Expert Systems in Artificial Intelligence — From Rules to Real‑World Impact
Artificial IntelligenceExpert SystemsKnowledge Based Systems

Expert Systems in Artificial Intelligence — From Rules to Real‑World Impact

Shenali Fernando

Shenali Fernando

Editorial Member at SLIIT Mozilla

January 28, 20266 min read

An Expert Systems is a core branch of Artificial Intelligence. It is a type of computer program designed to simulate the decision-making ability of a human expert in a specific domain such as medicine, engineering, or finance etc. An expert system doesn’t think like a human, but it follows if-then rules to make decisions and reach conclusions based on the knowledge stored in its system.

Expert system is not a data & information-based system. It is a knowledge-based system.

Data

Data refers to raw, unprocessed facts and figures that lack context on their own. Data can be difficult to understand or apply without extra processing.

Text, multimedia (pictures, videos, audio), and numerical numbers are formats for data collection.

Information

Information is data that has been processed. The transformation from data to information generally involves several key steps:

  1. Data Collection: Gathering raw data from various sources
  2. Data Cleaning: Ensuring the data is accurate
  3. Data Analysis: Applying statistical methods and algorithms to identify patterns
  4. Data Interpretation: Provide context and explaining what the data signifies.
  5. Data Presentation: Organising the data in charts, graphs, reports etc. to effectively communicate the information.

Knowledge

Knowledge is information that has undergone further analysis, resulting in a deeper understanding. Knowledge builds on information by adding experience, context and judgement, allowing it to be applied to solve problems, develop new products, or create innovative solutions.

Ex:

  • Data: Temperature readings (e.g., 68°F, 22°C)
  • Information: A report showing that the average temperature for the week was 69.8°F (21°C), indicating it was a warmer-than-average week.
  • Knowledge: A gardener’s understanding that plants are damaged when temperatures drop below 59°F (15°C), enabling them to take appropriate action to protect the plants.

Key Components of Expert Systems

  • Knowledge Base: An expert system acts like the brain which stores facts and rules.
  • Inference Engine: It acts as the thinker while applying rules to facts.
  • User Interface: It allows humans to interact with the system through the user interface.
  • Explanation Facility: Explains the reasoning behind decisions. (Medical decisions, Financial Analysis etc.)
  • Developer Interface: Allows experts in the field to update or improve the system.

Inference Engine

An Inference Engine is a logical processing unit that applies reasoning to the knowledge base to make conclusions. It solves problems using “if-then” logic and acts as the brain’s reasoning mechanism. The system uses logical steps to figure out a solution to a problem acting similar to how a human expert would “reason” through a case.

It supports Forward Chaining & Backward Chaining.

  • Forward Chaining (Data-Driven Reasoning): Forward Chaining starts with known facts (input data) and then apply rules until a goal is reached. The system matches the facts with rules. When it finds a rule, whose conditions are all true, it fires the rule. The conclusion is added to the known facts.
  • Backward Chaining (Goal-Driven Reasoning): Backward Chaining starts with a goal and then looks for rules that support this goal. It may ask questions to verify the conditions of those rules.

User Interface

User interface acts like a communication bridge between the user and the system. It can be text-based or graphical.

Developer Interface (Knowledge Acquisition Module)

Developer Interface helps experts and engineers input and update knowledge. It acts like a backstage control panel, where new rules or data are added or removed.

Explanation Facility

The explanation facility of an expert system explains how and why a certain conclusion was reached using its knowledge. This builds user trust and ensures transparency.

Expert Systems in Real World

MYCIN (Healthcare)

MYCIN is an early expert system for treating blood infections. MYCIN would attempt to diagnose patients based on reported symptoms and medical test results. It would also explain the reasoning that led to its diagnosis and recommendation. MYCIN can also recommend antibiotics, with the dosage adjusted for patient’s body weight.

DENDRAL (Chemistry)

DENDRAL is a chemical-analysis expert system. The substance to be analysed might, for example, be a complicated compound of carbon, hydrogen and nitrogen. Starting from spectrographic data obtained from the substance, DENDRAL would hypothesize the substance’s molecular structure.

XCON (Computing)

XCON is an automated configuration of computer systems. XCON was widely adopted across DEC, leading to significant improvements in configuration management and reductions in errors which saved $25 million for DEC. XCON has been implemented in various industries beyond its initial application in configuring computer systems.

Limitations of Expert Systems

  • Expert Systems are built entirely using manual rules created by human experts.
  • Expert Systems cannot learn from data. So, there would be no improvement over time.
  • Some of the typical Expert Systems at times are not able to make available common-sense knowledge and the broad-ranging contextual information.
  • Very narrow range of the knowledge is incorporated in the Expert Systems.
  • Not scalable to complex problems like vision, language, or natural speech.
  • The different types of problems that are faced by the various users while performing the various activities, cannot be efficiently tackled by the Expert Systems.

REFERENCES