[email protected].

et

1
KNOWLEDGE CREATION
Is defined as
 Formation of new ideas through interactions
between explicit and tacit knowledge in
individual human minds.
 As defined by Ikujiro Nonaka, Knowlwdge
creation consists of socialization (tacit to
tacit), externalization (tacit to
explicit), combination (explicit to explicit),
and internalization (explicit to tacit).
2
Nonaka’s Model of Knowledge
Creation and Transformation
TACIT TO TACIT TACIT TO EXPLICIT
(SOCIALIZATION) (EXTERNALIZATION)

e.g., Individual and/or Team e.g., Documenting a Team

Discussions Meeting

EXPLICIT TO TACIT EXPLICIT TO EXPLICIT

(INTERNALIZATION) (COMBINATION)

e.g., Learn from a report and e.g., Create a Website from

Deduce new ideas some form of explicit
knowledge
3
KNOWLEDGE CREATION
 Is a Dynamic activity that can enhance
organization success and economic well-
being
 Is a Driver of innovation
 Involves knowledge acquisition, selection,
generation and sharing
 Maturation - translates experience into
knowledge

4
Knowledge Creation and
Transfer via Teams
Initial
knowledge

Outcome is
realized
Team performs
a job Outcome compared
to action

New knowledge
reusable by same
team on next job
New experience/
Knowledge knowledge gained
captured and
codified in a form
usable by others
5
Impediments to Knowledge Sharing

Compensation Personality
Organizational
Recognition culture
Ability utilization
Lack of
Creativity Vocational Knowledge
Good work environment reinforcers sharing
Autonomy
Job security Attitude
Moral values Company
strategies and
Advancement policies
Variety Work Norms
Achievement
Independence
Social status
6
Knowledge Engineering
 Process of acquiring knowledge from
experts and building knowledge base
 Narrow perspective
 Knowledge acquisition, representation,
validation, inference, maintenance
 Broad perspective
 Process of developing and maintaining
intelligent system

7
Knowledge Engineering Process
 Acquisition of knowledge
 General knowledge or meta knowledge
 From experts, books, documents, sensors, files
 Knowledge representation
 Organizing knowledge
 Knowledge validation and verification
 Accuracy, timeliness and quality of knowledge
 Inferences
 Software designed to pass generalizations based on
statistically sampled data using logical reasoning
 Explanation and justification capabilities
 Explanation facility
8
9
Knowledge Sources
 Documented
 Written, viewed, sensory, behavior
 Undocumented
 Memory
 Acquired from
 Human senses
 Machines
 Books/documents
10
Knowledge Levels
 Shallow
 Surface level
 Input-output
 Deep
 Problem solving
 Difficult to collect, validate
 Interactions between system components

11
Knowledge Categories
 Declarative
 Descriptive representation

 Procedural
 How things work under different
circumstances
 How to use declarative knowledge
 Problem solving

 Meta-knowledge
 Knowledge about knowledge

12
Knowledge Engineers
 Are Professionals who elicit knowledge from
experts
 Should be: Empathetic and patient
 Should have: Broad range of understanding,
capabilities
 Should Integrate knowledge from various
sources
 Creates and edits code
 Operates tools
 Build knowledge base
 Validates information
 Train users 13
14
Elicitation Methods
 Manual
 Based on interview
 Track the reasoning process of experts
 Observation
 Semiautomatic
 Build knowledge base with minimal help from
knowledge engineer
 Allows execution of routine tasks with minimal
expert input
 Automatic
 Minimal input from both expert and knowledge
engineer
15
Manual Methods
 Interviews
 Structured
 Goal-oriented
 Walk through
 Unstructured
 Complex domains
 Data unrelated and difficult to integrate
 Semistructured

16
Manual Methods
 Process tracking
 Track reasoning processes
 Protocol analysis
 Document expert’s decision-making
 Think aloud process
 Observation
 Motor movements
 Eye movements
17
Manual Methods ….cont
 Case analysis
 Critical incident
 User discussions
 Expert commentary
 Graphs and conceptual models
 Brainstorming
 Prototyping
 Clustering of elements
 Iterative performance review

18
Semiautomatic Methods
 Repertory grid analysis
 Personal construct theory
 It is an organized, perceptual model of expert’s
knowledge
 Expert identifies domain objects and their attributes
 Expert determines characteristics and opposites for
each attribute
 Expert distinguishes between objects, creating a grid

 Expert transfer system

 Computer program that elicits information from experts
 Rapid prototyping
 Used to determine sufficiency of available knowledge

19
20
Semiautomatic Methods, continued
 Computer based tools features:
 Ability to add knowledge to base
 Ability to assess, refine knowledge
 Visual modeling for construction of
domain
 Creation of decision trees and rules
 Ability to analyze information flows
 Integration tools: ex from already existing
system
21
Using Multiple Experts
 Scenarios
 Experts contribute individually
 Primary expert’s information reviewed by
secondary experts
 Small group decision
 Panels for verification and validation
 Approaches
 Consensus methods
 Analytic approaches
 Automation of process through software usage
 Decomposition

