• US's connections with the outside world are insufficient,
• Grassroots knowledge is superficial.
• It is extremely dependent on human-expert labor.
• Almost no learning skills.
• Its usage area may be considered limited for now.
• Methods of reasoning are limited.
• Information presentation methods are limited.
1.2.4. Some Questions for a Human-Expert for Interrogating Expert Systems
Can a US perform as well as a human?
The power of the human mind is undoubtedly to understand complex concepts, implications and imprecise
even conflicting information and information that cannot be seen among them
It has the capacity to perceive similarities. Also useful information
the ability to make circuits and quickly separate relevant items from piles of details
has. This feature allows you to quickly see and use the correct method to solve problems.
Like rapidly eliminating many possibilities and dealing with only a few analyses.
It provides an extremely important advantage.
The main purpose of USs is to create a new synthesis by combining these two special capabilities.
is to be reached. In other words, it is the reasoning process of human intelligence,
New formations are aimed by adding the precision and speed of the computer. This
In areas where it can be applied, US will be stronger than humans, but judgment
In every field where intelligence is essential, human intelligence has great advantages.
will continue. (40)
Can USs improve themselves?
AI experts have already developed several tests, such as psychological testing and recognition of geometric shapes.
They have developed programs with learning ability in the field of simple problems. BASE
For learning, the results it initially produces are the result of externally given information.
It means knowing how to get/change new rules. One
The next stage is for the system to learn how to create new rules.
should be. All of this requires the system to examine its own reasoning process.
requires. Beyond that, there are many solutions or solutions for many problems.
There is a series; This makes the learning process difficult. For these reasons,
For now, there is no US that works with "real learning ability".
Isn't the ability to learn a necessary condition for intelligence?
Intelligence has many parts. Learning is just one of them. From this point
Apparently, USs cannot qualify; However, USs are related to intelligence.
It cannot be said that they are not. Just like older people whose learning abilities decrease,
As we would say he won't be smart.
Can a US adapt to a new situation as quickly as a human?
The answer to this question depends on what is meant by a "new" situation. If
If the new ones are just new data (input coefficient values), the answer is yes. If
If the “new” ones are problems for which no preparation has been made before, then the answer is
Generally no. Despite this, people-experts have never encountered before.
It may show poor performance regarding problems. Here is the human expert
The performance of the person will be directly affected by the reasoning power of the person.
It is relevant.
How does US act when given problems are outside its area of expertise?
A US has a problem that it cannot solve because no information is provided and data is missing or
cannot make a distinction between the problem it conflicts with. This inadequacy is the result of general knowledge
and the system has no reasoning ability that gives it the ability to reason.
This is because there are no multiples. It should not be forgotten that USs carry out certain tasks
occur without the intention of solving general intelligence problems, keeping in mind
have been placed.
What distinguishes US from an ordinary computer program?
For now, the main difference is the way the reasoning is presented. These are the rules in the US.
are kept separate from the machines that will interpret them, whereas in classical programs these two
are combined. Ultimately, the path the US will follow cannot be decided in advance, but
dynamically determined as rules are invoked; In classical programs, the decision
The stages are programmed clearly. Therefore, based on knowledge
systems allow more chances for change of information held in the system.
How can the differences between knowledge base and database be interpreted?
First of all, databases have the ability to store and contain many more things.
Since it creates the expert's memory, it takes up more space in the system, while
In contrast, the knowledge base is more about reasoning and problem solving.
According to one view, intelligence is effective when combined with a significant amount of factual knowledge.
It is accepted that it will happen. Good judgment and good memory complement each other.
complements each other and that is why these two approaches are brought together.
Can we be worried that US's are just a fad?
In general, fashion emerges without having much of an idea why.
It is a phenomenon that appears and disappears. Introducing USs and its advantages
computer programming that has influenced information scientists who see
It is a new style. If they come out and talk about US's for a few years, this will be US's.
It doesn't mean that things are fashionable, but I guess the ideas in this field are
It means that it is integrated with the technique.
1.2.5. Human Factor in Expert Systems
At least two people are involved in the development and use of US; expert and user.
Often the knowledge engineer and system builder are also included in this group.
Expert: Along with his special knowledge, judgment, experience and method, this
can apply skills to problems and provide advice. Task of the expert
It is to present to the system the tasks that the information system will perform and how it will be performed. Expert
which facts are important and the relationships between those facts
He is the one who knows the meaning.
