James Lester, and other members of the IntelliMedia Initiative at the Multimedia Laboratory of North Carolina State University, have two agents which are of interest to this discussion. The first is Herman the Bug, an impish bug-like creature that teaches children about biology, and the second is Cosmo, a frenetic advisor on internet protocols (see figure 11). Both projects share tremendous attention to detail in graphics, gesture logic, sequencing, and theoretical underpinnings. Herman the Bug has undergone rigorous testing, and Cosmo is headed for the same. The character of the two agents is quite different with Herman a likable clown, and Cosmo a hip and hyped internet cosmonaut.
The thrust of these efforts is rooted in both AI and education. The work seeks to clarify the issues raised when animated interface agents use gesture, speech, and guided locomotion to communicate important information (e.g., to focus a student's attention on an object in the world), but must also maintain believability in their characters' presence and actions. These two goals are not necessarily linked. Lester gives the examples of, on the one hand, a humorous, lifelike, joke-cracking, character that ultimately impedes problem solving through his distracting presence; and on the other, a dull assistant that always operates appropriately but yet fails to engage the student. When communications from an agent must be coordinated to be both engaging and purposeful issues in timing, and the multi-layering of actions arise. Like the USC-ISI, UPENN-HMS and NYU-IMPROV projects (although quite different in emphasis) the Lester group is pursuing deictic believability wherin the behavior planner for animated agents allows them to move through space, and refer to (sometimes dynamic) objects in its dialog with the user in a way that is both natural and unambiguous.
An important aspect of the Lester group's research is that it seeks to carefully document and test its findings about precisely what kinds of contributions lifelike agent capabilities make to learning. That the systems have the look and feel of computer games is a distant second to the core academic principles from which derive even the smallest gestures and action sequences in the systems. Intellimedia uses rigorous empirical testing to assess the contributions of agents to problem solving, higher-order learning, and the affective impact on students (e.g., see [Lester et al. 1997c], awarded Best Paper at AI-ED-97). Several extant studies discuss what might best be referred to as existence proofs that the intended influences on education are manifested by systems using these agents [Lester et al. 1997b, Lester et al. 1997a].