Specialty Digest

Global English editionLearners’ Study Dashboard — in development
The Knowledge Library for All

Expert System

nounArtificial Intelligencealso knowledge-based system, rule-based system

In one line

An expert system is a computer program that captures the knowledge of human specialists as rules and uses logical inference to give advice in a narrow domain.

An expert system is a computer program that captures the knowledge of human specialists as a set of rules and uses logical inference to give advice or make decisions in a narrow area, such as diagnosing a disease or configuring a computer order. Expert systems were among the first commercially successful forms of artificial intelligence, and they dominated the field in the 1980s. They belong to the tradition of symbolic AI, in which knowledge is written down explicitly instead of being learned from data.

History of expert systems

The first expert systems emerged in the 1960s and 1970s at Stanford University. DENDRAL, begun in the mid-1960s by Edward Feigenbaum, Joshua Lederberg and colleagues, helped chemists infer the structure of molecules from mass spectrometer data. MYCIN, developed in the early 1970s, used a few hundred rules to suggest likely bacteria in blood infections and recommend antibiotics, and it performed comparably to specialists in tests though it was never used routinely on patients. In 1980, a program called XCON, created for the Digital Equipment Corporation, began checking and configuring computer orders and became a landmark commercial success. During the 1980s, companies invested heavily in expert systems and specialized hardware, but the market collapsed in the late 1980s as maintenance proved costly and machines became less specialized, contributing to the second AI winter.

Components of an expert system

A classic expert system has several parts.

  • Knowledge base: the facts and if-then rules supplied by human experts.
  • Inference engine: the reasoning mechanism that applies rules to known facts to derive new conclusions.
  • User interface: the way a person asks questions and receives advice.
  • Explanation facility: a module that shows which rules led to a conclusion, which builds trust and aids debugging.
  • Knowledge acquisition tools: software and procedures used by knowledge engineers to capture expertise.

Separating the knowledge from the reasoning means that specialists can update the rules without rewriting the program.

How an expert system works

The inference engine uses one of two main strategies. In forward chaining, it begins with known facts and applies rules to derive new facts until it reaches a conclusion, which suits monitoring and configuration. In backward chaining, it starts from a goal, such as a possible diagnosis, and works backward to check whether the facts support it, asking the user for missing information along the way. Many systems also handle uncertainty with certainty factors or probabilities, so that a rule can say that a finding makes a conclusion likely, not certain. The result is a chain of reasoning that can be displayed and audited.

Examples of expert systems

Beyond DENDRAL, MYCIN and XCON, expert systems were built for geological prospecting, equipment troubleshooting, loan approval, tax preparation and airline scheduling. Modern descendants are less visible. Rule engines in banking and insurance, clinical decision support alerts in hospital records and troubleshooting wizards in customer support all use rule-based reasoning. Business rule management systems, used to encode policies such as eligibility or pricing, follow the same pattern of explicit rules applied by an engine.

Advantages and disadvantages

Expert systems preserve scarce expertise, give consistent answers and can explain their reasoning, which is valuable in regulated fields. They are also fast to run and do not require large datasets. Their weaknesses led to their decline. Capturing knowledge from experts is slow and expensive, a problem called the knowledge acquisition bottleneck. Rule bases become hard to maintain as they grow, and the systems are brittle: they fail outside their narrow domain and lack the common sense to recognize when a problem is unusual. They also cannot learn from experience unless the rules are updated by hand.

Expert systems vs. machine learning

Expert systems encode knowledge supplied by people, while machine learning systems infer patterns from data. The first approach is transparent and works with little data but depends on experts who can state their knowledge. The second can handle tasks that experts find hard to describe, such as recognizing images, but is often harder to explain. Modern systems frequently combine both, using rules for policy and safety constraints and learned models for perception and prediction.

Building an expert system: knowledge engineering

Creating an expert system is a process called knowledge engineering. A knowledge engineer interviews one or more human experts, observes how they handle real cases and turns what they say into formal rules. This is harder than it sounds, because experts often act on intuition they cannot easily describe. The engineer then builds a prototype, tests it on known cases, compares its answers with the experts’ and revises the rules. Tools known as expert system shells provide a ready-made inference engine and interface, so that teams only need to supply the knowledge base. Validation is continual, since rules must be updated whenever the domain changes, for example when a new drug is introduced or a regulation is revised.

Legacy of expert systems

Although the label has faded, expert systems left lasting marks on computing. They proved that AI could deliver commercial value and brought knowledge representation and explanation into mainstream software design. The ideas live on in business rules engines, clinical decision support and configuration tools. Their weaknesses also taught the field lessons that shaped later work: knowledge is costly to hand-code, systems need ways to learn from data, and explanation is valuable for trust. Today’s approaches, including neuro-symbolic AI, try to combine learning with explicit rules to capture the strengths of both traditions.

Building an expert system: knowledge engineering

Creating an expert system is a process called knowledge engineering. A knowledge engineer interviews one or more human experts, observes how they handle real cases and turns what they say into formal rules. This is harder than it sounds, because experts often act on intuition they cannot easily describe. The engineer then builds a prototype, tests it on known cases, compares its answers with the experts’ and revises the rules. Tools known as expert system shells provide a ready-made inference engine and interface, so that teams only need to supply the knowledge base. Validation is continual, since rules must be updated whenever the domain changes, for example when a new drug is introduced or a regulation is revised.

Legacy of expert systems

Although the label has faded, expert systems left lasting marks on computing. They proved that AI could deliver commercial value and brought knowledge representation and explanation into mainstream software design. The ideas live on in business rules engines, clinical decision support and configuration tools. Their weaknesses also taught the field lessons that shaped later work: knowledge is costly to hand-code, systems need ways to learn from data, and explanation is valuable for trust. Today’s approaches, including neuro-symbolic AI, try to combine learning with explicit rules to capture the strengths of both traditions.

FAQs

Are expert systems still used? Yes, though often under other names such as rule engines, decision support or business rules management.

Why did expert systems decline? High maintenance costs, brittleness and the collapse of the specialized hardware market in the late 1980s, along with the rise of machine learning.

Is an expert system the same as AI? It is one kind of AI, specifically a knowledge-based, rule-driven approach.

Key facts

  • Expert systems separate a knowledge base of rules from an inference engine that applies them.1
  • MYCIN, developed at Stanford in the 1970s, advised on bacterial infections and antibiotic therapy.2
  • Expert systems were the commercial center of AI in the 1980s and declined with the second AI winter.1

Related concepts

In the news

Quick checkWhat are the two core components of a classic expert system?Show answer

A knowledge base holding facts and rules from human experts, and an inference engine that applies those rules to reach conclusions.

Sources

  1. Britannica, Expert system, https://www.britannica.com/technology/expert-system
  2. Wikipedia, MYCIN, https://en.wikipedia.org/wiki/MYCIN
  3. Wikipedia, Expert system, https://en.wikipedia.org/wiki/Expert_system

Editorially reviewed by Specialty Digest Editorial TeamLast reviewed November 17, 2026Researched and drafted with AI assistanceReport an issue