Expert systems
Programs built from hand-crafted rules and knowledge that mimicked human experts — the dominant AI approach before machine learning ate everything.
Think of it like
A digital encyclopedia plus a referee: you feed it facts and rules ("if this symptom and that symptom, then likely diagnosis X"), and it walks through them to reach conclusions.
Example
XCON, a 1980s expert system for configuring computer systems, encoded hundreds of rules written by human experts and outearned its development costs many times over.
How it actually works
Expert systems worked by encoding domain knowledge as logical rules. A doctor-approved system might have hundreds of rules like "IF fever AND rash AND recent travel THEN suspect dengue." The system would fire rules in sequence to reach a diagnosis. They were expensive to build (hiring domain experts to articulate their knowledge is slow) and brittle (one missing edge case broke the whole chain). When machine learning proved it could learn patterns from data instead of waiting for humans to articulate them, expert systems fell out of favor.
For product teams
Early AI success story, but the manual knowledge-engineering bottleneck made it unscalable compared to learning-based approaches.
For engineers
A symbolic rule-based inference engine: given a set of IF-THEN rules and facts, it chains rules forward (forward chaining) or backward (backward chaining) to derive conclusions.
Related
- Machine learning — How it contrasts with modern learned models.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome