Body reading
Automated facial expression coding
Face / FACS
Measures: Facial movement during the interaction, and, with heavy caveats, the valence of the reaction
GSR says there was a reaction, but not whether it was good or bad, and this method suggests the direction. Useful when the flow cannot be interrupted to ask, as in games, video and onboarding. But it is the most contested method on this page, and using it without knowing that is signing off on a conclusion the current literature does not support.
The technique films the face during the interaction and automatically classifies the muscle movements, with a vocabulary that comes from the Facial Action Coding System, which describes each isolated movement before any interpretation of what it means. Melcher BrandãoProposed, with Brazilian colleagues, a method for measuring emotion in the use of digital products.One reference in this work:2025Proposta metodológica para medir emoções em contextos de experiência de uso de produtos di…See in the bibliography → and colleagues include it, in 2025, among the instruments for measuring emotion in digital products precisely because it requires no verbalisation and does not interrupt the activity; and AlmeidaTested the reading of facial microexpressions in interface evaluation sessions, in one of the few Brazilian studies in the set.One reference in this work:2009A técnica de análise de micro expressões faciais (METT II-Short) na avaliação da qualidade…See in the bibliography → tested a variant in digital game sessions, in 2009, and even with filming limitations identified expressions associated with the moments of greatest engagement. A stretch in which many people move their face the same way deserves a second look, but what the reading delivers is where to look, not what the person felt. That is answered by the question or the test that comes next.
Three caveats change what can be concluded. The first is one of vocabulary: a microexpression, in the strict sense, is a movement lasting fractions of a second linked to the suppression of an emotion, and that is not what the tools on the market capture. Treating the two as synonyms inflates the promise. The second is substantive, and it is big. The review by Lisa Feldman BarrettLed the review that dismantles reading emotion from facial expression: the same face does not mean the same thing everywhere.One reference in this work:2019Emotional expressions reconsidered: challenges to inferring emotion from human facial move…See on Wikipedia ↗See in the bibliography → et al. (2019) gathered the available evidence and concluded that facial configuration does not allow emotional state to be inferred reliably, because the same emotion is expressed in different ways across people, cultures and situations, and the same face appears in different emotions. What the technique delivers with confidence is the record that there was facial movement, and when. The emotional label hung on it is interpretation. The third is one of sampling, and it lives inside the tool itself. Comparisons between commercial facial-reading services show systematic attribution of more negative emotion to Black faces, even controlling for how much the person smiles (RhueOne reference in this work:2018Racial influence on automated perceptions of emotionsSee in the bibliography →, 2018). For valence with more rigour, the laboratory alternative is facial electromyography, which measures the activity of the brow and smile muscles even when the face does not move visibly (CacioppoShowed that facial electromyography detects emotional reaction even when the face does not move enough for anyone to notice.One reference in this work:1986Electromyographic activity over facial muscle regions can differentiate the valence and in…See in the bibliography → et al., 1986).
SeesNon-intrusive and continuous: it locates the instant of the reaction without interrupting the task.
Does not seeThe face is not a thermometer of emotion: saying which emotion occurred does not hold up (Barrett et al., 2019).