Course Title : Artificial Intelligence

Course Code: BTAIML 502-20

UNIT 5
Understanding Expert Systems in AI
An expert system is AI software that uses knowledge stored in a knowledge base to solve
problems that would usually require a human expert thus preserving a human expert’s
knowledge in its knowledge base. They can advise users as well as provide explanations to
them about how they reached a particular conclusion or advice.
Knowledge Engineering is the term used to define the process of building an Expert System
and its practitioners are called Knowledge Engineers. The primary role of a knowledge
engineer is to make sure that the computer possesses all the knowledge required to solve a
problem. The knowledge engineer must choose one or more forms in which to represent the
required knowledge as a symbolic pattern in the memory of the computer.
Types of Expert Systems in AI
In AI, expert systems are designed to emulate the decision-making abilities of human experts.
They are categorized based on their underlying technology and application areas. Here are the
primary types of expert systems in AI:
1. Rule-Based Expert Systems
 Description: Use a set of “if-then” rules to process data and make decisions. These
rules are typically written by human experts and capture domain-specific knowledge.
 Example: MYCIN, an early system for diagnosing bacterial infections.
2. Frame-Based Expert Systems
 Description: Represent knowledge using frames, which are data structures similar to
objects in programming. Each frame contains attributes and values related to a
particular concept.
 Example: Systems used for knowledge representation in areas like natural language
processing.
3. Fuzzy Logic Systems
 Description: Handle uncertain or imprecise information using fuzzy logic, which
allows for partial truths rather than binary true/false values.
 Example: Fuzzy control systems for managing household appliances like washing
machines and air conditioners.
4. Neural Network-Based Expert Systems
 Description: Use artificial neural networks to learn from data and make predictions or
decisions based on learned patterns. They are often used for tasks involving pattern
recognition and classification.
 Example: Deep learning models for image and speech recognition.
5. Neuro-Fuzzy Expert Systems
 Description: Integrate neural networks and fuzzy logic to combine the learning
capabilities of neural networks with the handling of uncertainty and imprecision
 offered by fuzzy logic. This hybrid approach helps in dealing with complex problems
where both pattern recognition and uncertain reasoning are required.
 Example: Automated control systems that adjust based on uncertain environmental
conditions or financial forecasting models that handle both quantitative data and fuzzy
inputs.
Examples of Expert Systems in AI
There are many examples of an expert system. Some of them are given below:
1. MYCIN
 Overview: MYCIN is one of the earliest and most influential expert systems
developed in the 1970s. It was specifically designed for medical diagnosis.
 Functionality: MYCIN uses backward chaining to diagnose bacterial infections,
such as meningitis and bacteremia. It identifies the bacteria causing the infection by
asking the doctor a series of questions about the patient’s symptoms and test results.
 Significance: Although not used clinically, MYCIN greatly influenced the
development of medical expert systems.
2. DENDRAL
 Overview: DENDRAL is another pioneering expert system, developed in the 1960s,
and is regarded as one of the first successful AI systems in the field of chemistry.
 Functionality: DENDRAL was designed to analyze chemical compounds. It
uses spectrographic data (data obtained from spectroscopy) to predict the molecular
structure of a substance.
 Significance: DENDRAL revolutionized chemical research by automating the
analysis of mass spectrometry data.
3. R1/XCON
 Overview: R1, also known as XCON, was developed in the late 1970s by Digital
Equipment Corporation (DEC) and is one of the most commercially successful expert
systems.
 Functionality: R1/XCON was used to configure orders for new computer systems. It
would select the appropriate hardware and software components based on the
customer’s requirements.
 Significance: R1/XCON streamlined system configuration, saving DEC millions by
reducing errors and improving efficiency.
4. PXDES
 Overview: PXDES is an expert system designed for the medical field, particularly in
the diagnosis of lung cancer.
  Functionality: PXDES could analyze patient data, including imaging results, to
determine both the type and the stage of lung cancer. It helps in deciding the best
course of treatment based on the patient’s specific condition.
 Significance: PXDES aids in accurate, timely diagnoses, improving treatment
decisions in oncology.
5. CaDet
 Overview: CaDet is a clinical support system developed to assist in the early
detection of cancer.
 Functionality: CaDet can identify potential signs of cancer in its early stages by
analyzing patient data and symptoms. It works by comparing patient data with known
patterns and indicators of cancer.
 Significance: Early detection by CaDet enhances survival rates by enabling prompt
treatment.
6. DXplain
 Overview: DXplain is a medical expert system developed at Massachusetts General
Hospital, used as a clinical decision support tool.
 Functionality: DXplain suggests possible diseases based on the symptoms and
findings provided by a doctor. It acts as a reference tool, offering a differential
diagnosis list that doctors can use to check their own diagnoses.
 Significance: DXplain broadens diagnostic possibilities, helping medical
professionals consider rare conditions.
Components and Architecture of an Expert System
1. Knowledge Base: The knowledge base represents facts and rules. It consists of
knowledge in a particular domain as well as rules to solve a problem, procedures and
intrinsic data relevant to the domain.
2. Inference Engine: The function of the inference engine is to fetch the relevant
knowledge from the knowledge base, interpret it and to find a solution relevant to the
user’s problem. The inference engine acquires the rules from its knowledge base and
applies them to the known facts to infer new facts. Inference engines can also include
an explanation and debugging abilities.
3. Knowledge Acquisition and Learning Module: The function of this component is to
allow the expert system to acquire more and more knowledge from various sources
and store it in the knowledge base.
4. User Interface: This module makes it possible for a non-expert user to interact with
the expert system and find a solution to the problem.
5. Explanation Module: This module helps the expert system to give the user an
explanation about how the expert system reached a particular conclusion.
Architecture of an Expert System
How Expert Systems Work?
Expert systems operate by following a structured approach:
1. Input Data: Users provide data or queries related to a specific problem or scenario.
2. Processing: The inference engine processes the input data using the rules in the
knowledge base to generate conclusions or recommendations.
3. Output: The system presents the results or solutions to the user through the user
interface.
4. Explanation: If applicable, the system explains how the conclusions were reached,
providing insights into the reasoning process.
Reasoning Strategies used by Inference Engine
Forward Chaining and Backward Chaining, which are two fundamental methods for
processing information and solving problems in an expert system:
1. Forward Chaining
This is a data-driven reasoning approach where the system starts with the available facts and
applies rules to infer new facts or conclusions. It’s typically used to predict outcomes or
determine what will happen next. An example given is predicting stock market movements.

