Category Construction Corollary

Canonical definition

A corollary to the Relative-Position Computation Principle stating that because the agent-type categories over which the Observer's position computation runs are Heuristics maintained by the Thinker rather than fixed by the architecture, they depart from objective agent-boundaries in two directions: outward over-extension, in which the Thinker classifies non-agentive systems as agents, or a system that acts without wanting as a person, and runs the position computation against them (the wind as a god, the market as a bear or bull, an artificial intelligence as a person), and inward over-partition, in which the Thinker subdivides a single agent class into ranked sub-categories against which positional computation differentiates (racial hierarchy, caste stratification). Both forms arise without new architectural machinery: outward over-extension is the same computation applied to a Thinker-constructed agent representation; inward over-partition is the same computation applied to Thinker-maintained sub-categories within a class. The departure is carried by Contactless Updating, because the Individual seldom contacts the Referent that settles an agent-boundary and holds the categories on the testimony of those who share them, and where a sub-category acquires positional value the Mimetic Drive to False Belief maintains it as well. The corollary does not require that the departure produces stable error; in some environments, treating the wind as an agent may be instrumentally correct. It predicts that wherever the Thinker's categorical maintenance diverges from objective agent-boundaries, the Observer's positional computation will track the Thinker's categories rather than correcting for their accuracy, and that the divergence persists unless the Individual holds the Referent prior to the category.