CITEC Central Lab Facilities
CITEC
British Flag   German Flag

Course on Joint Attention

During the CITEC Summer School 2012 I gave a course on Joint Attention in the frame of Stream 1 Development of Joint Action.

Modules 7-8: Joint Attention as a Building Block of Joint Action: A Practical Course using a Mobile Eye-Tracking System

Besides voice and gesture, the eye plays an important role in joint action. In a practical course, we will use a mobile eye-tracking system to record and analyze a short interaction sequence to demonstrate how to operate the eye-tracking system, what tools to use for the analysis and what eye movements can tell us about the interaction and the underlying cognitive processes. A special methodology followed in our research at the CITEC is the empirical-simulative loop: we observe human-human interactions to derive models that can be operationalized in artificial agents (robots or virtual humans), these agents – and by this the models – are in turn evaluated in human-robot interactions. This again provides us insights which may lead to modifications of our theory, new studies on human-human interaction and thus the cycle continues.
 
In this module, we will see, how complex models of interaction, here the establishing of joint attention via gaze, can be implemented in our virtual agent Max. These models can then be tested in an immersive virtual reality environment, where the human participants can be engaged in an interaction with a human-sized Max.
 
The module will be split into two parts, in the first we will present the general idea of the empirical-simulative loop and show how an operational model of joint attention can be realized with a virtual human. In the second part, we will visit the CAVE, the virtual reality environment, meet Max in person and give the implemented model a try-out. This scenario also includes real-time analysis of eye gaze by combining eye-tracking with motion capturing and a monitoring of all context objects

The slides from the course are available for download (about 160 MB).