22
Automatic Methods
 Data mining by computers
 Inductive learning from existing recognized
cases
 Neural computing mimicking human brain
 Genetic algorithms using natural selection

23
Automated Knowledge Acquisition
 Induction
 Activities
 Training set with known outcomes
 Creates rules for the examples
 Assesses new cases
 Advantages
 Limited application
 Builder can be expert
 Saves time, money

24
Automated Knowledge Acquisition
 Difficulties
 Rules may be difficult to understand
 Experts needed to select attributes/or need for
use of mathematical methods
 Rule-based classification problems
 Allows few attributes
 Many examples needed
 Examples may be insufficient

 Examples must be cleaned

 Limited to certainties
25
Automated Knowledge Acquisition
 Interactive induction
 Incrementally induced knowledge
 General models
 Based on interaction with expert
 interviews
 Computer supported
 Induction tables
 IF-THEN-ELSE rules

26
Evaluation, Validation, Verification
 Dynamic activities
 Evaluation
 Assess system’s overall value
 Validation
 Compares system’s performance to expert’s
 Similarity and differences
 Verification
 Building and implementing system correctly

27
28
Production Rules
 IF-THEN
 Independent part, combined with other
pieces, to produce better result
 Model of human behavior
 Examples
 IF condition, THEN conclusion
 Conclusion, IF condition
 If condition, THEN conclusion1 (OR) ELSE
conclusion2

29
Artificial Intelligence Rules
 Types
 Knowledge rules
 Declares facts and relationships
 Stored in knowledge base
 Inference
 Given facts, advises how to proceed
 Part of inference engines

30
Artificial Intelligence Rules
 Advantages
 Easy to understand, modify, maintain
 Explanations are easy to get.
 Rules are independent.
 Modification and maintenance are relatively easy.
 Uncertainty is easily combined with rules.
 Limitations
 Designers may force knowledge into rule-based
entities
 Systems may have search limitations; difficulties in
evaluation
31
Semantic Networks
 Graphical depictions
 Nodes and links
 Hierarchical
relationships
between concepts
 Reflects inheritance

32
Frames
 All knowledge about object
 Hierarchical structure allows for inheritance
 Allows for diagnosis of knowledge
independence
 Object-oriented programming
 Knowledge organized by characteristics and
attributes
 Slots
 Subslots/facets
 Parents are general attributes
 Instantiated to children
 Often combined with production rules
33
34
Knowledge Relationship Representations
 Decision tables
 Spreadsheet format
 All possible attributes compared to conclusions
 Decision trees
 Nodes and links
 Knowledge diagramming
 Computational logic
 Propositional
 True/false statement
 Predicate logic
 Variable functions applied to components of statements

35
Reasoning Programs
 Inference Engine
 Algorithms
 Directs search of knowledge base
 Forward chaining
 Data driven
 Start with information, draw conclusions
 Backward chaining
 Goal driven
 Start with expectations, seek supporting evidence
 Inference/goal tree
 Schematic view of inference process
 AND/OR/NOT nodes
 Answers why and how
 Rule interpreter
36
Explanation Facility
 Justifier
 Makes system more understandable
 Exposes shortcomings
 Explains situations that the user did not anticipate
 Satisfies user’s psychological and social needs
 Clarifies underlying assumptions
 Conducts sensitivity analysis
 Types
 Why
 How
 Journalism based
 Who, what, where, when, why, how

37
Generating Explanations
 Static explanation
 Pre-insertion of text
 Dynamic explanation
 Reconstruction by rule evaluation
 Tracing records or line of reasoning
 Justification based on empirical associations
 Strategic use of meta-knowledge

38
Expert System Development
 Phases
 Project initialization
 Systems analysis and design
 Prototyping
 System development
 Implementation
 Post-implementation

39
Project Initialization
 Identify problems
 Determine functional requirements
 Evaluate solutions
 Verify and justify requirements
 Conduct feasibility study and cost-benefit analysis
 Determine management issues
 Select team
 Project approval

40
Systems Analysis and Design
 Create conceptual system design
 Determine development strategy
 In house, outsource, mixed
 Determine knowledge sources
 Obtain cooperation of experts
 Select development environment
 Expert system shells
 Programming languages
 Hybrids with tools
 General or domain specific shells
 Domain specific tools

41
Prototyping
 Rapid production
 Demonstration prototype
 Small system or part of system
 Iterative
 Each iteration tested by users
 Additional rules applied to later iterations

42
System Development
 Development strategies formalized
 Knowledge base developed
 Interfaces created
 System evaluated, feedback received, system
improved

43
Implementation
 Adoption strategies formulated
 System installed
 All parts of system must be fully documented and
security mechanisms employed
 Field testing if it stands alone; otherwise, must be
integrated with existing systems and tested
 User approval

44
Post-implementation
 Operation of system
 Maintenance plans
 Review system, revision of rules
 Data integrity checks
 Linking to databases
 Upgrading and expansion
 Periodic evaluation and testing

45