From Personal Computers - PowerPoint PPT Presentation

About This Presentation
Title:

From Personal Computers

Description:

Personal Cognitive Assistant for Intelligence Analysis ... Edward Feigenbaum, 1993: Rarely does a technology arise that offers such a wide ... – PowerPoint PPT presentation

Number of Views:60
Avg rating:3.0/5.0
Slides: 61
Provided by: learningag
Learn more at: http://lac.gmu.edu
Category:

less

Transcript and Presenter's Notes

Title: From Personal Computers


1
From Personal Computers to Learning Assistants
Gheorghe Tecuci Learning Agents Center and
Computer Science Department School of Information
Technology and Engineering George Mason University
21 September 2005
2
Overview
Learning Agents Center Research Vision
Research Issues for Learning Agents
Personal Cognitive Assistant for Intelligence
Analysis
Agents for Centers of Gravity and Critical
Vulnerabilities
Virtual Experts for Multi-domain Collaborative
Planning
Agent for Course of Action Critiquing
Final Remarks
3
http//lac.gmu.edu
Mission
  • Conducts fundamental and experimental research on
    the development of knowledge-based learning and
    problem solving agents.
  • Supports teaching in the areas of intelligent
    agents, machine learning, knowledge acquisition,
    artificial intelligence and its applications.
  • Develops the Disciple theory, methodology and
    agent shells for building agents that can be
    taught how to solve problems by subject matter
    experts.

Basic Research
Tools
Applications
Transitions
4
Teaching as Alternative to Programming
Building an intelligent machine by programming is
too difficult. Instead of trying to produce a
programme to simulate the adult mind, why not
rather try to produce one which simulates the
child's? If this were then subjected to an
appropriate course of education one would obtain
the adult brain.
Turing, A.M. (1950). Computing machinery and
intelligence. Mind, 59, 433-460.
5
How are Expert Systems Built and Why it is Hard
Edward Feigenbaum, 1993 Rarely does a technology
arise that offers such a wide range of important
benefits.
6
Research Problem and Approach
7
Disciples Vision on the Future of Software
Development
Learning Agents
Personal Computers
Mainframe Computers
8
Vision on the Use of Disciple in Education
teaches
? 2005, Learning Agents Center
9
Overview
Learning Agents Center Research Vision
Research Issues for Learning Agents
Personal Cognitive Assistant for Intelligence
Analysis
Agents for Centers of Gravity and Critical
Vulnerabilities
Virtual Experts for Multi-domain Collaborative
Planning
Agent for Course of Action Critiquing
Final Remarks
10
The Overall Architecture of a Disciple Agent
? 2005, Learning Agents Center
11
Knowledge Base Ontology Rules
PROBLEM SOLVING TASK
ONTOLOGY FRAGMENT
Determine whether John Smith can be a PhD advisor
for Tom Even in Artificial Intelligence.

IF Determine whether ?O1 can be a PhD advisor
for ?O2 in ?O3.
Main condition ?O1 is PhD_advisor has_as_employer
?O4 has_as_position ?O5 ?O2 is PhD_student ?O
3 is research_area ?O4 is university ?O5 is tenure
d_position Except when condition ?O1 is person is
_likely_to_move_to ?O6 ?O6 is employer
THEN Determine whether ?O1 would be a good PhD
advisor for ?O2 in ?O3.
REASONING RULE
12
Main Idea of the Disciple Approach
KE
Model the reasoning of SME
Create object ontology
Define reasoning rules
Verify and update rules
SME
Traditionally
Determine whether John Smith can be a PhD advisor
for Tom Even in Artificial Intelligence.
13
Research Issues for Learning Agents
Problem Solving Paradigm for Expert-Agent
Collaboration
Learning with an Evolving Representation Language
Plausible Reasoning with Partially Learned
Knowledge
Integrated Teaching and Learning
Multistrategy Learning
Agent Architecture for Generality-Power Tradeoff
Knowledge Base Structuring for Knowledge Reuse
14
Problem Solving Paradigm for Expert-Agent
Collaboration
Task reduction and solution composition guided by
questions and answers

T1
S1
Q1

S11
A1n
A11
S1n
T1n
S11a
T11a
S11b
T11b

S11b
Q11b

S11b1
S11bm
A11bm
A11b1

T11bm
T11b1
15
Learning with an Evolving Representation Language
16
Plausible Reasoning with Partially Learned
Knowledge

