Chapter 11 - Repository UNIKOM
Simon’s Types of Decisions
Simon’s Types of Decisions
Programmed decisions Programmed decisions
- – repetitive and routine
repetitive and routine
- – have a definite procedure
have a definite procedure
Nonprogrammed decisions Nonprogrammed decisions
- – Novel and unstructured
Novel and unstructured
- – No cut-and-dried method for handling problem
No cut-and-dried method for handling problem
Types exist on a continuum Types exist on a continuum
Simon’s Problem Solving Phases
Simon’s Problem Solving Phases Intelligence
Intelligence
Searching environment for conditions calling for a solution
- – Searching environment for conditions calling for a solution
Design Design
Inventing, developing, and analyzing possible courses of action action
- – Inventing, developing, and analyzing possible courses of
Choice
Choice
Selecting a course of action from those available
- – Selecting a course of action from those available
Review Review
Assessing past choices
- – Assessing past choices
Definitions of a Decision Definitions of a Decision
Support System (DSS) Support System (DSS)
General definition - General definition -
a system providing both a system providing both problem-solving and communications capabilities problem-solving and communications capabilities for semistructured problems for semistructured problems
Specific definition - Specific definition - a system that supports a
a system that supports a single manager or a relatively small group of single manager or a relatively small group of
managers working as a problem-solving team in
managers working as a problem-solving team in
the solution of a semistructured problem by the solution of a semistructured problem by providing information or making suggestions providing information or making suggestions concerning concerning specific specific decisions. decisions.
The DSS Concept
The DSS Concept
Gorry and Scott Morton coined the phrase ‘DSS’
Gorry and Scott Morton coined the phrase ‘DSS’
in 1971, about ten years after MIS became popular in 1971, about ten years after MIS became popular Decision types in terms of problem structure
Decision types in terms of problem structure
- – Structured problems can be solved with algorithms and
Structured problems can be solved with algorithms and decision rules decision rules
- – Unstructured problems have no structure in Simon’s
Unstructured problems have no structure in Simon’s phases phases
- – Semistructured problems have structured and
Semistructured problems have structured and unstructured phases unstructured phases
Degree of Degree of problem problem structure structure
The Gorry and Scott Morton Grid
The Gorry and Scott Morton Grid Management levels Management levels Structured Structured Semistructured Semistructured Unstructured Unstructured Operational Operational control control Management Management control control Strategic Strategic planning planning Accounts receivable Order entry Inventory control Budget analysis-- engineered costs Short-term forecasting Tanker fleet mix Warehouse and factory location Production scheduling Cash management PERT/COST systems Variance analysis-- overall budget Budget preparation Sales and production Mergers and acquisitions New product planning R&D planning
Alter’s DSS Types
Alter’s DSS Types
In 1976 Steven Alter, a doctoral student In 1976 Steven Alter, a doctoral student built on Gorry and Scott-Morton framework built on Gorry and Scott-Morton framework
- – Created a taxonomy of six DSS types
Created a taxonomy of six DSS types
- – Based on a study of 56 DSSs
Based on a study of 56 DSSs
Classifies DSSs based on “degree of Classifies DSSs based on “degree of problem solving support.” problem solving support.”
Levels of Alter’s DSSs
Levels of Alter’s DSSs Level of problem-solving support from
Level of problem-solving support from lowest to highest lowest to highest
- – Retrieval of information elements
Retrieval of information elements
- – Retrieval of information files
Retrieval of information files
- – Creation of reports from multiple files
Creation of reports from multiple files
- – Estimation of decision consequences
Estimation of decision consequences
- – Propose decisions
Propose decisions
- – Make decisions
Make decisions
Importance of Alter’s Study
Importance of Alter’s Study
Supports concept of developing systems Supports concept of developing systems that address particular decisions that address particular decisions
Makes clear that DSSs need not be Makes clear that DSSs need not be restricted to a particular application type restricted to a particular application type
Alter’s DSS Types Alter’s DSS Types problem problem of of Degree Degree information entire reports decision decisions decisions information entire reports decision decisions decisions elements files from consequen- elements files from consequen- Retrieve Retrieve Analyze Analyze Prepare Prepare Estimate Estimate Propose Propose Make Make solving solving support support multiple multiple ces ces files files
Degree of Degree of Little
Much Little complexity of the complexity of the problem-solving problem-solving Much
