Computational Modeling of Appraisal Theory of Emotion

The central research concern is the empirical test of the Component Process Model of emotion (CPM), which predicts that emotions are elicited and differentiated on the basis of appraisal processes. Specifically, sequential appraisal results produce

Information gathering operations (e.g., attention, orientation)
A general state of action readiness (e.g., physiological tuning for effort)
A set of motivational tendencies for potential action
Motor expression which is partly subserving communication

The changes produced by these determinants together with the representation of the original appraisal processes are centrally integrated and form emotional experience.

Research in the project focuses on the CPM predictions for the response patterning in the different components, in particular emotional expression in different modalities, facial, vocal, postural, and gestural.
For the latter purpose we are partly using a large corpus of acted emotional expressions (GEneva Multimodal Emotion Portrayals, GEMEP) which has been made available to the research community.
We distinguish the processes of the production of expression in distal expression parameters (acoustic vocal parameters, facial action units) and the inference and attribution process on the observer side (the use of proximal percepts of the distal cues). We have developed a new model, the dynamic Tripartite Emotion Expression and Perception model (TEEP) that guides the research program. The model is currently tested with GEMEP stimuli and computer-synthesized facial expression, using structural equation modelling.
We also address issues linked to emotional intelligence such as individual differences in the ability to express, recognize, and regulate emotions and have developed several ability tests of such competences.
Finally, we conduct computational modelling of appraisal theory in order to investigate by simulation the appraisal mechanisms underlying emotion elicitation. To achieve this, we use data-driven methods of statistical machine learning, as well as time-dependent simulation in artificial neural networks.