Commander's Learning Agent

From MilcordWiki

CLearn is a software agent that passively monitors user behavior and discovers the patterns in the background, and offers to assist the user in the real-time inference mode.
blog ...
publications ...
request information

Overview

CLearn is a software agent that passively monitors user behavior and discovers the patterns in the background, and offers to assist the user in the real-time inference mode.

Need

Current decision aids require the manual input of commander’s intent in the field. Unfortunately, capturing commander’s intent manually requires an impractical amount of time to learn the various facets of the commander’s job, and, furthermore, requires an unrealistic adaptation capability as the commander’s mission changes dynamically in the field. Hence, there is a need to automatically capture the commander’s current mission, bring contextual knowledge, and assign priorities to resources supporting the commander’s mission.

Approach

CLearn Architecture

Clearn sits between the user and application User Interface, passively monitors user behavior in the background, maps low level events onto semantically relevant events from using contextual metadata, applies machine learning to discover the patterns in the background, and offers to assist the user in the real-time inference mode.

Benefits

• Government:

  • Increased situation assessment in network centric computing environments
  • Reduced cognitive loading in operations centers

• Commercial:

  • Personalization services

Applications

• Military: Air and Space Operation Centers (AOC) workflow automation
• Civilian: Personalized content delivery in Web services
• Competitive Advantages:

  • Unlike commercial user interface automation tools that mimic user behavior, CLearn generalizes the abstraction intent of user behavior.

References

• Caglayan, A. The age of Assistants? milcord blog. August 25, 2010.
• Caglayan, A., Burke, D. and Eaton, G. (2008) “Commander's Learning Agent", Technical Report. AFRL-RI-NY-IF- 2007-4638. AFRL, Rome, NY.
• Caglayan, A. and Harrison, C. G., Agent Sourcebook, ISBN 0-471-15327-3, July 1997, John Wiley & Sons, Inc., NY.
• Caglayan, A., and M. Snorrason, J. Jacoby, J. Mazzu, R. Jones and K. Kumar, Learn Sesame - A Learning Agent Engine” in N. Jennings and B. Crabtree (Eds.) International Journal of Applied Artificial Intelligence, Vol. 11, No. 5, p. 393, 1997.