Three DSS Objectives 1
Three DSS Objectives
1. Assist in solving semistructured problems Assist in solving semistructured problems 2.
2. Support, not replace, the manager Support, not replace, the manager 3.
3. Contribute to decision effectiveness, rather Contribute to decision effectiveness, rather than efficiency than efficiency
Based on studies of Keen and Scott-Morton
A DSS Model A DSS Model
Individual Individual problem problem Environment Environment group group Other Other solvers solvers members members software software writing writing Report Report Mathematical Mathematical GDSS GDSS Models software Models software software GDSS
Database Database Decision support system
Environment Environment Database Contents Database Contents
Used by Three Software Subsystems Used by Three Software Subsystems – Report writers Report writers » Special reports Special reports » Periodic reports Periodic reports » COBOL or PL/I COBOL or PL/I » DBMS DBMS
- – Mathematical models Mathematical models » Simulations Simulations » Special modeling languages Special modeling languages
- – Groupware or GDSS Groupware or GDSS
Computer-based system that supports groups of
Computer-based system that supports groups of people engaged in a common task (or goal) and people engaged in a common task (or goal) and
that provides an interface to a shared environment.
that provides an interface to a shared environment.
Used in problem solving
Used in problem solving
Related areas Related areas
- – Electronic meeting system (EMS)
Electronic meeting system (EMS)
- – Computer-supported cooperative work (CSCW)
Computer-supported cooperative work (CSCW)
- – Group support system (GSS)
Group support system (GSS)
- – Groupware
Groupware
How GDSS Contributes How GDSS Contributes
to Problem Solving
to Problem Solving
Improved communications Improved communications
Improved discussion focus Improved discussion focus
Less wasted time Less wasted time
GDSS Environmental Settings GDSS Environmental Settings
Synchronous exchange
Synchronous exchange
- – Members meet at same time
Members meet at same time
- – Committee meeting is an example
Committee meeting is an example
Asynchronous exchange Asynchronous exchange
- – Members meet at different times
Members meet at different times
- – E-mail is an example
E-mail is an example More balanced participation.
More balanced participation.
GDSS Types
GDSS Types Decision rooms Decision rooms
- – Small groups face-to-face
Small groups face-to-face
- – Parallel communication
Parallel communication
- – Anonymity
Anonymity Local area decision network
Local area decision network
- – Members interact using a LAN Members interact using a LAN
Legislative session Legislative session
- – Large group interaction
Large group interaction Computer-mediated conference
Computer-mediated conference
- – Permits large, geographically dispersed group interaction
Permits large, geographically dispersed group interaction
Decision
Room
Local AreaDecision
NetworkGroup Size and Location Determine
Group Size and Location Determine
GDSS Environmental Settings
GDSS Environmental Settings
Smaller Larger GROUP GROUP SIZE SIZE Face-to- face Dispersed
Legislative Session Computer- Mediated Conference MEMBER MEMBER PROXIMITY PROXIMITY
Groupware
Groupware Functions
Functions
- – FAX
FAX
- – Voice messaging
Voice messaging
- – Internet access
Internet access
Lotus Notes Lotus Notes
- – Popular groupware product
Popular groupware product
- – Handles data important to managers
Handles data important to managers Main Groupware Functions Main Groupware Functions
IBM TeamWARE Lotus Novell
IBM TeamWARE Lotus Novell Function Workgroup Office Notes GroupWise Function Workgroup Office Notes GroupWise Electronic mail
X X
X X FAX
X X O
X Voice messaging O
X Internet access
X X O
X Bulletin board system
X
3 O Personal calendaring
X X
3 X Group calendaring
X X O
X Electronic conferencing O
X
3
3 Task management
X X
3 X Desktop video conferencing O Database access O
X
3 Workflow routing O
X
3 X Reengineering O
X
3 Electronic forms O
3
3 O Group documents O
X X O Artificial Intelligence (AI)
Artificial Intelligence (AI)
The activity of providing such machines as computers with the ability to display behavior that would be regarded as intelligent if it were observed in humans.
History of AI
History of AI Early history
Early history
- – John McCarthy coined term, AI, in 1956, at
John McCarthy coined term, AI, in 1956, at Dartmouth College conference. Dartmouth College conference.