IF lttaskgt
Plausible Upper Bound ConditionltPUB conditiongt
Plausible Lower Bound ConditionltPLB conditiongt
THEN ltsubtask 1gt ltsubtask mgt
17
Mixed-Initiative Problem Solving
Problem
Creative solutions
Inventive solutions
Innovative solutions
Routine solutions
Solution
18
Integrated Teaching and Learning
examples, facts, rules
Input knowledge
learning hints
Explicit learning guidance
classification of examples, problem solutions
Problem solving behavior
Explicit teaching guidance
questions
19
Rule Learning Method
Analogy and Hint Guided Explanation
Analogy-based Generalization
Plausible version space rule
plausible explanations
PUB
guidance, hints
Example of a task reduction step
PLB
Incomplete explanation
analogy
Knowledge Base
20
Multistrategy Learning
21
Overview
Learning Agents Center Research Vision
Research Issues for Learning Agents
Personal Cognitive Assistant for Intelligence
Analysis
Agents for Centers of Gravity and Critical
Vulnerabilities
Virtual Experts for Multi-domain Collaborative
Planning
Agent for Course of Action Critiquing
Final Remarks
22
Challenges for the Intelligence Analyst
Difficult to share intelligence
Difficult to collaborate with other analysts and
experts
Overwhelmed by information
Intelligence analysis is very difficult
Difficult to consider multiple hypotheses
Difficult to train new analysts
Difficult to rigurously explain the analysis
Difficult to find time for critical analysis and
AARs
Difficult to analyze in reference to the culture
of the data source
Difficult to acquire and retain expertise
Difficult to avoid the analytic mindset
23
Investigated Solution
An integrated approach to intelligence analysis
research, education, and operations.
  • Develop a new type of intelligent agent that
  • can rapidly acquire expertise in intelligence
    analysis,
  • can train new intelligence analysts, and
  • can assist the analysts to solve complex
    problems.

