User behavior transformation through dynamic input mappings

Games increasingly use input devices – such as Kinect, Leap, and Move – that map the location of the player’s body into a virtual space. Designers are often faced with two options: an intuitive, direct mapping that is initially easier to learn but not suited for all tasks, or an advanced mapping that is better suited to carrying out tasks but harder to learn. In this paper we propose a method for achieving the best of both worlds: user behavior transformation. Rather than a single static mapping, our method uses a dynamic input mapping that starts out with a direct mapping but changes over time to a more advanced mapping that is more suited to the types of tasks the player needs to perform. As the mapping changes, the player automatically transforms their behavior to take advantage of the new input space. In this way the mapping can allow for more advanced interactions the designer would like, while still being intuitive. We validate this approach with a Kinect based interface for a game with a complex task: protein folding in Foldit. The initial direct mapping from the player’s hand position changes to a mapping that reduces occlusions caused by overlapping hands. In a user study, the dynamic mapping compares favorably to two static mappings with respect to task completion times.