Mechanism

Affect-modulated confidence sensitivity is a second-order modulation: the agent's affective state does not contribute a term directly to the confidence value, it changes the gain of the confidence computation function itself. The confidence evaluation function maps a structured input vector of agent state and task state into a confidence value and a confidence rate of change. Two opposing processes act on that value over time: confidence decay, which decreases the value as adverse conditions accumulate, and confidence recovery, which increases it as previously adverse conditions are ameliorated. Affect-modulated sensitivity adjusts how strongly the same objective inputs drive each of these processes.

Concretely, the affective state modulates the degree to which adverse inputs cause confidence to decay and the degree to which favorable inputs cause confidence to recover. The same objective adverse condition, a resource shortage of a given magnitude or a capability gap of a given severity, produces a larger or smaller confidence reduction depending on the agent's current affective state. The modulation is described in the specification in terms of gain on the decay pathway and gain on the recovery pathway, not in terms of any direct additive contribution of affect to the confidence value.

The Affective State as an Input

The affective state field is a structural field of the semantic agent schema. It is a deterministic, policy-bounded data structure that encodes a structured modulation vector derived from prior execution outcomes and environmental observations, not an emotion in the phenomenological sense. It is organized as a modulation layer of named control fields, each corresponding to a defined modulation axis with its own value range, update rules, and governance bounds. Among these named control fields are an uncertainty sensitivity field and a risk sensitivity field, which encode the agent's current responsiveness to epistemic uncertainty and its current weighting of potential negative outcomes.

The current value of the affective state field is one of the agent state inputs consumed by the confidence computation subsystem. In that capacity the affective state influences how the agent weighs uncertainty, tolerates partial information, and responds to adverse signals. It is this consumption of affective state by the confidence computation, rather than any separate pathway, that gives the affective state its purchase on the confidence value. Every mutation to the affective state field is recorded in the agent's lineage and subject to policy validation, in the same manner as every other agent field.

Anxiety and Risk Aversion: Amplified Decay

An agent whose affective state is characterized by elevated anxiety or risk aversion exhibits increased confidence sensitivity to adverse inputs. The same objective adverse condition produces a larger confidence reduction in such an agent than in a calm agent. The specification describes this as a higher gain on the decay pathway: adverse signals are amplified. Simultaneously the agent's confidence computation function has a lower gain on the recovery pathway, so favorable signals produce a smaller confidence increase.

The net effect is that an agent in this affective state decays faster and recovers more slowly. It is therefore more likely to reach the suspension threshold and less likely to recover from suspension without substantial improvement in objective conditions. The asymmetry follows from the underlying state, not from any change to the objective inputs themselves, which remain the same agent state and task state inputs the confidence evaluation function always consumes.

Engagement and Confidence: Attenuated Decay

An agent whose affective state is characterized by elevated engagement or confidence exhibits the opposite asymmetry: decreased confidence sensitivity to adverse inputs and increased confidence sensitivity to favorable inputs. The engaged agent's confidence computation function amplifies recovery signals and attenuates decay signals. The result is a confidence trajectory that is more resilient to transient adverse conditions and more responsive to improvements.

This asymmetry is bounded by policy constraints. The affective modulation of confidence sensitivity cannot reduce the gain on adverse signals below a configurable floor. Even a highly engaged agent therefore still responds appropriately to severe adverse conditions: the floor ensures that no affective state can make the agent indifferent to genuinely threatening signals. The bound is a policy-defined constraint on the modulation, consistent with the named control fields' general property of being clamped to a policy-defined ceiling and floor.

Interaction With Trajectory Analysis

Affect-modulated sensitivity interacts with the confidence governor's trajectory analysis. The governor performs differential rate analysis by comparing the confidence decay rate against the confidence recovery rate at each evaluation cycle, and it uses the resulting net rate of change to implement trajectory-based gating, which can suspend execution preemptively when confidence is decaying rapidly even while the absolute value remains above the authorization threshold.

Because affect changes the gain on the decay and recovery pathways, it changes the slope of the confidence trajectory those pathways produce. An agent whose affective state amplifies decay signals produces steeper negative confidence trajectories, which trigger preemptive suspension earlier. An agent whose affective state attenuates decay signals produces flatter negative trajectories, which extend the period of execution before suspension triggers. Affect does not alter the gate itself, it alters the trajectory the gate observes.

Distinction From Curiosity Modulation

The affective state field also includes a curiosity dimension, but curiosity acts on a different part of the pipeline. Curiosity modulates the agent's response to confidence interruption, biasing the agent toward inquiry and exploration rather than disengagement when confidence drops below the execution authorization threshold. An agent with elevated curiosity generates more hypotheses, seeks more information, and persists in the inquiry mode longer; an agent with suppressed curiosity performs minimal inquiry and waits.

Crucially, curiosity does not override the confidence governor's execution authorization gate. An agent with high curiosity that is in a suspended execution state remains suspended. Curiosity modulates the quality and intensity of the non-executing cognitive activity, not the gating decision itself: it influences what the agent does during suspension, not whether the agent is suspended. Affect-modulated sensitivity, by contrast, acts earlier, on the gain of the computation that produces the confidence value that determines whether suspension occurs at all.

Auditability of the Modulation

The affective modulation of confidence sensitivity is auditable. The governance infrastructure may audit the affective modulation parameters to verify that an agent's suspension timing was consistent with its affective state, and to confirm that no manipulation of affective state was used to inappropriately delay or accelerate suspension. This audit is possible because the affective state field, like the confidence field, participates in lineage recording: every mutation to the affective state is recorded, and the confidence trajectory is itself recorded as an auditable temporal record.

The separation of concerns described for the affective state field applies here as well. Affective state modulates how the agent thinks, not whether the agent is permitted to act. It cannot create authority, bypass policy constraints, or override the governance gate. Affect-modulated sensitivity therefore changes the dynamics of confidence computation within policy-bounded limits, but the execution authorization decision that the confidence governor enforces remains a hard gate that affect cannot reach around.

Disclosure Scope

This article describes affect-modulated confidence sensitivity as disclosed in the cognition filing (U.S. Application No. 19/647,395 and its international counterpart): the second-order modulation by which the agent's affective state changes the gain of the confidence computation function rather than contributing directly to the confidence value; the increased sensitivity to adverse inputs and reduced sensitivity to favorable inputs exhibited under elevated anxiety or risk aversion, producing faster decay and slower recovery; the converse asymmetry under elevated engagement or confidence, attenuating decay and amplifying recovery; the policy-defined floor below which the gain on adverse signals cannot be reduced; the interaction with the confidence governor's differential rate and trajectory-based gating, whereby modulated gain steepens or flattens the confidence trajectory and thereby advances or delays preemptive suspension; and the auditability of the affective modulation parameters against suspension timing.

The scope extends to embodiments in which the affective state is realized through any of the named control fields of the affective modulation layer, including the uncertainty sensitivity and risk sensitivity fields, and to deployments in which the modulation is bounded by the policy-defined ceilings and floors that govern those fields. It does not extend to mechanisms in which affect contributes a direct additive term to the confidence value, in which the modulation is unbounded by a policy floor, or in which the affective modulation alters the execution authorization gate itself rather than the gain of the computation that feeds it. Affect-modulated sensitivity is distinct from the curiosity modulation that governs the agent's behavior during suspension, which acts on inquiry intensity rather than on the gain of the confidence computation.