24
Vision Integration of Research, Education, and
Operations
Knowledge engineer
Agent Lifecycle
25
Vision Use of Disciple-LTA Agents in an
Operational Environment
Disciple Client
Disciple-LTA
GLOBAL KNOWLEDGE BASE
SEARCH ENGINES
Libraries Knowledge Repositories Massive Databases
Disciple-LTA
Intelligent agent
Disciple-LTA
26
Synergistic Integration of Research and Education
Develop a systematic approach to military
intelligence analysis
Experimentation with Disciple-LTA in the 589 MAAI
elective
Military Education Practice
Military Research
Working closely with the expert analysts in a
multi-disciplinary research
Working closely with the end user to receive
crucial and timely feedback
DiscipleLTA
IntelligentAgents Research
Agent development by expert analysts using
learning agent technology
? 2005, Learning Agents Center
27
Live Experiment
28
Assess whether Location-A is a training base for
terrorist operations
29
Assess whether Location-A is a training base for
terrorist operations
What type of factors should be considered to
assess the presence of a terrorist training base?
30
What type of factors should be considered to
assess the presence of a terrorist training base?
Political environment, physical structures, flow
of suspected terrorists, weapons and weapons
technology, other suspected bases in the region,
and terrorist sympathetic population
31
Political environment, physical structures, flow
of suspected terrorists, weapons and weapons
technology, other suspected bases in the region,
and terrorist sympathetic population
Assess whether there is a flow of suspected
terrorists in the region of Location-A
Assess whether there are other suspected bases
for terrorist operations in the region of
Location-A
Assess whether the political environment would
support a training base for terrorist operations
at Location-A
Assess whether there is terrorist sympathetic
population in the region of Location-A
Assess whether the physical structures at
Location-A support the existence of a training
base for terrorist operations
Assess whether there are weapons and weapons
technology at Location-A that suggest the
presence of a training base for terrorist
operations
32
Assess whether there is a flow of suspected
terrorists in the region of Location-A
Assess whether there are other suspected bases
for terrorist operations in the region of
Location-A
Assess whether the political environment would
support a training base for terrorist operations
at Location-A
Assess whether there is terrorist sympathetic
population in the region of Location-A
Assess whether the physical structures at
Location-A support the existence of a training
base for terrorist operations
Assess whether there are weapons and weapons
technology at Location-A that suggest the
presence of a training base for terrorist
operations
33
(No Transcript)
34
(No Transcript)
35
(No Transcript)
36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
(No Transcript)
40
(No Transcript)
41
(No Transcript)
42
Intelligence Experts Opinion Quotations
REVIEWER 1 a grand challenge to develop an
intelligent agent capable of learning, tutoring
and decision support if implemented it would
likely be pretty unique.
REVIEWER 2 This is an innovative idea that
could revolutionize the way we do business,
enable us to be more efficient, more effective,
more thorough.
REVIEWER 3 a very important RD area for next
generation intelligence analysis. The work is
well founded, and the execution of real software
to implement the ideas is substantial.
REVIEWER 4 I have seen a briefing on the work
presented here last year and was impressed with
the initial ease of use of capturing complex
concepts. This could be excellent for use in both
training analysts as well as capturing knowledge
from more senior analysts.
43
Overview
Learning Agents Center Research Vision
Research Issues for Learning Agents
Personal Cognitive Assistant for Intelligence
Analysis
Agents for Centers of Gravity and Critical
Vulnerabilities
Virtual Experts for Multi-domain Collaborative
Planning
Agent for Course of Action Critiquing
Final Remarks
44
DARPAs Rapid Knowledge Formation Program
Develop the Disciple technology to enable teams
of subject matter experts to build integrated
knowledge bases and agents incorporating their
problem solving expertise.
Disciple-RKF Assistant
Problem solver for a non-expert
KB1
...
Disciple-RKF Assistant
Expert
Assistant of an expert
Integrated KB
Disciple-RKF Assistant
Tutor to a student
KBn
Expert
Successful experiments and transition to the US
Army War College
45
Center of Gravity Analysis
The center of gravity of an entity is its primary
source of moral or physical strength, power or
resistance. Joe Strange, Centers of Gravity
Critical Vulnerabilities, Marine Corps War
College, 1996.
If a combatant eliminates or influences the
enemys strategic center of gravity, then the
enemy will lose control of its power and
resources and will eventually fall to defeat. If
the combatant fails to adequately protect his own
strategic center of gravity, he invites disaster.
P.K. Giles and T.P. Galvin US Army War
College, 1996.
46
Use of Disciple at the US Army War College
589jw Military Applications of Artificial
Intelligence
Students teach Disciple their COG analysis
expertise, using sample scenarios(e.g. Iraq
2003, War on terror 2003, Arab-Israeli 1973)
Students test the trained Disciple agent based on
a new scenario (North Korea 2003)
Global evaluations of Disciple by officers during
three experiments
I think that a subject matter expert can use
Disciple to build an agent, with limited
assistance from a knowledge engineer
Spring 2001 COG identification
Spring 2002 COG identification and testing
Spring 2003 COG testing based on critical
capabilities
47
Use of Disciple at the US Army War College
319jw Case Studies in Center of Gravity Analysis
Disciple helps the students to perform a center
of gravity analysis of an assigned war scenario.
Disciple was taught based on the expertise of
Prof. Comello in center of gravity analysis.
Problemsolving
Teaching
DiscipleAgent
