FACS in workplace research: how AU6 and AU12 reveal genuine employee sentiment

The problem with workplace smiles

When a CEO presents a restructuring at an all-staff town hall and asks for questions, most employees smile. When a manager delivers critical feedback and checks in, the recipient usually nods and says something positive. When a culture survey asks employees to rate their experience, most responses skew more positive than the underlying emotional reality.

None of this is dishonesty. It is social management — the entirely rational adaptation of emotional expression to professional context. The problem is that it makes traditional sentiment measurement systematically inaccurate. Engagement scores, culture survey ratings, and even qualitative focus groups all capture the managed expression rather than the underlying state. FACS analysis — specifically the Action Unit pattern beneath a workplace smile — is the most reliable method available for distinguishing the two.

AU6 and AU12: the Duchenne distinction

Action Unit 12 is the lip corner puller — the zygomaticus major muscle that produces the characteristic upward curve of a smile. It can be activated voluntarily. Anyone can produce AU12 on demand: say "cheese", think of something pleasant, or simply decide to smile at someone. AU12 fires in both genuine and performed smiles.

Action Unit 6 is the cheek raiser — the orbicularis oculi, pars orbitalis, the muscle that surrounds the eye and produces the characteristic crow's feet and cheek raise of a full smile. AU6 does not fire on demand. It activates involuntarily with genuine positive affect. You cannot reliably produce it without actually feeling something positive. Trying to do so results in a noticeable asymmetry — the eyes don't match the mouth — that trained observers and FACS systems recognise as a performed smile.

This is the Duchenne smile, named after 19th-century neurologist Guillaume Duchenne de Boulogne who first demonstrated the distinction using electrical facial stimulation. When AU6 and AU12 activate together — a co-activation pattern — the result is the Duchenne smile: the involuntary marker of genuine enjoyment. When AU12 activates without AU6 — the social smile — it indicates a deliberate expression without the underlying genuine affect.

Why this matters in workplace contexts

The Duchenne distinction is analytically important in workplace research precisely because workplace settings are high-performance social contexts. Employees are motivated to manage their emotional expression — to appear engaged, supportive, and positive — regardless of their actual emotional state. This is especially true in hierarchical contexts (all-staff meetings, performance reviews, senior leadership interactions) and in survey contexts where responses feel consequential.

Traditional sentiment measurement captures the performed expression. FACS captures the underlying signal. The gap between them is what EchoDepth calls masked dissatisfaction — and it is the most commercially significant signal in people and culture measurement. An organisation with high engagement scores but low Duchenne smile rates during leadership communications is sitting on a retention and culture risk that its survey data is not surfacing.

In practical terms, this plays out in three workplace scenarios:

Town halls and leadership communications

During live leadership events — all-staff presentations, change announcements, Q&A sessions — FACS analysis via webcam or room camera captures the moment-by-moment emotional response of the audience. Not what they say in the post-event survey; what their faces actually do during the presentation.

The AU6 co-activation rate across the audience during different segments tells you which parts of the communication landed as genuine — and which produced polite smiling (AU12 without AU6). A CEO's announcement of a new strategy that produces sustained social smiling but no Duchenne responses is an early warning signal, regardless of what the feedback survey subsequently shows. Conversely, a difficult message delivered with clear logic and honesty may produce lower overall smiling but higher AU6 co-activation rates — genuine, if sobered, engagement — than a positive framing that employees receive sceptically.

Combined with VAD (Valence-Arousal-Dominance) scoring of the same footage, the picture becomes richer: which moments produced genuine positive affect (high V, AU6+AU12 co-activation), which produced anxiety (high A, low V, low D), and which produced the suppressed contempt or frustration signals (AU4+AU7, AU14+AU15) that indicate trust erosion.

Survey open-text and text-emotion divergence

When FACS is applied to text-based survey analysis — which is EchoDepth's core workplace application — the Duchenne principle translates into what EchoDepth calls the performed-genuine distinction in language. Just as AU12 without AU6 is a socially managed expression, diplomatically worded survey text that scores high on surface-level positivity but carries suppressed frustration or disillusionment in its emotional signature is the textual equivalent of a social smile.

