Emotional AI research terms, defined.

Action Unit (AU)AU

A numbered code in the FACS system representing the activation of one or more facial muscles. EchoDepth tracks 44 Action Units per video frame in real time. AU combinations produce distinct emotional signatures — for example AU6 (cheek raiser) combined with AU12 (lip corner puller) indicates genuine enjoyment rather than social smiling. AU4 (brow lowering) paired with AU24 (lip pressor) signals suppressed concern.

Articulation gap

The gap between what people actually feel and what they are able to accurately describe in words. Emotional responses — particularly complex, ambivalent, or professionally threatening ones — cannot be reliably articulated in survey text or interview speech. The articulation gap is a primary reason traditional research consistently underestimates negative affect. FACS-based emotional AI bypasses the articulation gap by measuring involuntary physiological signals directly.

Arousal

The calm-to-activated dimension of emotional state in the VAD model, scored 0.0 to 1.0. High arousal paired with negative valence (e.g. 0.8A / 0.2V) indicates acute stress or distress. High arousal with positive valence indicates enthusiasm or excitement. Low arousal with negative valence (e.g. 0.2A / 0.3V) indicates emotional disengagement or resignation — often more concerning in a culture context than high-arousal anger.

Culture Index

EchoDepth's blended composite score combining ECI (Employee Culture Index) and CXI (Customer Experience Index), scored 0–10. A declining Culture Index while CXI remains stable is one of the earliest predictors of future customer experience degradation — typically appearing 90–180 days before it becomes visible in NPS or retention data. The Culture Index Trends chart plots both signals over time, with event markers for leadership changes and policy decisions.

CXI (Customer Experience Index)CXI

EchoDepth's composite measure of external customer sentiment, scored 0–10. CXI is derived from customer feedback, review platforms, support interactions, and market signals. When charted alongside ECI, the gap between employee and customer sentiment reveals whether internal cultural strain is yet visible externally — and how long before it becomes so. Sustained ECI decline with stable CXI indicates approximately 3–6 months before external impact.

Dominance

The controlled-to-in-control dimension of emotional state in the VAD model, scored 0.0 to 1.0. Low dominance during a research session indicates that a participant feels passive, uncertain, or overwhelmed — a critical signal in pharma patient research when evaluating how patients emotionally receive treatment information. In culture surveys, sustained low dominance scores on leadership-related topics indicate that employees feel they lack agency over decisions affecting their work.

ECI (Employee Culture Index)ECI

EchoDepth's composite measure of internal workforce sentiment, scored 0–10. ECI is derived from emotional AI analysis of employee survey responses, pulse feedback, exit interview data, and free-text input. Unlike engagement scores derived from self-report, ECI reflects the genuine emotional health of the workforce — including masked dissatisfaction that voluntary survey scores do not surface. ECI trends are tracked against CXI to identify the lag between internal strain and external impact.

Emotional Risk Score

A 0–100 forecast metric predicting organisational trust, resistance, and credibility risk based on current emotional signals. EchoDepth generates a 90-day Emotional Risk forecast from the current ECI trajectory and contributing signal patterns. The forecast surfaces leading risk indicators — Leadership Credibility Drift, Change Fatigue, Comms Defensiveness — before they manifest as measurable disengagement or voluntary turnover. Confidence ratings (Low / Medium / High) are assigned based on data volume and signal consistency.

eNPS Proxy

An estimated Employee Net Promoter Score derived from emotional AI analysis of internal signals, without requiring a direct survey question. EchoDepth generates an eNPS Proxy score to provide a promoter/detractor estimate even when formal eNPS data is unavailable, infrequently collected, or subject to social desirability bias. The Proxy score is recalibrated as new survey data is ingested.

FACS (Facial Action Coding System)FACS

A scientifically validated taxonomy of human facial muscle movements developed by Paul Ekman and Wallace Friesen, first published in 1978 and subsequently validated across six countries and 14 cultural cohorts. FACS codes facial expressions into numbered Action Units, providing an objective, language-independent method for measuring emotional state from facial behaviour. Because facial Action Units are produced by involuntary muscle contractions, FACS analysis bypasses the social desirability bias and articulation gap that distort self-report research methods.

Masked dissatisfaction

A culture measurement risk signal where an employee expresses concern in professionally moderated language while the underlying emotional scoring shows significantly elevated negative affect — typically Disapproval, Contempt, or Distress. Masked dissatisfaction is more operationally dangerous than overt complaints because it is invisible to standard survey platforms and indicates that the employee has already concluded that direct negative feedback is professionally unsafe. EchoDepth surfaces masked dissatisfaction through text-emotion divergence analysis.

Social desirability bias

The tendency for research participants to give responses they believe are socially acceptable rather than responses that reflect their genuine views. Social desirability bias is the primary source of distortion in culture surveys, exit interviews, and moderated focus groups. It is amplified by proximity to management, in-group dynamics, and perceived career consequences. FACS-based emotional AI bypasses social desirability bias by measuring involuntary physiological signals that occur before any conscious moderation of response.

Text-emotion divergence

The measurable gap between the written content of a survey or interview response and its underlying emotional signature as detected by EchoDepth's language emotion model. When an employee writes positively or neutrally but the emotional scoring shows Disapproval, Contempt, or Distress — or writes diplomatically about a concern while secondary signals show Doubt and Disappointment — EchoDepth flags this as divergent and scores it as masked dissatisfaction. Text-emotion divergence is the most commercially significant signal that emotional AI adds over traditional survey analysis.

Trust Risk Register

An EchoDepth output that compares employer brand promises — drawn from careers pages, job posts, press releases, and public communications — against the emotional reality reported by employees in internal surveys, Glassdoor reviews, Blind posts, and other sources. Each promise is scored a Gap Risk percentage and rated Overpromised or Aligned. A high Gap Risk score indicates that the employer brand positioning is creating credibility exposure that will affect recruitment quality, retention rates, and public reputation.

VAD (Valence-Arousal-Dominance)VAD

A dimensional model of emotion that maps emotional states onto three continuous axes: Valence (positive to negative feeling, 0.0–1.0), Arousal (calm to activated, 0.0–1.0), and Dominance (feeling controlled to in control, 0.0–1.0). VAD scoring converts FACS Action Unit activations into numerical, comparable emotional output suitable for statistical analysis, segment comparison, and longitudinal tracking. A VAD score of V:0.72 / A:0.45 / D:0.68 indicates moderately positive emotional state with moderate activation and high sense of agency.

Valence

The positive-to-negative dimension of emotional state in the VAD model, scored 0.0 (strongly negative) to 1.0 (strongly positive). In culture measurement, EchoDepth computes the average negative valence score across all survey responses — the ratio of negative-to-positive valence is the primary headline metric of workforce emotional health. A ratio above 1.5:1 indicates systemic strain; above 1.8:1 indicates the pre-exit condition most associated with sudden retention crises.