Affect-Modulated Inference Admissibility

Mechanism

Affect-modulated inference admissibility is the mechanism by which the semantic admissibility gate that governs inference-time semantic execution is modulated by the affective state of the invoking semantic agent. The admissibility gate is the central governance mechanism of the inference-time semantic execution substrate: it receives each proposed semantic mutation, evaluates it against the current semantic state object, and produces a deterministic determination of admit, reject, or decompose. Affect modulation does not change that gate's structure and does not change its outcomes vocabulary. It adjusts the quantitative parameters within which the gate operates, tuning evaluation stringency in response to the agent's current dispositional orientation.

The defining constraint is that the affective state does not override the admissibility gate's deterministic governance criteria. The policy constraint evaluation, the mutation descriptor validation, the lineage continuity validation, and the entropy bounds evaluation all continue to apply with their normal authority. Affect modulates the bounds those stages enforce; it cannot make the gate admit a mutation that policy forbids, and it cannot make the gate reject a mutation that policy permits at the unmodulated setting. The mechanism adjusts stringency within the governance envelope. It does not replace the deterministic floor.

The Affective State It Reads

The modulation reads the affective state field of the invoking agent, the structured seventh field of the semantic agent schema. That field is organized as a layer of named control fields, each encoding a distinct dimension of the agent's current dispositional orientation. The dimensions relevant to inference admissibility are the agent's uncertainty sensitivity, encoding the agent's current responsiveness to epistemic uncertainty, and its risk sensitivity, encoding the agent's current weighting of potential negative outcomes relative to potential positive outcomes when evaluating candidate mutations.

These dimensions are not produced by the inference engine and are not introspective guesses. They are maintained by a deterministic affective update function that operates on structured observations derived from the agent's execution environment, such as repeated failure patterns, novel environmental conditions, and reported low confidence or high entropy in prior inference outputs. Because the affective state is updated deterministically and carried as a typed field, the modulation it drives is reproducible: the same observations produce the same affective state, which produces the same modulation of the gate.

The Parameters It Modulates

The modulation operates on specific, enumerated parameters of the admissibility gate, not on the gate's acceptance criteria. The entropy and uncertainty bounds field of the semantic state object is modulated by the agent's uncertainty sensitivity and risk sensitivity dimensions. When uncertainty sensitivity is elevated, following repeated failure patterns, novel environmental conditions, or low-confidence outputs, the entropy bounds are tightened, requiring lower semantic uncertainty for a mutation to be admitted. When risk sensitivity is elevated, the lineage continuity threshold is also raised, so that a candidate mutation must sit more closely on the established semantic trajectory to pass lineage continuity validation.

The modulation is bidirectional. When the agent's affective state reflects a high-confidence disposition, the entropy bounds may be relaxed within the policy-defined ceiling, permitting broader exploration of candidate transitions. The policy-defined ceiling is the structural limit: affect can move the operating point within the band the policy permits, but it cannot move the bound past the ceiling the policy reference field establishes. Tightening and relaxing both occur inside that envelope.

Deterministic Modulation, Not Non-Determinism

The behaviorally significant result of the mechanism is that the same inference query, submitted by the same agent under the same policy constraints, may yield different admissibility determinations depending on the agent's affective state. An agent in a high-anxiety affective state produces more conservative inference outputs, outputs that stay closer to the established trajectory, introduce less novel content, and assert fewer uncertain claims, than the same agent in a high-confidence state.

This variation is not non-determinism. It is deterministic modulation within governance bounds. Given the same affective state, the same semantic state object, and the same candidate transition, the determination is identical. The affective state is one more typed input to a deterministic evaluation, recorded alongside the other inputs, rather than a stochastic perturbation of the gate.

Composition with the Gate's Stages

Because affect modulates the entropy bounds field and the lineage continuity threshold rather than introducing a separate stage, it composes with the admissibility gate's existing four-stage evaluation without adding a parallel governance path. A candidate mutation still passes through policy constraint evaluation, mutation descriptor validation, lineage continuity validation, and entropy bounds evaluation in sequence. Affect changes only the numeric bounds the third and fourth of those stages apply. Policy constraint evaluation and descriptor validation are unaffected, which is what preserves the deterministic floor.

The confidence-gating mechanism that transitions inference from executing mode to non-executing inquiry mode also reads the affective state: its admission-rate threshold may be modulated by the same affective state, so that an agent in a high-anxiety state transitions to inquiry mode more readily, while an agent in a high-confidence state permits a lower admission rate before transition. Affect-modulated admissibility and affect-modulated confidence gating therefore share the same affective input and reinforce the agent's overall posture: a stressed agent both tightens what it will admit and steps back into inquiry sooner.

Separation from Authority and Truth

The mechanism maintains the strict separation between the affective modulation layer and the governance infrastructure. Affective state cannot grant permissions: an agent's affective disposition cannot make a mutation that violates a policy constraint admissible. Affective state cannot validate or invalidate factual claims: a suppressed uncertainty sensitivity does not cause uncertain information to be treated as verified, because anchored semantic resolution and the policy stages still apply. Affect adjusts how stringently the gate inspects a candidate within the permitted band. It does not confer authority the agent lacks and it does not stand in for the gate's deterministic checks on policy, descriptor consistency, and reference resolution.

Prior-Art Distinction

Emotion-aware dialog and affect-conditioned generation are established techniques. This mechanism is not a claim over sentiment scoring or over routing inputs to different model heads. It is a claim over a deterministic admissibility gate, structurally interposed within the inference loop, whose entropy bounds and lineage continuity threshold, but not its acceptance criteria, are modulated by the agent's typed affective state. The modulation preserves the deterministic governance floor, keeps policy and descriptor validation outside the affective layer's reach, and records the affective state as a reproducible input to each determination. This distinguishes the mechanism from approaches that let affect mutate policy at runtime in ways an auditor cannot reconstruct, or that apply affect only to surface generation after the inference trajectory has already been committed.

Disclosure Scope

Affect-modulated inference admissibility, comprising the modulation of the inference-time semantic admissibility gate's entropy and uncertainty bounds by the agent's uncertainty sensitivity and risk sensitivity dimensions, the raising of the lineage continuity threshold under elevated risk sensitivity, the relaxation of entropy bounds within the policy-defined ceiling under a high-confidence disposition, and the property that modulation is deterministic rather than non-deterministic given a fixed affective state, semantic state object, and candidate transition, is disclosed in the cognition filing (U.S. Application No. 19/647,395 and its international counterpart). This article describes that disclosed mechanism. The scope is independent of the specific affective update function employed, provided it yields the named control fields deterministically, and it extends to the shared use of the affective state by the confidence-gating mechanism. Implementations in which affect overrides the deterministic governance criteria, grants authority the agent lacks, or validates factual claims fall outside the mechanism as disclosed; implementations that confine affect to modulating the gate's quantitative bounds within the policy ceiling are within scope.