Although it is well-known that bodily distance and orientation, also called proxemic behaviors, are key factors in human-human communication, many people cannot consciously monitor and optimize their behavior while interacting. We propose a serious game where users can train to optimize their proxemic behavior in interactions with a virtual human. We chose a first-encounter flirt scenario where proxemic factors play a prominent role in the outcome of the game. We built an interactive installation with a large screen and Kinect sensor where the user has to complete various tasks by using body actions alone (movement, gesture, dance etc.). The installation manages to raise awareness about proxemics while being fun to play and watch.
For our prototype, we decided to use a female virtual agent as the subject of the flirt scenario. The game consists, in our current prototype, of a sequence of tasks that the player has to fulfill: get her attention, approach her, show interest, dance with her etc. For each task, a score has to be reached for completion within a limited amount of time. After each task, the overall score is updated which ultimately determines the final outcome of the game.
Task completion depends on the user’s body actions alone. The following inputs are used to evaluate task completion:
Depending on the task, certain actions are evaluated against. For instance, during the task “dance” the activity of the hands and the hip are monitored while music is playing in the background. The measured activity is transformed to a task completion score. During the task “approach her” the optimal distance to the screen (note too close, not to far) has to be assumed.
In addition to the concrete task at hand a general rule system is running in the background at all times. This scheme evaluates the current level of interest expressed by the player’s body pose and movement. To trigger this and other behavioral rules in a consistent way, they are prioritized according to their current importance for the flirt task.