Transcript Document

BUSS 951
Critical Issues in Information
Systems
Lecture 9
Systems for Organisations 3:
Knowledge Technologies
Clarke, R. J (2001) L951-09:
1
Notices (1)
General
 Assignment 2 is due today, and will be available at the
beginning of week 11
 Assignment 3 will be put up on the web during the first
week of the holidays- download it at your leisure
 Please check that your Assignment 1 mark is correctly
recorded, current marks are available on the
departmental notice board and pick up assignments if
you have not yet done so
 BUSS951 is supported by a website (available from
Tomorrow), where you can find out the latest Notices
get Lecture Notes, Tutorial Sheets, Assignments etc
www.uow.edu.au/~rclarke/buss951/buss951.htm
Clarke, R. J (2001) L951-09:
2
Notices (2)
Readings for this week
 Last week we postponed seminar
discussion of the allocated readings in
order to discuss arguments for
Assignment 2,
 this week we will discuss these readings:
1. Yu, E. (1998) “Why agent-oriented
requirements engineering” Reading 6
2. Yu, E. S. K and J. Mylopoulos (1994) “From ER’ to A-R’- Modelling Strategic Actor
Relationships for Business Process
Reengineering” Reading 7
Clarke, R. J (2001) L951-09:
3
Agenda (1)
Discuss some problems with traditional
systems analysis views of work in offices
Promote a view which looks at office
work in terms of action and human
communication (similar to a Systems
Auditors View of an IS)
Introduce the ideas behind Action
Workflow (one type of LAP approach)
Clarke, R. J (2001) L951-09:
4
Decision Support Systems
Clarke, R. J (2001) L951-09:
5
Decision Support Systems (1)
 information producing system aimed at a particular
problem:
 that a manager must solve
 decisions that a manager must make
 Programmed Decisions
 repetitive and routine
 definite procedure for handling them
 Nonprogrammed Decisions
 novel and unstructured
 problem hasn’t arisen before
Programmed
Decisions
Nonprogrammed
Decisions
Continuum
Clarke, R. J (2001) L951-09:
6
Decision Support Systems (2)
Phases of Decision Making
Intelligence Activity
searching environment for problems
Design Activity
inventing, developing, analysing possible courses
of action
Choice Activity
selecting a course of action
Review Activity
assessing past choices
Clarke, R. J (2001) L951-09:
7
Developing DSS
using Simons phases
can build systems which perform this work
use a special form of systems
development called rapid prototyping (see
appropriate lecture)
Clarke, R. J (2001) L951-09:
8
DSS Capabilities (1)
May use one or more techniques
depending on the problem:
Mathematical Modelling
Simulation
Graphics
Visualisation (Virtual Reality)
Clarke, R. J (2001) L951-09:
9
DSS Capabilities (2)
Mathematical Modelling
Static Model (doesn’t include time)
Dynamic Model (includes time)
Deterministic Model (avoids probability)
Probabilistic Models (includes probability)
Optimising Model (selects best solution)
Suboptimising Model (‘satisficing’ solution)
Clarke, R. J (2001) L951-09:
10
DSS Capabilities
(3)
Steps towards Simulation:
 scenario & scenario data elements
managers input is referred to
decision variables (With McDonalds
Example: people, number of queues,
expected load over a day)
Clarke, R. J (2001) L951-09:
11
DSS Model
Individual
Problem
Solvers
Other
Group
Members
Report
writing
software
Math
Models
GDSS
Software
Database
Environment
Environment
Data
Info
Coms
Clarke, R. J (2001) L951-09:
12
DSS/ES
Decision Support Systems applications are
beginning use a technology called Expert
Systems
an Expert System (ES) comprises a
collection of facts and a set of rules which
are processed by an inference engine
we will discuss these in greater depth latter
in the lecture
Clarke, R. J (2001) L951-09:
13
Executive Information
Systems
Clarke, R. J (2001) L951-09:
14
Executive IS (1)
Executive Information Systems (EIS) are
designed specifically for managers on the
strategic planning level
executive activity is not very structureddifficulties in establishing an understanding
of executive problem solving
Warning: problems with acronyms- the term
EIS has a variety of definitions it is also been
defined as Enterprise Information Systems
Executive Information Systems
Clarke, R. J (2001) L951-09:
15
Executive IS (2)
three ways of achieving EIS:
develop custom software
application end-user development software
(spreadsheets, database, graphics)
install special EIS software
EIS characteristic- drill down
executive begins with an overview
gradually retrieve more information
Clarke, R. J (2001) L951-09:
16
Executive IS
EIS Characteristics- Drill down
 executive begins with
an overview and then
gradually retrieves
more specific
information
 often by accessing
lower level systems
(transaction
processing or
operational systems)
Clarke, R. J (2001) L951-09:
17
Executive IS
EIS Characteristics- Slice and Dice
 taking a large
database
 progressively
searching through the
data using various
combinations of
variables
 in order to understand
a complex problem
which is resolvable
Clarke, R. J (2001) L951-09:
18
Executive IS (3)
sometimes a distinction is made between
EIS and ESS
Executive Support Systems (ESS)
support information needs
also communications & analysis needs
does this by providing intelligence ie. by
understanding how information affects
operations.
