What a Promotion Threshold Is
A promotion threshold is the minimum score or confidence level a candidate must reach to advance from one evaluation stage to the next. The candidate may be a proposed mutation, an execution path, or a branch of a planning graph. The threshold is not a fixed constant: it is one of the deliberation parameters that the agent's affective state field modulates. Raising the threshold requires stronger evidence before a candidate advances; lowering it permits weaker candidates to advance for further evaluation. The promotion threshold is therefore a control surface through which the agent's accumulated execution experience changes how readily it commits to a line of reasoning.
In the cognition disclosure the affective state field is a seventh structural field added to the semantic agent schema, alongside the intent field, context block, memory field, policy reference field, mutation descriptor field, and lineage field. It encodes a structured modulation vector, not emotion in the phenomenological sense, and it modulates how the agent weighs alternatives, tolerates ambiguity, persists under partial failure, and escalates under constraint pressure. Promotion thresholds are the modulation target through which that disposition reaches candidate evaluation.
The Fields That Move the Threshold
The affective state field is organized as a structured modulation layer comprising named control fields, each encoding a distinct dimension of the agent's disposition. The fields that act on promotion thresholds are risk sensitivity, uncertainty sensitivity, and novelty appetite. Elevated risk sensitivity or elevated uncertainty sensitivity raises the promotion threshold, so that only candidates carrying stronger evidence advance. Suppressed risk sensitivity, or elevated novelty appetite, lowers the promotion threshold, permitting weaker candidates to advance for further evaluation.
Each named control field is represented as a tuple: a current magnitude within a defined range, a decay rate governing how rapidly the value returns toward a baseline absent reinforcing stimuli, a policy-defined ceiling and floor, and a timestamp of the most recent update. Because the threshold tracks these fields rather than a standalone parameter, the threshold's behavior is a direct consequence of the agent's modulation state at the moment of evaluation.
What Drives the Modulation
The control fields that move the threshold are themselves updated by a deterministic affective update function operating on structured observations drawn from the agent's execution environment. A sequence of execution outcomes matching a failure signature, for example repeated mutation rejections at the same validation gate, increases the agent's uncertainty sensitivity and risk sensitivity, which in turn raises the promotion threshold. Execution success patterns move the state toward a confidence disposition, reducing uncertainty sensitivity and increasing novelty appetite, which lowers the threshold. Novelty exposure, time pressure, competing objectives, and low reported confidence from an inference engine are additional observation classes that adjust these same fields.
The update function is deterministic: given the same agent state, the same observations, and the same policy configuration, it produces the same output. This makes the agent's affective evolution, and therefore the trajectory of its promotion thresholds, auditable and reproducible. Each update is recorded in the agent's lineage as a state mutation event, with the input observations and the resulting state change preserved.
Policy-Bounded Updates
Every update to a control field is a policy-bounded mutation, which means the resulting movement in the promotion threshold stays inside a governed envelope. The policy reference field specifies, for each named control field, range bounds clamping the value to a permissible minimum and maximum, rate limits capping the magnitude of change per update cycle, an admissible trigger set defining which observation types may drive the field, an update authority specification, and decay governance constraints.
The update sequence is: verify the observation type is admissible for the field, compute the raw update magnitude according to the field's update rule, clamp the magnitude to the rate limit, apply it to the current value, clamp the result to the range bounds, and record the complete transaction in lineage. The rate limit prevents a single extreme observation, such as a catastrophic failure, from producing a discontinuous jump in disposition and therefore a discontinuous jump in the promotion threshold. No single observation or sequence of observations can drive the threshold outside the policy-defined operating envelope.
Threshold Elevation Under Quarantine
When the system detects a volatility condition, one or more control fields oscillating rapidly at high magnitude, or composite deviation from baseline exceeding a policy-defined threshold, the agent is routed to an emotional quarantine state. Among the restrictions imposed in quarantine, the agent's promotion thresholds are elevated to their policy-defined maxima, preventing aggressive candidate advancement while the agent's disposition is unstable. Quarantine also suspends delegation authority, elevates mutation acceptance thresholds, and routes execution through an additional validation layer.
The quarantine condition is assessed by a volatility detector computing a composite metric from the recent update history of the control fields. The agent enters quarantine when the metric exceeds the quarantine threshold and is released when it falls below a recovery threshold set below the quarantine threshold, providing hysteresis that prevents oscillatory quarantine-release cycles. Quarantine does not suppress the affective state; the agent keeps processing observations and updating its fields. It restricts operational scope, with maximally elevated promotion thresholds as one of the levers, until the state stabilizes within governable bounds.
Temporal and Integrity Coupling
Promotion thresholds are also raised by two further cognitive domain fields that couple into the affective pathway. Under elevated temporal pressure, the temporal cognition field raises promotion thresholds within the confidence governor, requiring stronger evidence before candidate mutations advance to execution, reflecting the increased cost of error under time-constrained conditions. The temporal cognition field additionally produces negative-valence affective observations when temporal insufficiency is detected, which elevate risk sensitivity and escalation tendency through the affective modulation pathway and thereby raise the threshold indirectly as well.
Separately, the integrity field modulates promotion thresholds through cross-primitive integration with the confidence computation. The continuously evaluated deviation function influences candidate evaluation through integrity-modulated promotion thresholds: when the composite integrity score is degraded by recent deviation events or accumulated restoration gaps, candidate evaluation is gated more strictly, consistent with the reduced reliability of decision-making under compromised integrity. These couplings share the promotion threshold as a common control surface, so temporal pressure, affective disposition, and integrity state all converge on the same gate.
Separation From Governance
Modulating a promotion threshold changes how the agent deliberates; it does not change what the agent is permitted to do. The affective state field cannot create authority the agent does not possess, bypass policy constraints, override trust slope validation, or authorize execution that governance has denied. The governance gate evaluates execution admissibility from policy compliance, trust slope validation, and cryptographic provenance independently of the agent's affective state, and the affective state field is not an input to that gate. Even an agent whose disposition has lowered its promotion thresholds to admit weaker candidates still cannot advance any candidate that fails governance.
This separation is enforced structurally. Promotion thresholds adjust the stringency of an admissible process; they do not grant permissions, validate truth claims, relax policy bounds, or alter trust slope continuity requirements. Policy-imposed ceilings and floors on the control fields remain inviolable during every update cycle, which keeps the range over which the threshold can move bounded by policy at all times.
Disclosure Scope
The promotion threshold as a modulation target of the affective state field, the minimum score or confidence required for a candidate mutation, execution path, or planning graph branch to advance, together with its modulation by risk sensitivity, uncertainty sensitivity, and novelty appetite, the policy-bounded and rate-limited updates that move it, its elevation to policy maxima under emotional quarantine, and its coupling to the temporal cognition and integrity fields, 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 extends to embodiments in which the threshold is realized over different candidate types and evaluation stages, provided the threshold remains a policy-bounded modulation target driven by the named control fields rather than a parameter that creates authority, relaxes policy bounds, or bypasses the governance gate.