Forward Chaining
2. Backward Chaining
This is a goal-driven reasoning approach where the system starts with a hypothesis or a goal
(something to prove) and works backward to determine which facts or conditions would
support that conclusion. It’s often used to diagnose issues by determining the cause of an
observed effect. The examples provided include diagnosing medical conditions like stomach
pain, blood cancer, or dengue.

Backward Chaining
Applications of Expert Systems
1. Medical Diagnosis: Expert systems assist doctors by analyzing symptoms and
medical history to suggest possible diagnoses or treatment options. For example,
MYCIN, an early expert system, helped identify bacterial infections and recommend
antibiotics.
2. Financial Services: In finance, expert systems are used for credit scoring, fraud
detection, and investment advice. They analyze financial data and patterns to make
informed decisions.
3. Technical Support: Expert systems can troubleshoot and provide solutions for
technical issues. They guide users through problem-solving steps based on pre-
defined rules and knowledge.
4. Manufacturing: In manufacturing, expert systems help optimize production
processes, perform quality control, and manage inventory by analyzing data and
making recommendations.
Benefits of Expert Systems
1. Consistency: Expert systems provide consistent and reliable recommendations,
reducing the variability that can occur with human decision-making.
2. Availability: They are available 24/7 and can handle multiple queries simultaneously,
providing timely assistance and support.
3. Cost-Effectiveness: By automating expert-level decision-making, organizations can
save on the costs associated with hiring and training human experts.
4. Knowledge Preservation: Expert systems preserve valuable knowledge and
expertise, making it accessible even if the original experts are no longer available.
Limitations of Expert Systems
 1. Knowledge Limitation: The effectiveness of an expert system depends on the
completeness and accuracy of the knowledge base. If the knowledge is outdated or
incomplete, the system’s performance may be compromised.
2. Lack of Flexibility: Expert systems are limited to the rules and knowledge they are
programmed with. They may struggle with novel or ambiguous situations that fall
outside their predefined rules.
3. Maintenance: Regular updates and maintenance are required to keep the knowledge
base current and relevant, which can be resource-intensive.

Representing Domain Knowledge in Expert Systems:

1. Rule-Based Knowledge Representation:
o Most expert systems represent knowledge using production rules or if-then
rules. These rules encapsulate expert knowledge in a conditional format that
the system can use to make decisions.
o Example:
 If the patient has a high fever and sore throat,
 Then the patient may have a viral infection.
2. Semantic Networks:
o Some expert systems use semantic networks to represent domain knowledge
as a graph of interconnected concepts, where nodes represent entities and links
represent relationships between them.
o Example: A medical expert system might represent a disease and its symptoms
as a network where nodes like "fever" and "infection" are connected.
3. Frames:
o Another way to represent knowledge is using frames, which are data
structures for representing stereotyped situations. Frames consist of slots
(attributes) and values (attribute values), allowing the system to store
structured knowledge.
o Example: A frame for a disease might have slots for "name," "symptoms," and
"treatment."
4. Ontology:
o Some advanced expert systems use ontologies to define a formal
representation of the knowledge in a domain, including the relationships
between different entities and concepts.
o Example: An ontology in the financial domain might define concepts like
"assets," "liabilities," and "investments," and the relationships between them.
Using Domain Knowledge in Expert Systems:
1. Problem-Solving:
o Expert systems apply the knowledge from the knowledge base using the
inference engine to solve domain-specific problems. For example, a medical
expert system uses its knowledge of symptoms and diseases to diagnose a
patient.
2. Decision Support:
o Expert systems can assist in decision-making by offering recommendations
based on their knowledge. For instance, in a business context, an expert
system could help make financial or investment decisions by weighing
different factors and providing advice.
3. Learning and Updating:
o While traditional expert systems rely on human experts to update their
knowledge base, more modern systems incorporate machine learning
components, allowing them to refine and update their knowledge based on
new data or feedback.
4. Justification and Explanation:
o Expert systems provide users with explanations or justifications for the
decisions or diagnoses they make, increasing transparency and trust. This is
especially important in critical fields like law and medicine.
Example of an Expert System:
MYCIN (developed in the 1970s) is a well-known early medical expert system that
diagnosed bacterial infections and recommended treatments. MYCIN used a rule-based
approach with an inference engine that employed backward chaining. It would ask the user
(usually a doctor) a series of questions about the patient’s symptoms and medical history and
then apply its rules to suggest a diagnosis and appropriate antibiotic treatment.
Advantages of Expert Systems:
 Consistency: Expert systems can provide consistent decisions based on the same
facts, unlike human experts who may vary.
 Availability: Expert systems can operate 24/7 and be accessible anytime, which is
especially useful in fields like healthcare.
 Knowledge Preservation: They allow the preservation and dissemination of expert
knowledge, which is beneficial in fields where expert knowledge may be scarce.
Disadvantages of Expert Systems:
 Limited Flexibility: They are often limited to the specific knowledge they are
programmed with and may struggle with new or unexpected situations.
  Dependence on Knowledge Quality: The accuracy of an expert system depends
entirely on the quality and completeness of the knowledge base.
 Maintenance: Keeping the knowledge base up-to-date with current knowledge can be
time-consuming and challenging.
shell
In the context of Artificial Intelligence (AI), the term "shell" refers to a software
framework or system that provides the infrastructure to develop expert systems. A shell is a
ready-made environment that includes an inference engine and tools for knowledge
representation but does not contain domain-specific knowledge. Users can add their own
knowledge base (rules, facts, etc.) to the shell to create an expert system for a particular
domain.
Key Features of AI Shells:
1. Inference Engine:
o The core component of an AI shell is the inference engine, which drives the
reasoning process. It applies logical rules to the facts in the knowledge base to
derive conclusions.
o The inference engine can perform:
 Forward chaining (data-driven reasoning) or
 Backward chaining (goal-driven reasoning).
2. Knowledge Representation:
o The shell provides the tools and formats to input and structure knowledge,
such as rules, facts, and relationships. This knowledge can be represented
using:
 Production rules (if-then statements),
 Frames, or
 Semantic networks.
3. User Interface:
o Most shells come with a user-friendly interface to allow non-programmers to
input knowledge and interact with the system. This interface helps the user
define the rules and facts that the system will use for decision-making.
4. Explanation Facility:
o One of the essential features of a shell is its ability to explain how it arrived at
a conclusion. The system can trace back its reasoning steps and provide an
explanation of the decision-making process.
5. Modularity:
 o AI shells are modular, meaning that the knowledge base and inference engine
are separate. Users can add or modify the knowledge base without changing
the inference engine, making it flexible to adapt to different domains.
How an AI Shell Works:
1. Setup: The shell provides an environment for users to input a domain-specific
knowledge base (e.g., rules, facts).
2. Inference Engine: The built-in inference engine uses the rules from the knowledge
base to process the data provided by the user.
3. Reasoning: The system applies forward or backward chaining to derive conclusions
from the given data.
4. Explanation: The shell can explain how it reached a particular decision or conclusion
based on the rules applied.
Examples of AI Shells:
1. CLIPS (C Language Integrated Production System):
o A widely used expert system shell for building AI applications. It supports
rule-based, object-oriented, and procedural programming paradigms.
2. EXSYS:
o A commercial expert system shell used in many fields such as engineering,
healthcare, and business for developing expert systems without the need for
extensive programming knowledge.
3. EMYCIN:
o Based on the earlier MYCIN system, EMYCIN (Essential MYCIN) is a
general-purpose expert system shell that can be used to develop expert
systems in different domains.
4. OPS5:
o An older production system that served as a shell for developing expert
systems by using a forward-chaining rule-based system.
Advantages of Using AI Shells:
 Time-Saving: Developers do not have to write the inference engine or framework
from scratch. They can focus on building the knowledge base.
 User-Friendly: AI shells often provide graphical interfaces or easy-to-use
programming environments to facilitate knowledge input.
 Reusable: Since the inference engine is separate from the knowledge base, it can be
reused across different projects and domains.
Disadvantages of AI Shells:
  Limited Flexibility: Shells are designed to be general-purpose and may not be
suitable for highly specialized or complex applications.
 Dependent on Knowledge Base: The quality and effectiveness of the expert system
depend entirely on the completeness and accuracy of the knowledge base provided.