Does this distinction make any sense?
Clarke, R. J (2001) L951-09:
19
EIS Model
Executive Workstation
Executive
database
Personal
Computer
To other Workstations
Corp. Database
Electronic Databases
Info
Requests
Info
Display
To other Workstations
Make
Corp. Info.
available
External
data &
info.
Current news,
explanations
Software Library
Corporate Mainframe
Clarke, R. J (2001) L951-09:
20
Concept of Organisational
Information Subsystems (1)
subsets of MIS tailored to meet
users needs in functional areas
examples:
marketing information systems
manufacturing information systems
financial information systems
human resources information
systems
Clarke, R. J (2001) L951-09:
21
Concept of Organisational
Information Subsystems (2)
Remember: Nothing physically
separates these systems- rather these
systems are logical distinct
databases used by one organisational
subsystem can be used by others
programs may also be shared
Clarke, R. J (2001) L951-09:
22
Concept of Organisational
Information Subsystems (3)
these systems are not an alternative to
a firm-wide MIS
need these subsystems to have a
complete organisation wide MIS
development strategy: MIS then
implement organisational information
subsystems
Clarke, R. J (2001) L951-09:
23
Organisational Information
Systems
EIS
Marketing
Financial Human
Resource
Manufacturing
Information Systems
Clarke, R. J (2001) L951-09:
24
Knowledge Systems
Artificial Intelligence & Expert Systems
Clarke, R. J (2001) L951-09:
25
Knowledge Systems
AI vs ‘Non-Intelligent’ programs
 Artificial Intelligence (AI) has emerged as one of the
most significance technologies of this century
 subfield of computing science that is concerned
with symbolic reasoning and problem solving, by
manipulation of knowledge rather than mere data
 classical ‘non-intelligent’ computer programs:
 are rigid, structured procedures that deal with
specific problems
 programs may be flexible and may be capable of
dealing with complex situations
 everything non-intelligent programs do is predictable
or preordained
Clarke, R. J (2001) L951-09:
26
Knowledge Systems
Intelligent Programs
 Intelligent Programs
 can on occasion exhibit behaviours that were
not programmed
 consists of a complex set of rules on how to
process data as well as having information
stored in a database
 thee programs are generally goal directedthey are designed to behave in order to reach
their goals rather than being told how to
achieve them
Clarke, R. J (2001) L951-09:
27
Knowledge Systems
Heuritsics- Rules of Thumb
A key characteristic of AI programs is heuristics or
rules of thumb which guide the execution of the program:
Conventional Programs
 Often primarily numeric
 Algorithmic- solution steps
explicit
 Integrated information and
control
 Difficult modification
 Correct answer required
 Best possible solution usually
sought
AI Programs
 Primarily symbolic processes
 Heuristic search- solution steps
implicit
 Control structure usually separate
from knowledge domain
 Usually easy to modify, update and
enlarge
 Some incorrect answers are often
tolerable
 Satisfactory answers usually
acceptable
Clarke, R. J (2001) L951-09:
28
Knowledge Systems
AI: A Broad Spectrum of Technologies










Expert systems
Natural language
Speech processing
Vision
Robotics
Cognitive modelling
Knowledge representation and utilization
Problem solving and inference
Learning- knowledge acquisition
Special purpose computer hardware
Clarke, R. J (2001) L951-09:
29
Knowledge Systems
Expert Systems
 “An expert system is an intelligent computer program
that uses knowledge and inference procedures to solve