EchoDepth's text analysis scores each open-text response across 53 emotional dimensions, derived from the same VAD framework that underpins FACS-based analysis. When the text says "I think management communication has been improving" but the emotional signature scores high on Disappointment, Doubt, and Scepticism, EchoDepth flags this as text-emotion divergence — the written equivalent of AU12 without AU6. The employee is performing positivity in their response while the underlying emotional state is negative.

This divergence is invisible to any sentiment analysis tool that only reads keywords and phrase patterns. It requires an emotional model — derived from the same principles that make AU6+AU12 distinguishable from AU12 alone — applied to the linguistic signal rather than the facial one.

The key FACS Action Units in workplace sentiment analysis

Beyond the Duchenne distinction, several AU combinations are particularly significant in workplace emotional analysis:

AU combination	Muscle groups	Workplace signal
AU6 + AU12	Orbicularis oculi + zygomaticus major	Genuine positive affect — Duchenne smile. Reliable marker of authentic engagement or satisfaction.
AU12 alone	Zygomaticus major only	Performed or social smile. Common in hierarchical interactions; elevated rate indicates managed affect.
AU1 + AU4 + AU15	Medial frontalis + corrugator + depressor anguli	Suppressed sadness or distress. Often seen during difficult change communications or critical feedback.
AU4 + AU7 + AU17	Corrugator + upper lid + mentalis	Controlled frustration or anger. Suppression is partial — the combination is detectable even when the employee believes they are presenting neutrally.
AU14 + AU15	Buccinator + depressor anguli	Contempt or disgust — the most significant trust erosion signal. Typically micro-expression duration (40–200ms), meaning employees are unaware of producing it.
AU46	Orbicularis oculi (winking portion)	Elevated blink rate — cognitive load or fatigue. In survey sessions or extended listening, may indicate information processing overload or attention strain.

Cultural calibration in workplace FACS

One important nuance in applying FACS to workplace research across diverse organisations is cultural display rules. Ekman's original research established the cross-cultural universality of basic emotional expressions at the muscular level — but subsequent research has identified meaningful variation in how those expressions are displayed in social contexts. Cultures differ in the degree to which emotional suppression is expected in professional settings, the threshold at which negative affect is expressed openly, and the contexts in which the Duchenne smile is withheld even with genuine positive affect (out of modesty or professional formality).

EchoDepth's models are trained across 14 cultural cohorts in 6 countries, with culturally calibrated baseline models that account for display rule variation. This does not eliminate cultural confounds — no system can — but it means the Duchenne distinction is calibrated against what is typical for that cultural context, rather than applied from a single-culture norm.

For organisations with diverse or global workforces, this calibration is not optional: applying a single-culture FACS baseline to a multi-cultural organisation will systematically misclassify emotional display patterns, producing false negatives (missing suppressed negative affect in high-suppression cultures) and false positives (reading cultural formality as disengagement) that undermine the analysis. See the full FACS methodology reference for detail on how EchoDepth handles this.

From FACS measurement to ECI scoring

In EchoDepth's workplace application, FACS-derived signals — whether from live session camera analysis or text-based emotional AI — feed into the Employee Culture Index (ECI): a composite 0–10 score reflecting genuine workforce emotional health. The ECI tracks alongside the Customer Experience Index (CXI), enabling organisations to see the lag correlation between internal emotional state and external performance.

The Duchenne distinction is embedded in ECI scoring: an organisation where AU12-without-AU6 is the dominant response pattern — performed satisfaction across the board — will score differently on ECI from one where AU6+AU12 co-activation is consistently high. The difference in ECI reflects the difference between a workforce that presents as engaged and one that genuinely is.

For the full methodology on measuring employee sentiment accurately, including how FACS principles apply to survey text analysis, see our primary guide. For the science of FACS analysis in research contexts, including VAD scoring and cultural calibration detail, see our FACS explainer.