Knowledge & Reasoning›Core· 35 min read

Knowledge & Reasoning (Expert Systems)

Store facts and IF–THEN rules, then let the computer combine them to reach new conclusions.

What you will learn

Represent facts and rules
Build a tiny rule-based “expert system”
See the strengths and limits of rules

Facts + rules = reasoning

A knowledge-based (or expert) system stores two things: facts about the situation, and rules that say how to draw conclusions. The computer then applies the rules to the facts to work out something new.

Fact: “the patient has a high temperature.”
Rule: “IF high temperature AND cough THEN possible flu.”

A tiny symptom checker

Let us encode a few medical-style rules and reason over a set of symptoms — a classic expert system.

A rule-based expert system: it fires every rule whose conditions are met

facts = {'fever', 'cough'}          # what we know about the patient

rules = [
    ({'fever', 'cough'},        'possible flu'),
    ({'sneezing', 'runny nose'},'possible cold'),
    ({'fever', 'rash'},         'see a doctor'),
]

def diagnose(facts, rules):
    found = []
    for needed, conclusion in rules:
        if needed.issubset(facts):   # are all required symptoms present?
            found.append(conclusion)
    return found or ['no rule matched']

print(diagnose(facts, rules))

Note: Output: ['possible flu'] The patient has fever and cough, so the first rule fires. Add 'rash' to the facts and the “see a doctor” rule fires too.

Strengths and limits

Strengths	Limits
Easy to read and explain	A human must write every rule
Predictable and reliable	Cannot handle messy, fuzzy data well
Great when rules are known	Breaks on cases nobody wrote a rule for

Expert systems are still used (tax software, safety checklists). But writing rules for everything — like recognising a cat in a photo — is impossible. That limit is exactly why machine learning took over: instead of writing rules, we let the computer learn them from data. That is Unit 4.

Tip: A rule “fires” when all its conditions are true. Real expert systems chain rules together — one rule’s conclusion becomes a fact another rule uses.

Q. What is the main limitation of a rule-based expert system?

Answer: Expert systems rely on rules written by people. For tasks like image recognition you cannot write all the rules — so machine learning is used instead.

✍️ Practice

Add a rule: IF sore throat AND fever THEN “possible strep”. Test it.
Build a 3-rule expert system that recommends what to wear based on weather facts.

🏠 Homework

List one real product that uses hand-written rules and one that clearly must learn from data. Explain the difference.