What is a rule-based system?

• Consists of a rule-base (permanent data); an inference engine (process); and a workspace or working memory (temporary data). Not part of the basic reasoning process, but essential to applications, is the user interface.

  In the cognitive models, the rule-base is usually equated with LTM and the workspace with STM. There will be restrictions on these to correspond with assumptions about mental architecture: e.g. limited size of STM.

• Knowledge is stored as rules in the rule-base. (Also known as the knowledge base. Rules are of the form

  IF some condition THEN some action

  The condition tests working memory, e.g. for the presence of certain symbols or patterns of symbols. In many systems, the conditions are expressed logically as conjunctions (occasionally, disjunctions) of predicates. In some systems, some conditions correspond to sensor data. In a cognitive model, this would correspond to direct access to sensory input, rather than to its representation in STM.

  The action can be one of the following:

  • Another symbol or set of symbols to be added to STM. In many systems these will be expressed logically as conjunctions of predicates.

  • Some other action on STM, e.g. ``delete the symbols XyZ''.

  • Some other action, e.g. turning a motor off, printing. In a cognitive model, this would correspond to a motor command.

• The inference engine applies the rules to working memory. There are various ways of doing this: see later. Why so called? See Crevier p 157.

• The user interface sits between the inference engine and the user. It translates the system's answers from an internal representation to something the user can understand; it passes questions from the system to the user and checks the replies (rejecting, for example, a negative number as the answer to a request for your weight).

• The components are independent, in that (as long as they are compatible), you can apply the same inference engine to many different knowledge bases. Crevier pp 152, 156.