- –
Logic Theorist (first AI program. Herbert Simon
Logic Theorist (first AI program. Herbert Simon
played a part) played a part)- – General problem solver (GPS)
General problem solver (GPS)
Past 2 decades Past 2 decades
- –
Research has taken a back seat to MIS and DSS
Research has taken a back seat to MIS and DSS development development Areas of Artificial Intelligence
Areas of Artificial Intelligence Expert Expert systems systems AI AI hardware hardware Robotics Robotics Perceptive Perceptive systems systems (vision, (vision, hearing) hearing) Neural Neural networks networks Natural Natural language language Learning Artificial Intelligence Artificial Intelligence
Appeal of Expert Systems
Appeal of Expert Systems
Computer program that codes the Computer program that codes the knowledge of human experts in the form of knowledge of human experts in the form of heuristics heuristics
Two distinctions from DSS Two distinctions from DSS
- – 1. Has potential to extend manager’s problem-
1. Has potential to extend manager’s problem- solving ability solving ability
- – 2. Ability to explain how solution was reached
2. Ability to explain how solution was reached
Instructions & Solutions & Knowledge information explanations
User User interface Know- Inference Problem ledge engine
Domain base Development engine Expert Expert An Expert
An Expert system system
System Model System Model
Expert and
User interface
Expert System Model Expert System Model
User interface
- – Allows user to interact with system
Allows user to interact with system
Knowledge base Knowledge base
- – Houses accumulated knowledge
Houses accumulated knowledge
Inference engine Inference engine
- – Provides reasoning
Provides reasoning
- – Interprets knowledge base
Interprets knowledge base
Development engine Development engine
- – Creates expert system
Creates expert system
User Interface User Interface
User enters: User enters:
- – Instructions
Instructions
- – Information
Information
Expert system provides: Expert system provides:
- – Solutions
Solutions
- – Explanations of
Explanations of »
Questions Questions
» Problem solutions
Problem solutions }
Menus, commands, natural language, GUI
Knowledge Base
Knowledge Base
Description of problem domain Description of problem domain
Rules Rules
- – Knowledge representation technique
Knowledge representation technique
- – ‘
- – Networks of rules
‘
IF:THEN’ logic
IF:THEN’ logic
Networks of rules »
Lowest levels provide evidence Lowest levels provide evidence
» Top levels produce 1 or more conclusions
Top levels produce 1 or more conclusions » Conclusion is called a Goal variable.
Conclusion is called a Goal variable.
A Rule Set That Conclusion
Produces One Final Conclusion Conclusion Conclusion Evidence Evidence Evidence Evidence Evidence Evidence Evidence Evidence Rule Selection Rule Selection
Selecting rules to efficiently solve a Selecting rules to efficiently solve a problem is difficult problem is difficult
Some goals can be reached with only a few Some goals can be reached with only a few rules; rules 3 and 4 identify bird rules; rules 3 and 4 identify bird
Inference Engine Inference Engine
Performs reasoning by using the contents of Performs reasoning by using the contents of knowledge base in a particular sequence knowledge base in a particular sequence
Two basic approaches to using rules Two basic approaches to using rules
- – 1. Forward reasoning (data driven)
1. Forward reasoning (data driven)
- – 2. Reverse reasoning (goal driven)
2. Reverse reasoning (goal driven)
Forward Reasoning
Forward Reasoning
(Forward Chaining)
(Forward Chaining)
Rule is evaluated as: Rule is evaluated as:
- – (1) true, (2) false, (3) unknown
(1) true, (2) false, (3) unknown
Rule evaluation is an iterative process Rule evaluation is an iterative process
When no more rules can fire, the reasoning When no more rules can fire, the reasoning process stops even if a goal has not been process stops even if a goal has not been reached reached
Start with inputs and work to solution
Rule 1 Rule 1
Rule 3 Rule 3 Rule 2 Rule 2 Rule 4 Rule 4 Rule 5 Rule 5 Rule 6 Rule 6 Rule 7 Rule 7 Rule 8 Rule 8 Rule 9 Rule 9 Rule 10 Rule 10 Rule 11 Rule 11 Rule 12 Rule 12 IF A THEN B IF C THEN D IF M THEN E IF K THEN F IF G THEN H IF B OR D THEN K IF E THEN L IF K AND L THEN N IF M THEN O IF N OR O THEN P F
OR J THEN M IF (F AND H) OR J THEN M
The Forward
The Forward
Reasoning
Reasoning
Process
Process T T T T T T T T T F
T Legend: Legend: First pass Second pass Third pass Reverse Reasoning Steps Reverse Reasoning Steps
(Backward Chaining) (Backward Chaining)
Divide problem into subproblems Divide problem into subproblems
Try to solve one subproblem Try to solve one subproblem
Then try another
Then try another
Start with solution and work back to inputs
Are Identified
IF M THEN E
Rule 8
Rule 7 Rule 10 Rule 12
Are Identified
The First Five Problems
The First Five Problems
IF M T
IF (F AND H) OR J IF M THEN O
IF E THEN L
T Rule 1
Rule 2 Rule 3
IF N OR O THEN P
IF K AND L THEN N
IF B OR D THEN K
IF A THEN B
