AIM Framework glossary

Agent-Type Vector

Canonical definition

The output of the Agent-Type classifier (Axiom 9) for a given Individual-Object: a four-dimensional vector v = (C, R, X, Y), where C is Collaborator-weight, R is Rival-weight, X is Predator-weight, and Y is Prey-weight. The coordinates are continuous, real-valued, non-negative, and updated by the classifier on incoming observable evidence, both behavioural (what the Object has been observed to do: cooperate, contest, threaten, yield) and non-behavioural (physical attributes, group affiliation, proximity, apparent resource value or threat capacity, and other perceptible properties that the classifier's ancestral circuitry was selected to read directly from the Object's presentation). The X and Y axes update faster than the C and R axes: false-positive costs under threat and prey contexts were asymmetrically high in ancestral selection environments, and Source Opacity (Axiom 3) impedes error-correction in the slower social-relationship axes. The update runs in two limbs of the architecture's loop. The Observer's limb assigns the Individual-Object a salience weight on what its presentation supplies, within the Preconscious Transmission window; the Thinker's limb carries the same vector as an Agent-Type hypothesis in the Hypothesis Space over which Ground-Resolution operates, reaching the Observer as a prior through the Belief Register. On a first encounter no specific-Object vector exists, so the read is conditioned by the inherited population-level template and the Thinker's resolution of the resulting Scalar commits the first one; on later encounters the committed vector conditions the read before it is made. The rate asymmetry is the timescale separation between the limbs: presentation reaches the Observer's, which runs at the architecture's shortest time constant, and a behavioural record accumulates in the Thinker's, which runs at its longest. The vector is computed Individual-Object by Individual-Object, not at the population level. The moral consideration applied to an Individual-Object is the vector-weighted combination of the four basis-profile contents: C-profile carries cooperation-reciprocity and associated obligations; R-profile carries contest-reciprocity and fair-combat conventions; X-profile carries threat-response permissions and community protective obligations; Y-profile carries extraction-ethics with stewardship constraints and prohibitions on gratuitous cruelty. Because most Individual-Objects carry non-zero weight on more than one axis, the phenomenology of moral consideration is typically a weighted blend rather than the pure content of any single profile. The canonical formalism and the five formal properties of the vector are specified in Chapter 11, Section 11.13.1.

Explanation

When your brain classifies another person, it does not produce a single label. It produces a weighted blend across four profiles, Collaborator, Rival, Predator, Prey, and the moral duties and permissions you feel toward that person are the weighted mix of those four sets of content. Most people score non-zero on more than one axis, which is why moral experience is usually complex rather than pure. Threat signals (Predator-weight, Prey-weight) shift the blend quickly; alliance signals (Collaborator-weight, Rival-weight) shift it slowly, partly because Source Opacity makes social relationships harder to read accurately at speed.

Epistemic status

Framework-Specific Construct. Canonical formalism in Chapter 11.

Derivation

Chapter 11, Section 11.13.1.

Related terms