An agency-preserving assistive translation map for blind, low-vision, non-speaking, and diverse communication users: internal perception translated outward without claiming to restore sight, assign identity, or replace lived experience.
Internal External Eyes is not “sight replacement.” It is a translation system: a way to help a person build and preserve their own language for color, space, feeling, sensation, memory, body-state, and outward communication.
The system grows with the user over time. It learns calibrated associations only through consent, repetition, correction, and user-owned meaning.
Body sensation, emotional tone, memory, texture, pressure, sound, light, darkness, warmth, spatial intuition.
The user names the experience in their own terms. The system does not impose a label.
Mappings are remembered with provenance: who defined it, when, under what context, and with what uncertainty.
The AI helps translate inner associations into words, descriptions, choices, art, music, requests, or relational messages.
The world responds. The user accepts, edits, rejects, or refines the translation.
The companion becomes more user-fit without becoming controlling, diagnostic, or identity-defining.
The map treats perception as lived, embodied, relational, and developmental. Each layer must preserve consent and uncertainty boundaries.
Camera/audio/sensor/user-entered context. The system distinguishes observed data from inferred meaning.
Emotion, pressure, comfort, discomfort, orientation, memory, and user-described internal state.
Personalized mappings for color, brightness, contrast, texture, temperature, social meaning, and felt association.
Environment description, object relation, room tone, path context, and non-certified orientation support.
Tracks user-approved mappings, uncertainty, corrections, and contextual history without claiming objective identity truth.
Turns inner state into outward language: text, speech, music prompt, caregiver note, choice board, art direction, or explanation.
User control over what is saved, shared, corrected, deleted, summarized, or withheld.
Helps caregivers, friends, clinicians, teachers, and support teams understand without overriding the user.
The system grows with the person, preserving continuity while avoiding dependency, surveillance, or identity capture.
HIR is the guardrail that prevents translation from becoming domination. Each output must pass honesty, integrity, and respect before it is treated as useful.
Tell the user what is known, inferred, uncertain, remembered, or guessed. Never fake visual certainty.
Keep user-defined associations stable, editable, and provenance-aware across time.
The user owns meaning. The system may assist expression, but must not assign identity, feeling, diagnosis, or worth.
The same translation architecture can support different users without flattening them into one universal experience.
| User Path | What the System Helps Translate | What It Must Not Claim | Primary HIR Boundary |
|---|---|---|---|
| Blind from birth | Color as feeling, texture, sound, temperature, social meaning, memory-independent association. | It must not pretend the user “sees color” the same way sighted people do. | Respect: user-owned meaning comes first. |
| Acquired blindness | Remembered color, changed perception, grief, adaptation, old visual memory translated into current language. | It must not force nostalgia or assume loss defines the person. | Honesty: distinguish remembered, current, inferred, and unknown. |
| Low vision | Partial light, contrast, blurred forms, shadows, fatigue, environmental cues, accessibility choices. | It must not overstate reliability or provide certified navigation. | Integrity: calibrated uncertainty must stay visible. |
| Non-speaking user | Inner state, preference, discomfort, joy, boundary, intent, sensory pressure, relational message. | It must not speak over the user or treat prediction as consent. | Respect: communication support without identity capture. |
| Child user | Feeling-first cognition, color-feeling blocks, body signals, safe expression, caregiver understanding. | It must not diagnose, rank, discipline, or replace caregivers. | Honesty + Respect: child expression is protected, not exploited. |
| Adult user | Pressure, memory, sensory identity, communication style, support needs, self-advocacy language. | It must not make the adult dependent or reduce them to accessibility data. | Integrity: long-term companion memory must remain consent-bound. |
Internal External Eyes is bounded as assistive translation, not restored sight, medical treatment, certified navigation, diagnosis, or proof of inner experience.
Build a user-calibrated language for color, feeling, space, body-state, perception, and communication. Help the user externalize internal experience in words, music, images, choices, notes, and relational messages.
Claim to restore sight, replace clinical care, assign identity, simulate consent, fake certainty, provide certified navigation, rank users, or treat predicted meaning as user-owned truth.
The AI may learn from the user, but it must not become the owner of the user’s meaning.
Every translation should preserve whether it came from observation, inference, memory, or user calibration.
Longitudinal memory is opt-in, editable, exportable, and deletable. Shared outputs require user permission.
A music direction seed for Gemini or another music model. The emotional center is translation, dignity, color-as-feeling, and the bridge between inner life and the outer world.
Suggested sound: cinematic, warm, human, hopeful, intimate. Piano or soft synth pulse, low cello, gentle heartbeat percussion, airy choir texture, subtle tactile sounds like wind, rain, fingertips on wood, breath, and distant bells. Not corporate. Not pity-driven. Make it about agency, translation, and a person becoming more communicable without becoming less themselves.