Analogical Processing

NU CS Dept.
Home
Research
CrossJam
Classes
Toolbox

CPOF SI Info

Command Post Laboratory
CPOF Proposal Summary
Multi-Dimensional Presentation
Nemesis
Sketch Understanding

DARPA CPOF SI Support Information

Co-PIs: Ken Forbus Brian M. Dennis

Analogical Processing for Command Post reasoning and visualization

Analogical processing plays a central role in the reasoning involved in creating, evaluating, and monitoring COAs:

  • Analogical processing plays a central role in the reasoning involved in creating, evaluating, and monitoring COAs. In creating COAs from a concept sketch, operation plans are reused, based on libraries and templates. In evaluating COAs, likely enemy responses are estimated based on knowledge of their doctrine and prior experience. In choosing a COA, alternatives are compared along dimensions established by the commander. In monitoring COAs, the changing situation description must be compared against the unfolding plan, in order to highlight divergences. Inferring enemy intent involves explaining their actions in terms of likely plans and goals, which depends in part on knowledge about how they operate.
  • Analogical processing also plays a central role in creating understandable explanations and visualizations. Analogical encoding helps organize perceptual and conceptual material, including identifying symmetries, detecting regularities in complex descriptions, and suggesting ways to resolve ambiguities that increase conceptual coherence. Generating concise explanations often involves comparison, to identify relevant similarities and highlight important differences. In comparing COAs, highlighting differences helps in decision-making. In explaining the potential impact of an unexpected event, comparing the event against expectations highlights potentially important consequences.
The deliverables for this task include software prototypes that use analogical processing to provide interactive generation, evaluation, comparison, and monitoring of COAs, inferring enemy intent, and generating new cases based on Command Post operations. This includes developing visualization techniques for comparisons that make it easier to compare COAs and to detect divergence of events from expectations.


This page was last built on 9/21/98; 1:30:16 AM by Brian M. Dennis

NU CS Dept. Northwestern University