Self-Conscious Modeling


Kirstie L. Bellman, Ph.D.

The Aerospace Corporation, California





Self-conscious modeling is one of the 21st century challenges. Over the next century, a large diversity of models will underlie computer-based support for most of our critical human functions, including education, commerce, medicine, science, and defense systems.  Such will require “self-conscious modeling” in two senses.  The first emphasizes the need to raise our awareness of the features and the limitations underlying the representations and methods we use to build our models and simulations.  We will discuss the progress in developing methods to explicitly reason about the extent to which a representation or method appropriately matches the requirements of a modeling problem. 


Once we have the methods to match representations to appropriate problems and to explicitly analyze and compare representations, we then can introduce the second sense of self-consciousness, which is to allow the modeling systems to use such methods themselves.  The ability of a system to reason about its own abilities is part of a family of adaptive capabilities known as computational reflection.  We will briefly describe our work in reflective architectures as one example of implementing computationally reflective systems. 




  Dr. Kirstie L. Bellman returned to the Aerospace Corporation after four years at DARPA to start up a new bi-coastal research and development center, called the Aerospace Integration Sciences Center (AISC).  The center serves as a research and development capability for a number of DoD and government agencies.  AISC's focus is on the development of advanced system and model integration methods, new analytic techniques, and evaluation tools for assessing the impacts of new technologies.  Upon completion of her term at DARPA as a Program Manager for the Domain-Specific Software Architectures (DSSA) program, Prototech (rapid prototyping technology), projects in the Formal Foundations program, the large Computer-Aided Education and Training Initiative (CAETI), and several TRP programs, she received an award from the Office of the Secretary of Defense for excellence in her programs.  During her years at DARPA, she had the honor of working with Dr. Anita Jones, then DDR&E at OSD, with the Office of Science and Technology Policy in the White House, and a wide range of other government agencies. 


Dr. Bellman has over thirty-five years of academic, industrial, and consulting experience in both laboratory research and the development of models and information architectures for large military and government programs.  Her published research spans a wide range of topics in Cognitive Science, Neuroscience, and Computer Science.  In addition to playing a leading role in the development of programs in the error analysis and evaluation of Artificial Intelligence programs, her group did internationally recognized research in conceptual design environments, software integration and architectures, and 'enterprise evaluation'.  Fifteen years ago, she started the VEHICLES project, an environment for the conceptual design of space systems that incorporates both conventional and artificial intelligence methods.  Ten years ago, with Dr. Landauer, she started the Wrappings approach to system integration.  Eight years ago, with Dr. Landauer and others, she started extending the concept of Virtual Worlds to education, business and research environments.  Her recent work focuses on the use of domain specific languages and formally based architectural description languages to design and analyze information architectures.  In this work, she has also been developing methods for modeling organizational as well as technical aspects of complex systems.  With a number of academic partners, she is also developing new mathematical approaches to the analysis of Virtual Worlds containing collaborating humans, artificial agents, and heterogeneous representations, models and processing tools.  Lastly she has been working on reflective architectures that use models of themselves to manage their own resources and to reason about appropriate behavior.  Recently with both national and international partners, she has been applying the above experience and methods to theoretical work and experiments on emotional agents, cybermedicine applications, bio-computation, and "biologically-inspired" architectures and operating systems.  Besides her active research and management efforts, Dr. Bellman gives many invited talks every year and is a popular lecturer.