The gap

Every existing approach to affective computing treats emotion as something to recognize in humans or to simulate in responses. Sentiment analysis classifies text; affective generation produces emotionally-toned output. Neither 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, and no system carries them as a persistent field inside the agent.

The invention

Affective state is a set of named, typed control fields carried by the agent — not metaphors and not displayed output. Each field has defined update dynamics that are asymmetric, with fast rise and slow decay, and undergoes exponential temporal decay. The fields are deterministically coupled to other cognitive subsystems, so an agent's confidence computation, forecasting horizon, and integrity self-assessment are all modulated by its current affective state.

The result is a persistent, temporally-decaying set of 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 raise its self-monitoring frequency. The emotion is computed, maintained, and functionally operative — and every effect it produces is deterministic and auditable.

The inventive step

The departure from prior art is that affect is a control primitive internal to the agent rather than a recognition task or a presentation layer. Existing work measures emotion in humans or renders an emotional tone in responses; here, named affective fields are first-class state with their own asymmetric update rules and decay curves, and they modulate the agent’s own deliberation.

What makes this non-obvious is the deterministic coupling. Because affective fields feed confidence, forecasting, and integrity self-assessment through defined dynamics, the agent achieves the adaptive caution and contextual sensitivity that biology gets from emotion — without requiring consciousness, subjective experience, or any philosophical commitment about what emotion is. The behavior follows structurally from the fields, not from an after-the-fact heuristic.

Alone, and in composition

On its own, affective state is a control layer for any autonomous agent that must act under uncertainty: a typed, decaying set of caution and urgency signals that make an agent behave with contextual sensitivity instead of uniform confidence. It serves companion, therapeutic, customer-service, and care applications that need adaptive emotional consistency rather than simulated sentiment.

In composition, it is a cognitive subsystem the rest of the architecture reads from. Confidence governance, forecasting, and integrity self-assessment all take affective state as input, so a single persistent field shapes how the wider platform weighs risk and decides what to act on. It is the primitive that gives the other cognition layers their adaptive caution.