Relative-Position Computation Principle

Canonical definition

The Relative-Position Computation Principle states that the architecture contains no Object-class-specific or human-specific positional module; instead, the Observer computes a signed relative-position estimate of the Individual against any agent the Thinker has parsed and represented as an agent, using a single position-computation operation indifferent to the agent's class. The agent-type categories over which this computation runs (Prey, Predator, Rival, Collaborator, and any other categories the Thinker constructs and maintains) are not architectural primitives but are Thinker-maintained constructions; the positional computation itself privileges no category. Because the computation is agent-class-neutral, the same operation that tracks relative-standing against a human rival also tracks relative-standing against a predator to be evaded, a prey target to be assessed, or a collaborator whose contribution is to be estimated. The Thinker's resolution of the Scalars to which this computation contributes commits posteriors about each encountered agent to the Belief Register, and those posteriors persist across encounters as positional residues that bias future Salience weighting through the Observer's read of the Belief Register (Chapter 4, Sections 4.4.7 and 4.4.8). The computation's directionality (that it presses upward and yields downward only under duress) is not posited at the level of the Observer but is inherited from Differential Satiation Dynamics (Axiom 6); a falling reading among conspecifics carries the sign of a reading against a predator, because a fall acquires the criteria that make an agent Prey (Chapter 10, Section 10.3.5): the survival function that M-escalation serves is satisfied by rising relative position, not by a fixed absolute level, which makes the upward direction structurally privileged at the level of the whole system even though the Observer's computation is itself directionally symmetric.