User: USs have many users. They appear before us with the following identities:
They can come out:
The non-expert client who wants direct advice is a person who wants to learn.
student, US maker who wants to develop or increase their knowledge base, expert, etc.
Users' knowledge about computers or in-depth knowledge of problems
They may not have information. But many people find it quicker and easier by using US's.
They probably want to reach decisions that cost less. US's
Since their capabilities are developed to save time and effort, they
provides users with the shortest answers, unlike traditional computer systems.
They provide.
Knowledge Engineer: Knowledge engineer helps experts evaluate problem areas
It helps for. We do this by interpreting and integrating human-expert answers,
He does this by making analogies and giving reverse examples. This person is also
It is (mostly) the person who makes the system. Knowledge engineers in US construction
Its deficiency is a significant problem. Those who designed the US faced this difficulty
the need for knowledge engineers using production tools to defeat
They are reducing.
Other Participants: Many other people can attend USs. For example system
The constructor helps USs integrate with other computer systems.
The seller and other support staff can also be mentioned under this heading.
2. PROBLEM IDENTIFICATION AND PROBLEM SOLVING
STRATEGIES
From the perspective of machines, the most important purpose of artificial intelligence is to
especially to enable people to make decisions like humans. decision making or
Problem solving is a difficult task even for humans. The process of finding a cause while solving a problem,
It is the basic characteristic of intelligent behavior. Artificial intelligence and especially expert
problem solving used by artificial intelligence to understand how systems work
It is necessary to know the process and approaches.
2.1. Problem Solving and Decision Making
Problem solving is mostly about the performance outcomes of thinking beings.
Problem solving is the mental activity performed while searching for a solution to a problem. problem term
It's a bit misleading. We often view problems as situations of sadness and danger.
we think. Although this is true in some cases, it is not true in all cases.
For example, analyzing a potential merger is an opportunity search and problem solving.
can be considered as solving. Likewise, researching a new technology is also a problem.
is the solving process. The term problem solving was first used by mathematicians.
used. The term decision making in the business world means problem solving.
is used.
3.2. Problem Solving Process
The definition of the problem-solving process varies depending on the training and experience of the researchers.
For example, Bel et al. found that people rely on their intuition to solve problems and make decisions.
He has put forward several approaches that vary quantitatively. Generally six basic steps
can be observed in the process: detecting and defining the problem, finding a solution
determining criteria, creating alternatives, searching for and evaluating solutions, making choices
and making and implementing recommendations (Figure 2.1). Some scholars differ
They use classifications. For example, there are three phases in Simon's classical approach:
intelligence, design and choice.
Although the process in Figure 2.1 is shown to be linear, it is rarely linear. Real
In life, some of these steps can be combined, some steps can be basic steps or
Revisions can be made to the initial steps. In short, this process is iterative. Each
A brief description of the step is given below.
Step 1: Identifying and Defining the Problem: A problem (or opportunity)
should be noticed. The size and importance of the problem (or opportunity) is determined and
is defined.
Step 2: Identifying Criteria: Comparing possible alternatives to solve a problem
Depends on the criteria used. For example, searching for a good investment requires security, liquidity and returns.
It depends on criteria such as conversion rate. In this step, the criteria and their relative
We determine their importance.
Step 1
Step 2
Step 3
Step 4
Step 5
Step 6
Figure 2.1 Problem Solving Process
Identifying the Problem and
Describing
Application Criteria
Specification
Alternatives
Creating
Solution Search and
Evaluation
Making Choices and
Making Recommendations
APPLICATION
Step 3: Creating Alternatives: According to the definition, in order for there to be a decision situation, two or more
There should also be more opportunity for movement. Creativity and intelligence to create potential solutions
requires.
Step 4: Solution Search and Evaluation: In this step, predetermined criteria
solution options are examined in this light. This step identifies the best or "good enough" solutions.
It's basically a search process as we try to find it. In this step there are several
search, evaluation and cause finding methodologies can be used.
Step 5: Making Choices and Making Recommendations: The result of the search indicates a problem.
It is choosing a solution to recommend as a remedy.
Step 6: Application: The recommended solution to solve the problem is successfully implemented.
should be carried out.
In fact, this process is more complex because each step follows a similar process.
There may be several interim decisions.
