Series
Ethics, Autonomy, and Delegation
Seven mechanisms that determine what the system can do, what it must refuse, and how the person controls the boundary between them. Series 04 is the ethical architecture that governs every other layer of the platform.
BMT-04.01
The Human Agency Scale
A 0.0-to-1.0 autonomy spectrum with domain modifiers and cognitive state adjustment
A 0.0-to-1.0 autonomy spectrum with nine domain modifiers and a real-time cognitive state adjustment. The person sets the dial. The architecture enforces the consequences per …
BMT-04.02
Earned Autonomy
Five progression levels, the evidence package, dependency detection, and the reverse direction
Autonomy earned through demonstrated competence across five progression levels, with dependency detection that works against the system's own indispensability. Trust moves in both …
BMT-04.03
Contextual Consent
Three consent tiers, four capacity scenarios, lucid window handling, and consent propagation
Three tiers of consent matched to risk: foundational at onboarding, domain per agent, transactional in the moment. Lucid windows confirm existing consent but never request new …
BMT-04.04
The Escalation Hierarchy
Five levels with failure modes, the cognitive state paradox, and escalation timeout
Five escalation levels with named failure modes. The cognitive state paradox: over-escalating to a person with reduced capacity creates worse outcomes than careful automated …
BMT-04.05
Cognitive Capacity and Consent
Three governing principles, the capacity spectrum, and the decision-maker transition
Three principles held in tension for the person whose capacity changes: prior preferences as anchor, current capacity as scope of modification, dignity never traded for safety.
BMT-04.06
What the System Must Refuse
Eight hard constraints with override prevention architecture
Eight behaviors the system will not perform regardless of who asks. Hard constraints enforced at the infrastructure level through three redundant mechanisms.
BMT-04.07
Privacy as Architecture
Four tiers with distinct implementations, aggregation detection, and five engineering principles
Four privacy tiers with distinct architectural implementations, not just labels. Aggregation detection catches the inference risk that individual data elements in light tiers …
Synthesis