Step 5
Step 3 Step 2 Step 1
Legend: Problems to be solved Step 4
Rule 9 Rule 11
IF C THEN D
The Next Four Problems Are
The Next Four Problems Are
Rule 12 Step 8 Identified
Identified
If N Or ORule 4 Then P T
If K Then F T Step 7 Step 6 Step 9
Rule 5 If G
If M
IF (F And H) Then H Or J Then O T T Then M T
Legend: Rule 9 Rule 11 Rule 6 Problems to be solved
If I
Forward Versus Reverse Reasoning
Forward Versus Reverse Reasoning
Reverse reasoning is faster than forward Reverse reasoning is faster than forward reasoning reasoning
Reverse reasoning works best under certain Reverse reasoning works best under certain conditions conditions
- – Multiple goal variables
Multiple goal variables
- – Many rules
Many rules
- – All or most rules do not have to be examined in
All or most rules do not have to be examined in the process of reaching a solution the process of reaching a solution
Development Engine Development Engine
Programming languages
Programming languages
- – Lisp
Lisp
- – Prolog
Prolog
Expert system shells Expert system shells
- – Ready made processor that can be tailored to a
Ready made processor that can be tailored to a particular problem domain particular problem domain
Case-based reasoning (CBR)
Case-based reasoning (CBR)
Decision tree Decision tree
Expert System Advantages
Expert System Advantages For managers
For managers
- – Consider more alternatives
Consider more alternatives
- – Apply high level of logic
Apply high level of logic
- – Have more time to evaluate decision rules
Have more time to evaluate decision rules
- – Consistent logic
Consistent logic
For the firm For the firm
- –
Better performance from management team
Better performance from management team
- – Retain firm’s knowledge resource
Retain firm’s knowledge resource
Expert System Disadvantages
Expert System Disadvantages
Can’t handle inconsistent knowledge Can’t handle inconsistent knowledge
Can’t apply judgment or intuition Can’t apply judgment or intuition
Keys to Successful ES
Keys to Successful ES
Development
Development
Coordinate ES development with strategic planning
Coordinate ES development with strategic planning
Clearly define problem to be solved and understand
Clearly define problem to be solved and understand
problem domain problem domain Pay particular attention to ethical and legal feasibility
Pay particular attention to ethical and legal feasibility of proposed system of proposed system
Understand users’ concerns and expectations
Understand users’ concerns and expectations concerning system concerning system
Employ management techniques designed to retain
Employ management techniques designed to retain developers developers
Neural Networks
Neural Networks
Mathematical model of the human brain Mathematical model of the human brain
- – Simulates the way neurons interact to process
Simulates the way neurons interact to process data and learn from experience data and learn from experience
Bottom-up approach to modeling human Bottom-up approach to modeling human intuition intuition
The Human Brain The Human Brain
Neuron -- the information processor Neuron -- the information processor
- – Input -- dendrites
Input -- dendrites
- – Processing -- soma
Processing -- soma
- – Output -- axon
Output -- axon
Neurons are connected by the synapse Neurons are connected by the synapse
Soma (processor ) Axon Synapse
Dendrites (input) Axonal Paths (output)
Simple Biological Neurons Simple Biological Neurons
Evolution of Artificial
Evolution of Artificial
Neural Systems (ANS)
Neural Systems (ANS)
McCulloch-Pitts mathematical neuron McCulloch-Pitts mathematical neuron function (late 1930s) was the starting point function (late 1930s) was the starting point
Hebb’s learning law (early 1940s) Hebb’s learning law (early 1940s)
Neurocomputers Neurocomputers
- – Marvin Minsky’s Snark (early 1950s)
Marvin Minsky’s Snark (early 1950s)
- – Rosenblatt’s Perceptron (mid 1950s)
Rosenblatt’s Perceptron (mid 1950s)
Current Methodology Current Methodology
Mathematical models don’t duplicate Mathematical models don’t duplicate human brains, but exhibit similar abilities human brains, but exhibit similar abilities
Complex networks Complex networks
Repetitious training Repetitious training
- – ANS “learns” by example
ANS “learns” by example y 1 y 2 y 3 y w 1 w 2 w 3 w n-1
Single Artificial Neuron Single Artificial Neuron
The Multi-Layer
The Multi-Layer
Perceptron
Perceptron Y n2
OUT OUT n n
OUT OUT 1 1
Y Y 1 1 Input Input Layer Layer OutputL OutputL ayer ayer
Knowledge-based Systems Knowledge-based Systems
in Perspective
in Perspective
Much has been accomplished in neural nets Much has been accomplished in neural nets and expert systems and expert systems
Much work remains Much work remains
Systems abilities to mimic human Systems abilities to mimic human intelligence are too limited and regarded as intelligence are too limited and regarded as primitive primitive
Summary [cont.]
Summary [cont.]
AI AI
- – Neural networks
Neural networks
- – Expert systems
Expert systems
Limitations and promise Limitations and promise