The HyperPlex consists of a many-dimensional virtual building inhabited by digital movie clips and other visual objects (photographs, text, graphics). The user can navigate around the building exploring the different rooms in each floor and moving from one floor to another through virtual doors, corridors, and elevators. Each part of the building is associated with a particular cluster of topics. For example, one set of rooms might concern a particular set of people, places, politics, and time; in addition, nearby rooms would share some common theme (e.g., adventure, special effects, fun, or romantic).
Figure: Movie clip competing for attention in the magic room
Movies live in this space and can appear at many different locations in the building according to the subject/information of their component clips (subsegments) (see fig. 1). Each movie clip appears in the form of a window on the screen showing a keyframe from the clip. Keyframe windows can move around in a manner that tries to reflect the ``personality'' of each clip (see [Lasseter1987,Johnson1995]); keyframe windows also react to the user's gestures and voice in a manner characteristic of the clip.
The behavior of each visual object is modeled using Blumberg's computational model of action-selection [Blumberg1994a]. In Blumberg's framework an agent's set of activities is organized as a loose hierarchy with the top of the hierarchy representing more general activities and the leaves representing more specific activities. Activities compete on every time step for the the control of the agent that engages in a single activity at a time. A movie's goal is to play itself and to compete with other movies to catch the user's attention. Movies form a community where movies associated with similar concepts collaborate with each other, whereas movies pertaining to distant concepts compete to play or to have a central position on the display screen.
The user can ``call a movie'', ``grab a movie'', ``play a movie'', ``play a movie again'', ``send a movie away'', ``send a movie to another user as a postcard'', ``take a movie'', ``ask more info about a movie'' (that comes in the form of text), ``stop a movie that is playing''---like using a smart gesture-driven VCR---or just let the movies organize themselves dynamically on the screen and play as a result of the interaction amongst themselves and the graphical world they are immersed in.