BMT-04.05 Executive Summary#
BlueMirror.tech | May 2026#
Margaret set her healthcare autonomy to 0.55 three years ago, when her memory was sharp. She made considered choices about what the system should handle and what she wanted to keep for herself. She cannot recall making those choices now.
The system that freezes her 2022 preferences in amber and executes them forever has failed her. The system that interprets her current cognitive state as authorization to take on more authority has also failed her. The space between those two failures is where the hardest ethical question in this architecture lives.
The common framing of “safety versus autonomy” is the wrong frame. The actual question: what does it mean to serve a person whose capacity is changing, without either abandoning her to risk she cannot assess or imprisoning her in safety she did not choose? The architecture answers with three principles held in tension, not resolved into a clean hierarchy.
Prior capacity preferences anchor current behavior. What Margaret wanted when she could clearly express it remains the baseline. The system does not override her prior preferences because her current capacity is reduced. It implements them more carefully. Current capacity determines the scope of modification: Margaret can maintain or narrow her prior preferences but cannot expand them, because expansion requires capacity commensurate with the decision. Dignity is never traded for safety: the system that removes all autonomy to protect Margaret from every possible risk has erased her, not protected her.
Cognitive capacity fluctuates across multiple timescales. Daily fluctuation (lucid mornings, confused evenings) is normal and handled through continuous adjustment. Weekly variation is averaged to avoid whiplash. Trend trajectory over months or years produces gradual, imperceptible adaptation. Acute events from infection, medication change, or hospitalization require a faster response.
Three concrete scenarios demonstrate the principles. Margaret’s prior preferences anchor behavior: the system implements her 0.6 autonomy setting more carefully as capacity declines, using simpler language and clearer options. Margaret’s daughter Elena cannot expand Margaret’s delegation scope without legal authority; if she lacks it, the system explains the pathway for establishing it. Margaret’s desire to continue her cooking class despite a low-capacity assessment this week is served with modifications: shorter duration, a cognitive check-in at thirty minutes, and caregiver awareness.
The decision-maker transition has four non-negotiable properties. It is legal, not algorithmic: a verified legal instrument authorizes it, not the system’s own capacity assessment. It is domain-specific: a healthcare proxy gets healthcare authority, not financial. It is baseline-preserving: the decision-maker cannot undo the person’s prior preferences without documented justification. It is auditable and reversible: if capacity returns, the system offers to restore the person’s direct authority.
The architecture does not claim the Cognitive State Estimator is a clinical diagnostic tool. It detects behavioral patterns, not diagnoses. Edge cases will arise that the framework handles imperfectly. The default in ambiguity is to preserve the person’s most recent clearly-expressed preference and escalate to a human decision-maker.
The full article is available at bluemirror.tech.