EMPAT: Emphatic agents for virtual job interviews


Project start: 01.03.2015
Duration: 3 years
Funded by: BMBF (Bundesministerium für Bildung und Forschung)
Scientifically responsible: Prof. Dr. Elisabeth André
Involved researchers at our lab:  Dr. Tobias Baur
Patrick Gebhard, DFKI

About the project


EmpaT is developing a virtual empathic training system which will help users to practice application situations and to better assess their own behavior.



Job interviews often present a great challenge. Presenting yourself successfully at the interview, correctly, interested and motivated, and appearing authentic and binding, is an unusual situation, especially for young applicants, who have not received sufficient training before. Virtual empathic training systems can create additional opportunities to practice application situations and to learn to better assess one's own behaviour.

In the EmpaT project, a real-time analysis of social signals is coupled for the first time with an emotional real-time user model in order to adapt the behavior of an interactive avatar to the socio-emotional situation of a user. This mechanism makes it possible to design the application dialogue in a sensitive way. Such dialogue systems can become personal assistance systems in different situations. A challenge here is the realistic design of the emotional feedback by the avatar.

For this purpose, a project consortium of computer scientists, psychologists and 3D graphics experts researches and realizes the interpretation of social and emotional signals that are necessary for empathy and adaptation to the respective user and his or her social situation. Highly developed hardware sensors are used. These record social communication signals such as eye, eye, hand and body movements. The real-time processing of the information thus obtained is used for adaptive-reactive control of the EmpaT training avatar. This approach allows for coordinated communication between user and avatar. In this way, various emotional-social interaction patterns can be simulated. The results of the interpretation of social and emotional signals are also used to calculate possible emotional states of a user. For this purpose, real-time capable emotion simulation and adaptive behavioral models are used, which integrate current findings from psychology, cognitive science and artificial intelligence.

The project is accompanied in conception and implementation by an advisory board, which supports the project in ethical, social and legal questions. The Ethics Advisory Board is composed of experts in the fields of coaching, youth work, law and pedagogy. In continuous acceptance and evaluation studies, the effect of personal interaction with a virtual avatar is examined in a realistic simulation environment. The committee also focuses on possible long-term social effects of virtual empathic training systems.

EmpaT is funded by the Federal Ministry of Education and Research (BMBF) as part of the research program "Socially and emotionally sensitive systems for optimized human-technology interaction - from technical tool to interactive companion (InterEmotio)". The project management agency is the VDIVDE-IT GmbH.