What Is the VAD Model? Valence, Arousal and Dominance Explained
Sentiment analysis tells you positive or negative. The VAD model tells you how positive, how activated, and how in control. The difference between those two answers is the difference between a customer who is quietly unhappy and one who is about to escalate.
The VAD model (Valence-Arousal-Dominance) is a three-dimensional framework for mapping human emotional states, developed by James Russell and Albert Mehrabian (1977). It provides a continuous space where any emotional state can be located by its position on three independent axes: how positive or negative it is (Valence), how activated or calm it is (Arousal), and how dominant or submissive it is (Dominance). EchoDepth uses VAD scores as the foundation for all composite outputs including Trust Score.
Why three dimensions instead of one
The dominant paradigm in enterprise sentiment analysis is a single axis: positive versus negative. A customer contact is positive or negative. A social media post is positive or negative. An earnings call transcript is positive or negative. This framing has commercial value — it is measurable, scalable, and better than nothing.
The limitation is that a single axis cannot distinguish between emotional states that have completely different implications. Grief and boredom are both negative — but a customer expressing grief needs a very different response from a customer expressing mild disengagement. Excitement and anxiety are both high-arousal — but an excited customer is converting, while an anxious customer is at risk. Valence alone cannot make those distinctions. Arousal and Dominance are the dimensions that make enterprise-relevant emotional measurement possible.
Submit content for a free EchoDepth analysis — receive actual V, A and D scores for your specific context.
The three dimensions in detail
The emotional quality dimension. Negative valence (V < 0.5) indicates displeasure, discomfort, unhappiness. Positive valence (V > 0.5) indicates pleasure, comfort, satisfaction. Valence is what most sentiment analysis attempts to measure — but as a single continuous dimension rather than a binary positive/negative classification.
Example: A customer declining a renewal communicates low valence. An investor leaving a confident earnings call communicates high valence. A candidate performing well in an interview communicates rising valence.
The activation dimension. Low arousal (A < 0.5) indicates calmness, stillness, disengagement. High arousal (A > 0.5) indicates excitation, agitation, alertness. Arousal is what most sentiment tools completely miss — yet it is often the most diagnostically important dimension for enterprise use cases.
Example: A customer in financial distress shows high arousal even when their words are controlled. A resistant employee shows elevated arousal during transformation announcements. An investor losing confidence shows arousal rising as valence falls.
The control dimension. Low dominance (D < 0.5) indicates powerlessness, vulnerability, feeling controlled. High dominance (D > 0.5) indicates agency, confidence, assertiveness. Dominance is the dimension most relevant to vulnerability detection and to credibility assessment.
Example: A vulnerable customer shows low dominance — they feel the firm has power over them. A leader losing credibility shows dominance collapse mid-presentation. A candidate performing with genuine confidence shows consistently high dominance.
How EchoDepth uses VAD
EchoDepth generates per-frame VAD scores from facial Action Unit data and vocal pattern analysis. These scores are continuous — not categorical — meaning the output is a time series of emotional positions rather than a sequence of labelled emotional states. This matters because the trajectory is often more informative than the position: V dropping from 0.6 to 0.3 over 90 seconds while A rises from 0.4 to 0.8 is a different signal from a static V:0.3, A:0.6 baseline.
The composite outputs — Trust Score, Credibility Rating, Resistance Indicator — are derived from VAD patterns over time, not from any single moment. A trust score is a function of whether V and D maintain a consistent positive trajectory during delivery, whether A spikes at specific moments (suggesting stress), and whether the overall pattern is consistent with what a credible, confident communicator would produce. VAD provides the measurement substrate; the composite scores provide the enterprise-interpretable output.
Frequently Asked Questions
What does VAD stand for?
VAD stands for Valence, Arousal and Dominance. It is a three-dimensional model of emotional state developed by James Russell and Albert Mehrabian in 1977. Valence describes the positive or negative quality of an emotion. Arousal describes the activation level. Dominance describes the sense of control or power.
Why does EchoDepth use VAD instead of emotion labels?
Emotion labels like 'happy', 'angry' or 'fearful' are culturally variable, context-dependent, and difficult to validate objectively. The VAD model maps emotional state in a continuous three-dimensional space, allowing gradations and combinations that discrete labels miss. It also produces scores that can be tracked over time and processed algorithmically.
What is the difference between valence and sentiment?
Sentiment analysis typically produces a single positive/negative/neutral classification. Valence is a continuous 0–1 score on a single dimension of the full VAD model. Sentiment cannot capture arousal (how activated the emotion is) or dominance (how in control the person feels). EchoDepth measures all three dimensions independently.
See actual VAD scores from your content.
Submit a video, audio file or transcript. EchoDepth returns V, A and D scores with a full Trust Score analysis — free, within 5 working days.