Emotion as a computational primitive, not a simulation.
Existing affective AI models treat emotion as output to display. No system treats it as a persistent field that modulates deliberation. Named control fields with asymmetric update, exponential decay, and deterministic coupling to confidence, forecasting, and integrity.
Simulated emotion is not functional emotion
Every existing approach to affective computing treats emotion as something to recognize in humans or simulate in responses. Sentiment analysis classifies text. Affective generation produces emotionally-toned output. Neither approach gives the agent itself an emotional state that influences its reasoning, modulates its confidence, or shapes its planning.
Without functional affect, agents lack the computational machinery that emotion provides in biological cognition: urgency signals, risk aversion, approach-avoidance modulation, and the asymmetric weighting of losses versus gains. These are not luxuries. They are control signals that prevent catastrophic decisions under uncertainty.
Affective state in this architecture provides named, typed control fields — not metaphors. Each field has defined update dynamics (asymmetric: fast rise, slow decay), exponential temporal decay, and deterministic coupling to other cognitive subsystems. An agent's confidence computation, forecasting horizon, and integrity self-assessment are all modulated by affective state. The emotion is not displayed. It is computed, maintained, and functionally operative.
Control dynamics that mirror biological cognition
The architecture does not simulate human emotion. It provides the same functional role: a persistent, temporally-decaying set of control signals that modulate deliberation in real time. An agent that has recently encountered unexpected failure carries elevated caution fields that reduce its confidence, narrow its forecasting horizon, and increase its self-monitoring frequency. These effects are automatic, deterministic, and auditable.
This is not sentiment analysis. This is not empathetic response generation. This is a structural control primitive that makes autonomous agents behave with the adaptive caution and contextual sensitivity that biological cognition achieves through affect — without requiring consciousness, subjective experience, or any philosophical commitment about what emotion is.
Functional affect for autonomous agents. Published and available to license.
No guarantee of issuance or scope. No rights granted by this page. Any license requires issued claims (if any) and a separate written agreement.
