Cultural Context Identification
AI notes language, origin, and cultural signals before drafting.
- 40+: Cultural profiles tracked (dialect and region tags)
- 3: Language preference fields (primary, literacy, interpreter)
- 6: Context signals used (name, locale, referral notes)
- Flagged: Misidentification risk (human review triggered)
Signals the system reads
Locale, language choice, and referral notes seed the profile.
Self-reported data always outranks inferred signals.
Avoiding stereotyping
Context is a starting hypothesis, not a fixed rule.
Human oversight
Clinicians can correct or override any inferred context.
Directness & Family-Involvement Norms
AI estimates preferred bluntness and family role before writing.
- 1–10: Directness scale (indirect to direct)
- 1–10: Family-involvement scale (individual to family-centered)
- 3: Norm sources (literature, patient input, notes)
- Live: Update frequency (sliders adjust in real time)
Directness spectrum
Some cultures favor plain talk, others favor softened framing.
Family-centered decision norms
Many cultures involve family in medical choices by default.
Assumptions must yield to the patient's stated preference.
Health-belief sensitivity
Traditional and biomedical beliefs are weighed together.
Baseline Medical Message Drafted
AI writes one neutral, clinically accurate message first.
- Neutral: Draft tone (no cultural styling yet)
- Pass: Clinical accuracy check (content locked before styling)
- Grade 8: Reading level (plain-language baseline)
- Pending: Adaptation layer (applied next stage)
Why draft first
Separating content from style prevents accuracy drift.
Plain-language baseline
Baseline avoids jargon regardless of eventual audience.
Content lock
Medical facts stay fixed through every later rewrite.
Style adapts; clinical substance never changes.
Tone, Framing & Involvement Adjusted
AI rewrites tone, phrasing, and family framing to fit context.
- ±100%: Tone shift range (soft to direct rewrite)
- 2: Family phrasing variants (individual vs inclusive wording)
- Swapped: Belief-sensitive phrases (matches health-belief framing)
- Rising: Adaptation confidence (recalculated per edit)
Tone restyling
Sentence structure softens or sharpens with directness norm.
Family-inclusive wording
Pronouns shift toward "you and your family" when expected.
Wording never overrides the patient's own stated wishes.
Belief-sensitive framing
Explanations bridge biomedical and traditional health beliefs.
Culturally-Adapted Message Delivered
Final message respects tone, family, and belief norms together.
- 90%+: Adaptation confidence (at delivery threshold)
- Higher: Patient comprehension lift (versus generic messaging)
- Lower: Cultural mismatch complaints (reported in pilot use)
- Required: Clinician review step (before final send)
Final quality gate
A clinician reviews tone and content before sending.
Continuous feedback
Patient response refines future cultural-style estimates.
Adaptation improves the model; it never fully overrides discretion.
Scope boundary
This adapts routine communication, not high-stakes decisions.