Applied Artificial Intelligence technologies to support all six steps
available. But most of the AI moves happen in steps 4 and 5.
In particular, expert systems are used to find solutions from assumed alternatives.
is used. The role of artificial intelligence is basically search and evaluation (itlk) with some
is to manage by using conclusion-drawing abilities. Today, artificial intelligence has limited
Despite its role, after a certain time, technologies play a greater role in the steps of the process.
It is hoped that he will play.
However, artificial intelligence technologies have another big advantage. artificial intelligence
Although it effectively uses only two steps of the problem-solving process, artificial
intelligence, problem solving, and many other tasks not classified as decision making
is used. For example, expert systems help shape computer commands.
They help. Expert systems are also people's "help centers"
To imitate (centers that provide information in catalogs and user manuals)
They are used for planning, complex scheduling, and information interpretation.
2.3. Human Problem Solving Techniques: An Information Processing Approach
The main purpose of artificial intelligence is to understand human intelligence. Human intelligence on computer
Learning how we store and use information when we try to model it
we start. Our thought patterns, reasoning techniques and problem solving
We begin to understand our approaches. We learn how we learn, strong and weak
Our points emerge. As a result, we understand our intelligence better and this makes us better.
It leads us to ways of learning and applying our intelligence to real-world problems.
The science that conducts such studies is called conceptual science, and an interesting area of this science is
It is the way humans and computers process information. The problem when applying artificial intelligence
We consider a specific approach to solving and deciding. This approach is a problem
It is based on the belief that decoding can be thought of as an information processing process.
approach is based on a conceptual approach that uses qualitative descriptions of the ways people think.
is based on. Below is the Newell-Simon theory that describes the information processing process of humans.
The model is explained.
Newell-Simon Model
Allen Newell and Herbert A. Simon explain how computers and humans process information
They proposed a model of human problem solving based on similarity. this model
Helps us understand how artificial intelligence works and what its limitations are
It is possible. The human information processing system consists of the following subsystems: perceptual subsystem, conceptual
subsystem, engine subsystem and external memory. Figure 2.2 shows the memories in each subsystem
and shows the processors.
Perceptual Subsystem
External stimuli are inputs to the human information processing system. These stimuli, our eyes and
It is received by sensors like our ears. The perceptual subsystem consists of these sensors and their
It consists of buffer memories. These memories, perceptual subsystem, conceptual subsystem
It stores incoming data while it is waiting to be used by.
Conceptual Subsystem
Human Senses take up a huge amount of space in buffer memories. To make a decision
When necessary, the conceptual subsystem selects appropriate information. center in computer
Like the processing unit, the underlying processor uses sensor memories to make this decision.
It retrieves information and transfers it to temporary memory. Processors are the "pick-and-run" processors in the computer.
It works in loops similar to . In these cycles, the processor retrieves information from a memory.
It receives, evaluates and stores this information in another memory.
There are three parts in the conceptual subsystem: core processor, volatile memory, and problem memory.
interprets part or all of the program consisting of decoding commands
It consists of an interpreter. This program examines several factors, such as the intelligence and task of the problem solver.
depends on the variable.
In the simplest tasks, the conceptual subsystem only transfers the information from the sensor inputs to the engine.
It works as a unit that transfers data to its outputs. To turn off the light switch
These include routine tasks such as lying down. The movement of the person doing this job is coordinated
It should, but "deep thinking" is still not required for this. Such a
The “thinking” that occurs during behavior cannot be cured.
More complex tasks require more knowledge. More detailed procedures when the time comes.
requires. To accomplish these tasks, the conceptual subsystem uses a second memory system,
will require long-term memory.
Long-term memory is the storage of large amounts of data stored by a complex indexing system.
It consists of symbols. What are the basic symbols and how do they define themselves?
There are competing hypotheses about what they regulate. The simplest memory
In the model, symbols related to each other are combined. In a more detailed model
symbols are arranged in temporary alphabets. According to another view, memory symbol
It consists of sets. A cluster is a unit of stored data. It's a number, symbol
or it may be a word combined with a set of stimuli. Clusters are smaller
They are hierarchically organized collections of clusters. According to this thought
Memory is a large network of clusters.
People can support the decision-making process with another external memory. external memory, a
It includes external mass media such as deck of paper and chalk board. in computers
processing, storing and retrieving data is thousands and millions of times more complex than humans.
