knowledge elicitation and knowledge representation
natural language processing, statistical linguistics and
language learning
information retrieval systems, text summarization methods,
and adaptive information filtering
connectionist modeling, adaptive critic architectures,
machine learning, artificial intelligence and expert system
design
Bayesian and non-Bayesian methods of inference
cognitive psychology; human factors and human-computer
interface design
Computational Model
Since 1995, Mr. Thompson has lead a research
effort to develop a computational model that is based on, and
expands, a cognitive theory of naturalistic decision making.
The theory focuses on two levels of cognition, the reflexive
elaboration of interpretative mental models in a recognitional
system and a second-order, meta-cognitive system. Meta-cognition
is built on simple, but necessary, attention shifting processes
that permit focused inference and allow dynamic decomposition
of priors to reveal assumptions. More complex meta-cognitive
behaviors include critical thinking skills, by which uncertainty
is structured and understood, and by which we may more accurately
measure stakes.
Current Research
In current research Mr. Thompson builds on this
earlier work by suggesting an extension of the computational
model that combines a world model with a model of the agent's
own utility function -- a necessary step for developing a computational
planning system. The research is focused on the recognitional
inference mechanism (vs metacognition), and the inclusion of
utility is expected to result in a novel reflexive, knowledge-based,
adaptive control architecture. This research addresses a critical
issue -- developing adaptive control architectures for higher-order
cognitive processes. It is informed, and constrained, by data
from cognitive psychology, adaptive-control, neurobiology, knowledge
representation, non-bayesian inference, and computational complexity.
Human-Computer Interactions
Mr. Thompson has developed an extension of these
methods to inferential information retrieval using mental
models to motivate search and collaborative research
agents to manage information flow with the user. In that work,
he intends to explore the utility of mental models as a new
modality for Human-Computer Interactions, one made possible
by the existence of computational methods for reasoning with
mental models and uncertainty.
CTI Research Efforts
Mr. Thompson leads CTI research efforts in computational
modeling of cognitive processes, novel information retrieval
methods, adaptive control and language processing. Other recent
work includes the development of a tool for eliciting mental
models, methods for representing mental models in computational
models, and the development of an intelligent training environment
for critical thinking skills.
Leadership
Mr. Thompson coordinates and leads all CTI technical
design and implementation efforts.
National Science Foundation (1988-1994)
Working with the Director and the National Science
Board, Mr. Thompson developed systems for the National Science
Foundation, for coordinating the NSF long-range planning effort
and informing the Director's Policy Group. Mr. Thompson also
designed systems for improving management of inter-disciplinary
proposals, identifying scientific areas where research has provided
significant contributions to the nations welfare, providing
management information for technical and executive planning
and decision making throughout the agency and providing methods
for improving the content of material used to communicate outside
the Foundation.