problems that are difficult enough to require significant
human expertise for their solution. The knowledge
necessary to perform at such a a level, plus the inference
procedures used, can be thought of a s a model of the
expertise of the practitioners of the field”
 “the knowledge of an expert consists of facts and
heuristics. The ‘facts’ constitute a body of information
that is widely shared, publicly available, and generally
agreed upon by experts in a field. The performance level
of an expert system is primarily a function of the size and
quality of the knowledge base that it processes”
(Feigenbaum, E. A. & P. McCorduck 1983)
Clarke, R. J (2001) L951-09:
30
Knowledge Systems
Expert Systems Applications
Typical expert systems applications
(O’Brien, 355):
Decision Management (recommendations)
Diagnostic and troubleshooting
Maintainence and Scheduling (prioritizing,
scheduling)
Intelligent text and documentation (legal
documents)
Design/configuration
Selection/classification (suspect id)
process monitoring/control (chemical testing)
Clarke, R. J (2001) L951-09:
31
Knowledge Systems
Expert Systems-Architecture
 architecturally expert
systems are considered
as rule-based systems
 they provide a separation
between the knowledge
base, and the inference
engine, and may contain a
built-in explanation
facility (reasoning) and a
dialogue component(based on computation
linguistics)
Clarke, R. J (2001) L951-09:
32
Knowledge Systems
ES Application Areas & Problem Classes
 Expert systems application
areas are often divided into
particular domains
 analysis problems which
include:
 debugging,
 diagnosis, and
 interpretation
 synthesis problems which
include:
Classes
Diagnosis
Design
Planning
Simulation
Approach
Pattern recognition
Object Generation
Action Generation
Modelling
 Configuration
 Planning, and
 scheduling
Clarke, R. J (2001) L951-09:
33
ES System Prescriptions
 developed by Stefik et al
(1984), this table consists of
the requirements and
possible development
prescriptions that might be
used when developing
expert systems
 three major development
options include:
 when there is unreliable data
or knowledge
 when the data are time varying,
and
 when the problem to be solved
involves a big solution space
Clarke, R. J (2001) L951-09:
34
Knowledge Systems
Expert Systems Development
 interestingly although we might
think of expert systems as
somehow privilege and special,
in fact building them is similar to
other systems
 notice the prototyping cycle in
the lower half of the diagram
 this is what is referred to as
Rapid Prototyping and is found
in conventional systems
development as well here it is
rather well suited to the process
of rule creation and improvement
( = knowledge acquisition)
Clarke, R. J (2001) L951-09:
35
Expert System Classification
Knowledge vs Technological Complexity
Broad scope, great depth,
information ambiguity,
high database mgmt, integrity
Clarke, R. J (2001) L951-09:
36
Knowledge Systems
Types of Expert Systems 1: Low Tech
1.
2.
Personal productivity systemssimplest system, eg. Personal
budgeting systems running on
PCs, built using an expert
systems shell
Power Decision systemknowledge intensive
incorporating skills of highly
skilled decision makers, goal top
improve decision making of a
group eg. REVA- repairs and
faults with electric pumps
centrifuges etc
Clarke, R. J (2001) L951-09:
37
Knowledge Systems
Types of Expert Systems 2: High Tech
3.
4.
Integrated production systems- limited
amounts of domain complexity but
involve advanced technology. Tend to
target organisational productivityimproving throughput, reducing
headcounts and lowering costs, eg.