It could be faster. People also create probabilistic data,
Their ability to integrate and interpret is limited.
Three memories are shown in Figure 2.2. The essentially unlimited capacity of long-term memory
has. Short-term (temporary) memory is very small. It can only hold 5 or 7 clusters.
However, up to 2 clusters can be kept while doing another job. This is temporary
Indicates that this piece of memory is used for input and output operations. This is a
It is one of the most important limitations of humans compared to computers. temporary memory
limits may be increased. For example, based on similarities, combinations and graphic usage.
can be increased. A graph represents information such as large numbers of pieces of data in several clusters.
may require presence. Thus, charts help managerial decision making.
They play an important role in supporting
Sensors Buffer
memories
External Memory:
Paper, Blackboard
Interpretive Foundation
Processor
Long Term Memory
Short-term
(working) memory
Buffer
memories
Person
muscles
Perceptual Subsystem Conceptual Subsystem Motor Subsystem
Entrance
Stimulants
Output
Answers
Figure 2.2 Newell-Simon Human Information Processing Model (Source: A. Newell and H. A. Simon's Human
Adapted from Problem Solving (Englewood Cliffs, N.J.: Prentice Hall, 1972)
According to this model, humans work serially rather than parallelly. This means;
While a computer can work in serial and parallel designs, humans can only work one at a time.
They can perform information processing tasks.
Engine Subsystem
After scanning and searching memories, the processor sends the information to the engine subsystem.
Motor processors initiate the movements of muscles and other internal human systems. This
Eventually, observable activities such as speech occur.
2.4. Problem Solving in Artificial Intelligence
Applied artificial intelligence technologies are, first of all, the search for problem solving process.
and combined with evaluation steps. First, general search strategies
Let's examine it and then identify those used in artificial intelligence.
Search Methods
Many search methods and strategies to find appropriate solutions to problems
is used. Some of these methods are informal and based on intuition or instinct.
It requires movement. Formal methods can be classified into three categories:
Optimization, blind search and use of heuristics. Blind search and heristic numeric or
requires qualitative (symbolic) analysis, while optimization requires numerical and quantitative analysis.
The categories are shown in Figure 2.3 and then briefly discussed.
Figure 2.3 Formal Search Methods
Optimization
Optimization using mathematical formulas that model a specific situation
tries to find the best possible solution. The problem area is defined in accordance with the rules
should be structured in such a way that optimization can be done by a one-step formula or an algorithm.
is managed. The algorithm is used to generate solutions and test them for possible improvements.
Remember that it is a step-by-step search process.
Call
Methods
Optimization
Analytical
blind
Call
heristic
Full
Arrangement
Partial
Call
Better
produces solutions
or directly to
finds the best solution.
all possible
Answers
Is controlled.
just some
alternatives
Is controlled;
systematic
as bottom
It comes down to solutions.
Only
promised
address solutions
is taken.
Improvement
possible
If not, stop.
All
alternatives
control
when
comparisons
stops.
Comparisons
Solution
good enough
it stops.
optimal
(The best)
optimal
(The best)
Controlled
of alternatives
the best
Good enough
Optimization
Optimization using mathematical formulas that model a specific situation
tries to find the best possible solution. The problem area is defined in accordance with the rules
should be structured in such a way that optimization can be done by a one-step formula or an algorithm.
is managed. The algorithm is used to generate solutions and test them for possible improvements.
Remember that it is a step-by-step search process.
Call
Methods
Optimization
Analytical
blind
Call
heristic
Full
Arrangement
Partial
Call
Better
produces solutions
or directly to
finds the best solution.
all possible
Answers
Is controlled.
just some
alternatives
Is controlled;
systematic
as bottom
It comes down to solutions.
Only
promised
address solutions
is taken.
Improvement
possible
If not, stop.
All
alternatives
control
when
comparisons
stops.
Comparisons
Solution
good enough
it stops.
optimal
(The best)
optimal
(The best)
Controlled
of alternatives
the best
Good enough
Search Process Testing Solution
Where possible, improvement is made and the new solution is subjected to an improvement test.
is kept. This process continues until improvement cannot be made.
Optimization; non-AI, such as business research (management science) and mathematics
It is widely used in technologies. Other analytical quantitative with optimization
The main differences between methods and heristics are shown in Table 2.1. in artificial intelligence
Blind search and heuristic search are widely used.