KB
Learning
Global evaluations of Disciple by officers from
the Spring 05 course
Disciple helped me to learn to perform a
strategic COG analysis of a scenario
The use of Disciple is an assignment that is well
suited to the course's learning objectives
Disciple should be used in future versions of
this course
48
Parallel development and merging of KBs
432 concepts and features, 29 tasks, 18 rules For
COG identification for leaders
Initial KB
Domain analysis and ontology development (KESME)
Knowledge Engineer (KE)
All subject matter experts (SME)
Training scenarios Iraq 2003 Arab-Israeli
1973 War on Terror 2003
Parallel KB development (SME assisted by KE)
37 acquired concepts and features for COG testing
Extended KB
DISCIPLE-COG
DISCIPLE-COG
DISCIPLE-COG
DISCIPLE-COG
DISCIPLE-COG
stay informed be irreplaceable
communicate
be influential
have support
be protected be driving force
Team 1
Team 2
Team 3
Team 4
Team 5
5 features 10 tasks 10 rules
14 tasks 14 rules
2 features 19 tasks 19 rules
35 tasks 33 rules
3 features 24 tasks 23 rules
KB merging (KE)
Learned features, tasks, rules
Integrated KB
Unified 2 features Deleted 4 rules
Refined 12 rules Final KB 9 features ? 478
concepts and features 105 tasks ?134 tasks 95
rules ?113 rules
5h 28min average training time / team 3.53
average rule learning rate / team
COG identification and testing (leaders)
DISCIPLE-COG
Testing scenario North Korea 2003
Correctness 98.15
49
Current Project Distributed Knowledge
Acquisition, Validation, and Maintenance
PROBLEM SOLVING AND LEARNING ASSISTANT
PROBLEM SOLVING AND LEARNING ASSISTANT
Operational Use and Non-Disruptive Learning
After Action Review and KB Refinement
Knowledge acquired by the agents is validated and
integrated into an improved Disciple Knowledge
Base
Integration Team Knowledge engineer Subject
matter experts
PROBLEM SOLVING AND LEARNING ASSISTANT
Operational Use and Non-Disruptive Learning
KB INTEGRATION ASSISTANT
KB Integration, Validation and Maintenance
PROBLEM SOLVING AND LEARNING ASSISTANT
After Action Review and KB Refinement
Copies of Disciple agents support users
decision-making and all learn from these
experiences.
PROBLEM SOLVING AND LEARNING ASSISTANT
PROBLEM SOLVING AND LEARNING ASSISTANT
Operational Use and Non-Disruptive Learning
After Action Review and KB Refinement
50
Army War College
Co-PI, SME Dr. Jerome Comello Experiments in
2005, 2006, 2007
PROBLEM SOLVING AND LEARNING ASSISTANT
PROBLEM SOLVING AND LEARNING ASSISTANT
Operational Use and Non-Disruptive Learning
After Action Review and KB Refinement
Integration Team Knowledge engineer Subject
matter experts
PROBLEM SOLVING AND LEARNING ASSISTANT
Experimentation Environment 2005, 2006, 2007
Operational Use and Non-Disruptive Learning
KB INTEGRATION ASSISTANT
KB Integration, Validation and Maintenance
PROBLEM SOLVING AND LEARNING ASSISTANT
George Mason University
After Action Review and KB Refinement
Air War College
Co-PI, SME Col Jeffrey Hightaian LtCol Todd
Kemper Experiments in 2006, 2007
Marine Corps War College
Co-PI, SME Dr. Joseph Strange Experiments in
2006, 2007
PROBLEM SOLVING AND LEARNING ASSISTANT
PROBLEM SOLVING AND LEARNING ASSISTANT
Operational Use and Non-Disruptive Learning
After Action Review and KB Refinement
51
Overview
Learning Agents Center Research Vision
Research Issues for Learning Agents
Personal Cognitive Assistant for Intelligence
Analysis
Agents for Centers of Gravity and Critical
Vulnerabilities
Virtual Experts for Multi-domain Collaborative
Planning
Agent for Course of Action Critiquing
Final Remarks
52
Virtual Experts for Multi-Domain Collaborative
Planning
Sample scenario Planning the response to an
emergency situation involving a tanker truck
leaking red-fuming nitric acid near a student
residential area of GMU.
Users Assistant
Scenario Specification
Report Generator
Virtual Experts (VE) Library
Profile-based Team Selector
Plan Browser
Knowledge Management
User
Local Knowledge Base
Disciple-VE
Disciple-VE
Disciple-VE
Disciple-VE
Disciple-VE
Disciple-VE
Assistant Training Modules
Plan Abstraction
Ontology
Rules
Disciple-VE
Disciple-VE
Disciple-VE
Virtual Team Manager
Disciple-VE
Disciple-VE
Team of Virtual Experts
DISTRIBUTED KNOWLEDGE BASE
Disciple-VE
CollaborativePlanner
KB
IndicatorsIdentification
Plan Brainstorming
Disciple-VE
Knowledge Management
KB
KB
VE Training Modules
Local Knowledge Base
External- Expertise Agent
KB
KB
KB
Ontology
Rules
Plan Grading
Assumption-based Reasoning
Disciple-VE
Disciple-VE
53
Overview
Learning Agents Center Research Vision
Research Issues for Learning Agents
Personal Cognitive Assistant for Intelligence
Analysis
Agents for Centers of Gravity and Critical
Vulnerabilities
Virtual Experts for Multi-domain Collaborative
Planning
Agent for Course of Action Critiquing
Final Remarks
54
DARPAs HPKB Challenge Problem
Rapid development and evaluation of a Course of
Action critiquer
To what extent does this course of action conform
to the principle of surprise?
55
DARPAs HPKB Program Evaluation
100
  • High knowledge acquisition rate
  • Better results than the other evaluated systems
  • Better performance than the evaluating experts
    (many unanticipated solutions).

56
Overview
Learning Agents Center Research Vision
Research Issues for Learning Agents
Personal Cognitive Assistant for Intelligence
Analysis
Agents for Centers of Gravity and Critical
Vulnerabilities
Virtual Experts for Multi-domain Collaborative
Planning
Agent for Course of Action Critiquing
Final Remarks
57
Disciples Vision on the Future of Software
Development
Mainframe Computers
Software systems developed and used by computer
experts
58
Vision on the Use of Disciple in Education
? 2005, Learning Agents Center
59
Acknowledgements
The research performed in the Learning Agents
Center was sponsored by several US government
agencies including Defense Advanced Research
Projects Agency, Air Force Office of Scientific
Research, Air Force Research Laboratory, National
Science Foundation, and Army War College.
60
Questions
Write a Comment
User Comments (0)
About PowerShow.com