Expert Systems Overview.

Expert systems are programs that mimic people's cognitive abilities. They can truly perform at an expert level, or they can do something we might consider less extraordinary but still cognitive.

Introduction to Exp Systems

Mahogany Demonstration

Stress: Built by M/BBA, SILS with no AI background in one semester

Parts of an Expert System:

• Knowledge Base
• Inference Engine
• Interface

Knowledge Base

A set of rules for a particular "domain"

(Also: in Mahogany, objects associated with those rules)

Check: What were the domains in the demos? What are some others?

Inference Engine

Makes logical deductions based upon the facts provided in the KB.

Supports these deductive processes (= Goals)

Magognay demos used back chaining (vs. forward chaining)

Interface

For system building

For run time display

Connections to other systems, databases, ...

Comparison: 3GL and ES

	3 GL	Expert System
Focus	Data structures bits bytes arrays	Knowledge level sensible objects sensible rules
Programming	Algorithmic "while do" order matters exact, unforgiving	Non-procedural Write rules, objects Let I.E. string together

What we'll cover this term

technical techniques:

objects and inheritance

types of searching permitted (in Mahogany)

design good rule based systems

modeling techniques

issues surrounding:

interface

uncertainty

KB verification, testing

integration with other Info Systs

Look ahead

Warning about what ES may never do