Knowledge Acquisition in AI
Knowledge Acquisition refers to the process of gathering, organizing, and incorporating
knowledge into an AI system, particularly expert systems. In AI, it involves collecting
knowledge from human experts or other reliable sources and storing it in a format that the
system can use to make decisions, solve problems, or give recommendations.
This process is crucial for building intelligent systems, as the quality of their decision-making
ability depends heavily on the quality and accuracy of the knowledge they possess.
Types of Knowledge in AI:
1. Declarative Knowledge:
o Refers to factual knowledge about the world, such as facts, concepts, and
objects. It can be easily stated or described.
o Example: "The capital of France is Paris."
2. Procedural Knowledge:
o Knowledge about how to perform tasks or solve problems, such as algorithms
or processes.
o Example: A recipe for baking a cake or steps for diagnosing a medical
condition.
3. Heuristic Knowledge:
o Knowledge based on experience, intuition, or expert judgment. It involves
rules of thumb or best practices.
o Example: "If the computer is running slowly, try restarting it."
Steps in Knowledge Acquisition:
1. Identifying the Domain:
o The first step is to define the specific area or domain for which the knowledge
is being acquired. For example, the domain could be medicine, engineering,
finance, or law.
2. Extracting Knowledge from Experts:
o Knowledge is often extracted directly from human experts who possess
specialized experience in a domain. Techniques include:
  Interviews: Directly questioning experts about how they make
decisions.
 Observation: Observing experts while they perform tasks to capture
their knowledge and problem-solving approaches.
 Protocol Analysis: Asking experts to think aloud as they solve
problems, revealing their thought process.
3. Codifying Knowledge:
o Once knowledge is collected, it must be codified or structured in a way that
the AI system can understand. Common methods include:
 Rule-Based Representation: Representing knowledge in the form of
if-then rules.
 Frames: Using data structures to store stereotypical situations or
objects.
 Semantic Networks: Structuring knowledge as a network of
interconnected concepts.
4. Building the Knowledge Base:
o After the knowledge has been extracted and codified, it is stored in a
knowledge base, which is the repository of facts, rules, and procedures used
by the system to reason and make decisions.
5. Verification and Validation:
o The acquired knowledge is tested and refined to ensure its accuracy and
reliability. Experts and system users verify that the system’s decisions are in
line with real-world expert judgment.
6. Updating and Maintaining Knowledge:
o Knowledge acquisition is an ongoing process, as new information or
discoveries need to be incorporated into the system. This ensures the AI stays
relevant and up-to-date.
Knowledge Acquisition Methods:
1. Manual Knowledge Acquisition:
o Knowledge is manually gathered and entered into the system by knowledge
engineers, who work with domain experts to extract, formalize, and structure
knowledge.
2. Automated Knowledge Acquisition:
o AI systems can automatically gather and learn knowledge from data, using
machine learning techniques. This is common in modern AI systems, which
use algorithms to learn patterns and rules from large datasets.
 3. Collaborative Knowledge Acquisition:
o Multiple experts or users contribute to building the knowledge base
collaboratively, often through crowd-sourcing platforms or expert
communities.
Challenges in Knowledge Acquisition:
1. Tacit Knowledge:
o Many experts possess tacit knowledge, which is difficult to articulate. It’s the
type of knowledge gained through personal experience and intuition that
cannot always be easily captured in rules or data.
2. Knowledge Transfer:
o It can be challenging to transfer knowledge from human experts to machines
in a way that preserves the nuances of decision-making.
3. Dynamic Knowledge:
o Knowledge in certain domains changes frequently (e.g., medical research or
technology). Keeping the system's knowledge base updated can be difficult.
4. Knowledge Representation:
o Deciding how to best represent knowledge in a format that an AI system can
efficiently use is a critical aspect of the acquisition process. Too much
complexity can slow down processing, while oversimplification can lead to
errors.