Affect as Cross-Primitive Input
by Nick Clark | Published March 27, 2026
Affective state serving as structured input to confidence computation and forecasting operations, creating a feedback loop where cumulative experience modulates willingness to execute and speculative planning.
What It Is
The affective state field serves as structured input to both the confidence computation function and the forecasting engine. Confidence receives affect as a modulator of execution sensitivity. Forecasting receives affect as a modulator of speculation parameters. This creates feedback loops where cumulative experience, encoded as affect, modulates both the willingness to execute and the breadth of speculative planning.
The coupling is read-only from the perspective of confidence and forecasting: they consume affective state but do not write to it.
Why It Matters
Without cross-primitive coupling, the cognitive subsystems operate in isolation. An agent with elevated caution from negative experiences would still speculate with default breadth and compute confidence with default sensitivity. Cross-primitive input ensures that experiential context propagates across all cognitive functions.
This coupling is what produces coherent behavioral profiles: an agent that is cautious is cautious in all aspects of its cognition, not just in one subsystem.
How It Works Structurally
The confidence computation function receives the affective field as an input parameter. Specific dimensions modulate the gain applied to capability, knowledge, and environmental inputs. The forecasting engine receives the affective field and uses it to modulate branching factor, pruning aggressiveness, and delegation preference.
The coupling weights are defined in the agent's policy reference, enabling domain-specific tuning of how strongly affect modulates each subsystem.
What It Enables
Coherent cognitive agents where experience in one domain propagates naturally to all evaluation processes. An agent that has learned caution through failed execution also plans more conservatively and computes lower confidence under similar conditions.
The cross-primitive feedback loops produce emergent behavioral patterns that resemble experienced judgment without requiring explicit behavioral rules for each pattern.
