11 emotional states as computational modulators of agent behavior.
Emotions aren't decorative. They change how the agent thinks, decides, and acts.
Emotions exist in a 2D valence-arousal space:
- Valence: positive (approach) to negative (avoid)
- Arousal: calm (deliberate) to excited (reactive)
| Emotion | Valence | Arousal | Effect on Agent |
|---|---|---|---|
| Joy | + | + | Faster decisions, broader exploration |
| Trust | + | - | Cooperation, information sharing |
| Surprise | 0 | + | Attention spike, context reset |
| Anticipation | + | + | Planning, preparation |
| Fear | - | + | Rapid avoidance, deliberation suppression |
| Anger | - | + | Aggressive action, risk tolerance |
| Disgust | - | - | Rejection, avoidance |
| Sadness | - | - | Reduced activity, conservative choices |
| Curiosity | + | + | Increased exploration |
| Calm | + | - | Deep deliberation, thorough analysis |
| Frustration | - | + | Strategy switching, escalation |
- Emotional contagion: States spread to nearby agents via A2A messages
- Behavioral modulation: Emotion affects attention breadth, decision speed, and risk tolerance
- Emotional memory: Events tagged with emotional valence; emotional events persist longer in episodic memory
- Return to baseline: Emotions decay exponentially toward neutral state
- cuda-neurotransmitter -- Neurochemical foundation
- cuda-attention -- Arousal modulates attention scope
- cuda-deliberation -- Emotion affects deliberation depth
- cuda-memory-fabric -- Emotional valence strengthens encoding
- cuda-communication -- Emotional state in messages
- cuda-narrative -- Stories carry emotional arcs
MIT OR Apache-2.0