Synthetic Teammates – A View from the Trenches


Dr. Wayne Zachary

Over the last decade, a practical application of cognitive modeling has been intelligent synthetic teammates for virtual environments. Synthetic characters have been used as teammates in team-training systems, role-players in mission simulation systems, non-player characters in serious computer games, and even as synthetic instructors in simulation-based training environments. I have led cognitive modeling applications of this kind in a broad range of domains. My presentation will summarize some of the key issues encountered (both theoretical and engineering) and the lessons learned during the development of more than twenty synthetic teammate projects. Subjects I hope to cover include:

• The limitations of cognitive task analysis as a source of model structure and knowledge,
• The issue of level of granularity in architecture, time, knowledge, perceptual/motor interactions with the environment
• The subtle but pervasive effects of the virtual environment design and engineering
• The three Es – expertise, errors, and evaluation – and how cognitive model design interconnects them and
• Problems in scalability from the temporal, task, and team perspective, and
• The persistently challenges of language, and related “happy valley” issues.

Dr. Wayne Zachary is an internationally recognized leader in applied cognitive science, and specifically in the development and application of novel instructional models for computer-mediated learning, and the development and application of cognitive architectures. His degrees are a BA in Anthroplogy (Summa cum laude, Cleveland State University), MA Anthropology (Temple University), MS in Computer Science (Temple University), and Ph.D. in Cognitive Anthropology and Computer Science (Temple University). A cognitive anthropologist by training, he has lead numerous ethnographic studies of differing military, business, and non-western cultural settings. He developed the widely-used COGNET/iGEN cognitive architecture and led its application to numerous intelligent training applications for the Army, Navy, Air Force, and DARPA. He also co-developed the Intelligent Guided Practice Instructional Model, which integrates the use of synthetic instructors and scenario-based curricula to provide individualized tutoring with guided practice and objective-based pacing and student feedback. He created the SCOTT architecture, which adapted the intelligent guided practice instructional model to distributed and team training uses, and led the application of SCOTT to multiple command and control teamwork training systems. He advises the Defense Science Board and is on the editorial boards of the several journals, including Human Factors and Cognitive Engineering and Decision Making.