Background

Rule-based systems (RBSs henceforth) are a simple but successful AI technique. This technology originates in the early days of AI (mid-twentieth century), when the intention was to create intelligent systems by manipulating information. One of the key issues encountered was knowledge representation (KR), discussed in depth during Chapter 9, "Specifications and Knowledge Representation." How should facts about the world (or problem) be described to allow intelligent reasoning?

Inspired by work in psychology, human reasoning is understood as a characteristic behavior that can be modeled. Rule-based systems model human reasoning by solving problems step by step, as experts apply their experience. Knowledge is stored in a highly implicit representation. (That is, many facts need to be inferred.)

In practice, RBSs are extremely widespread. Many domains benefit from them, ranging from medicine to the manufacturing industry, including tech support. Many fields profit from deductive systems manipulating data according to simple rules. As such, they constitute one of the major successes of classical AI. There are literally tens of thousands of them, falling into various subcategories of knowledge-based systems, as shown in Figure 11.1.

RBSs are also known as production systems, but we'll avoid the term as it is less intuitive and often ambiguous within the industry domains it is applied to. Generally speaking, RBSs fall into the category of knowledge-based systems, dealing with the processing of information (for instance, sensory data in our case).

Specifically, the examples in this chapter as well as most other RBS—are expert systems. This means that the rules within the system have been crafted by a domain expert. This flattering word is used quite loosely; almost any solution where humans directly encode their wisdom and experience passes as an expert system.