Emotional AI: PAD Affective State Simulator

Stimulus (P·A·D)

Pleasure +0.60

Arousal +0.50

Dominance +0.20

Decay half-life 2.5s

Agent State

Neutral

baseline mood

0.00

Intensity |E|

0.00

Nearest octant dist.

0.00

How it works

An affective agent's mood is modelled as a point E = (P, A, D) in the Pleasure–Arousal–Dominance space (Mehrabian & Russell). Between stimuli it relaxes back to a neutral baseline E₀ = (0,0,0) with first-order exponential decay:

dE/dt = -k · (E - E₀),   k = ln(2) / τ½

where τ½ is the decay half-life you set below. A stimulus event instead jumps the state partway toward the target vector (P,A,D) you dial in, blending 55% of the target in on each click — this is the same "impulse then relax" pattern used by affect-simulation architectures such as WASABI and ALMA in games and social robots.

Every emotion word is just a region of this cube. Mehrabian defined eight temperament octants from the sign of each axis:

Exuberant (+P+A+D)   Dependent (+P+A-D)
Relaxed   (+P-A+D)   Docile    (+P-A-D)
Hostile   (-P+A+D)   Anxious   (-P+A-D)
Disdainful(-P-A+D)   Bored     (-P-A-D)

Sliders — set the (P, A, D) target of the next stimulus.
Apply Stimulus — pushes the agent's state toward that target; the marker leaves a fading trail as it moves.
Decay half-life — how quickly the mood forgets the stimulus and returns to neutral; short = volatile agent, long = persistent mood.
Nearest octant — the mood label is whichever of the eight reference points is closest in Euclidean PAD distance to the live state.

Real-world relevance: affective-computing systems (companion robots, game NPCs, driver-monitoring AI) track a continuous internal state like this instead of a single emotion tag, because moods blend and fade the way this vector does.

Agent state E(t) Octant reference points

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.