Automated facial analysis using Valence Recorder: When recognition drives emotion
Résumé
This paper focuses on a new data source available to marketing researchers, ValenceRecorder, a software for automated facial analysis recently developed by GfK Verein. It uses a parallel approach: on the basis of specific features of a respondent's face, a first classifier has been trained to identify positive expressions while another classifier has been trained to identify negative emotions. Subtracting the negative emotion score from the positive emotion score produces an overall valence score, at a high frequency (15 frames per second). We analyze how the "net valence" (mean overall valence during exposure minus "baseline" valence just before exposure) of the emotion triggered by a positive stimulus (a photograph of a male or female movie star) is higher when the star is recognized. Further, stronger emotions during exposure lead to a more accurate recall of having seen the stimulus.