Telex Router receives all telexes
coming into a banks headquarters,
reformats them and routes them to
appropriate employees
Strategic Impact Systems- are both
broad and deep in their domains. The
decision process is long and intricate,
and often requires testing of numerous
hypotheses. Example: Lincoln National
Life Underwriting System which knows
about medical, financial and insurance
fields
Clarke, R. J (2001) L951-09:
38
Knowledge Systems
Benefits of Expert Systems
1. Improve personal
decision making
2. Improve group decision
making and enhance
the quality of product
service
3. Improve organisational
throughput and costs
4. Create market barriers
and improve
organisational decision
making and productivity
Clarke, R. J (2001) L951-09:
39
Knowledge Systems
Benefits of Expert Systems
1.
2.
3.
4.
Personal Productivity Systems:
end user- supported by
information systems
Power Decision Systems:
existing business unit- close to
experts
Integrated Production Systems:
Data processing/MIS
departments- driven by
technical needs
Strategic Impact Systems: new
system business unit- full time,
resource hungry, new business
generating
Clarke, R. J (2001) L951-09:
40
Integrating Expert Systems
 Bowerman & Glover (1988)
identified five different types of
integration between IS and ES
based on production
environment requirements:
 Standalone eg. Personal
Consultant Plus
 Business eg. ACORN & Lotus
1-2-3
 Scientific eg. Rulemaster &
Informix
 Process Control: eg. PICON &
GKS & TDC-3000
 Decision Support: eg. TIMM &
Comshare & Nichols &
Versatec
Clarke, R. J (2001) L951-09:
41
Integrating Expert Systems
 a major area of research activity
was how to construct
standardised interfaces between
numerous expert systems
 in the diagram on left- a
hypothetical process control
function- the inputs constitute
data collected from sensors and
condensed by pre-processing
computer
 outputs consist of conventional
computer for controlling
processes
Clarke, R. J (2001) L951-09:
42
Conclusions
Some Critical Issues
 we have concentrated on expert systems as new of the
most useful ‘knowledge’ technologies in information
systems
 yet are expert systems knowledgeable- would we be
thinking of these technologies as intelligent if the
metaphor we used was not anthropomorphic (human-like
or of human form)
 if we though of these technologies as simply
reproducing patterns where the patterns were in form of
sets of rules- would we kid ourselves that there were
intelligent
 this is probably simply the same kind of discourse (way
of speaking & way of thinking) as is illustrated in the
cybernetic organisation metaphor we talked about earlier
Clarke, R. J (2001) L951-09:
43
Conclusions
Some Critical Issues
 the facts and heuristics of expert systems for
example constitute a particular ontological and
epistemological position on the world
 this definition of expertise ignores for example that
with humans these facts are usually acquired
through the senses and with the aid of a physical
body rather than simply being a consequence of
which rational cognition
 this is not to deny the usefulness of these systems
in the niche markets which they address- for many
kind of computation these and related technologies
are simply the only sensible way of solving certain
class of computational problem
Clarke, R. J (2001) L951-09:
44
Conclusions
Some Critical Issues
 it is to remind us that equating this kind of
programmatic execution to intelligence is at
best a trope
 … a word or expression used in a figurative
sense
Clarke, R. J (2001) L951-09:
45
References
 Gottinger, H. W. and H. P. Weinmann (1990) Artificial
Intelligence, A Tool for Industry and Management
Chapters 1 and 3, pp. 9-12 & pp. 22-31
 Meyer, M. H. & K. F. Curley (1991) Putting Expert
Systems Technology to Work Sloan Management
Review 32 (2), pp.31
 Bowerman, R. G. and D. E. Glover (1988) Putting
Expert Systems into Practice Van Nostrand Reinhold
Chapter 8, pp. 287-297 & pp. 307-317
 Murray, J. T. & M. J. Murray (1988) Expert Systems in
Data Processing: A Professional Guide, Pitman
Publishing Co., Chapter 1
Clarke, R. J (2001) L951-09:
46
 Feigenbaum, E. A. & P. McCorduck (1983) Fifth
generation Reading, Mass.: Addison-Wesley
 Stefik, M et al (1984) “The organization of expert
systems: a perspective tutorial” in Hayes-Roth, F. et al
(1984) Building expert systems Addison-Wesley:
Redaing, Mass.
Clarke, R. J (2001) L951-09:
47