI assists; only clinicians diagnose. |
| 3 | What happens emotionally after the patient sees it? | If a screen spikes heart rate or triggers dread, the presentation fails. High-stakes findings must be gated by calm, grounded framing and mindful friction. |
| 4 | Are uncertainty and data limitations clear? | Confidence bounds, missing outside records, and model limits are displayed directly on-screen—never buried in tiny terms of service. |
| 5 | Does the next step match the seriousness of the finding? | Routine variations suggest “Discuss at next routine visit”; acute crisis indicators offer immediate 1-tap human clinical bridges. |
| 6 | Is human care easier to reach from here? | Every summary screen provides 1-tap bridges: generate a 2-page Specialist Appointment Brief, message a doctor, or access emergency support lines. |
| 7 | Would I still consider this responsible if the data were wrong? | If an OCR scan misread a character or a wearable recorded an artifact spike, our design protects the patient from catastrophic fear or dangerous medical overreaction. |