Expert Systems Questions and Answers

Answer: A

The knowledge base stores the domain expertise as rules, frames or other structures, separate from the machinery that uses it. The inference engine applies that knowledge, working memory holds the facts of the current case, and the explanation facility reports the reasoning afterwards. This separation of knowledge from control is the defining architectural commitment of the field, and it is what makes a shell reusable across domains. Most errors here come from carrying a result across from a neighbouring case where it does not hold.

Answer: B

The inference engine matches rules against working memory, resolves the resulting conflict set and fires the chosen rule. The knowledge base merely stores rules and imposes no order of its own. The interface handles dialogue and the acquisition module handles editing. Keeping the control strategy in the engine is precisely what allows a knowledge base to be written declaratively, without its author thinking about execution order at all. Working one numerical case through both cases side by side fixes the distinction better than any rule of thumb.

Answer: C

Working memory holds the assertions specific to the current consultation and is emptied between cases, while the knowledge base persists. The distinction between long-term knowledge and short-term facts mirrors human problem solving and is what makes the same rules serve every case. A rule index is an implementation detail and the trace is a by-product. Confusing the two stores is the commonest beginner's error in this topic. A single measurement chosen to respond to the one asymmetry beats several that respond to shared behaviour.

Answer: D

A production rule pairs a condition part matched against working memory with an action part that asserts, retracts or acts. It is deliberately less expressive than full predicate logic, which is what keeps matching efficient. Frames, implications and semantic networks are alternative representations covered elsewhere in this topic. The modularity of rules, each independently meaningful, is the property that made knowledge bases maintainable.

Answer: A

Forward chaining is data-driven: it fires whatever rules the current facts enable and keeps going until nothing more can be concluded. It suits monitoring and configuration, where the data arrive first and the goal is unknown. Backward chaining works the other way, and abduction and means-ends analysis are different strategies again. Matching the direction to the shape of the task is the first design decision in building any rule-based system.

Answer: B

Backward chaining takes a hypothesis, finds the rules that would conclude it and tries to establish their conditions in turn, asking the user when nothing else will settle them. It suits diagnosis, where the candidate conclusions are few and known in advance. Forward chaining, search order and conflict resolution are separate notions. The characteristic question-asking style of a diagnostic consultation falls directly out of this control regime.

Answer: C

When data arrive continuously and any of many conclusions might follow, running rules forward from the data is natural, which is why process control and configuration systems chain forward. A short candidate list instead favours backward chaining, since each hypothesis can be tested cheaply. Single readings and yes-or-no queries are goal-directed. The rule of thumb is that few facts and many goals chain backward, many facts and few goals chain forward.

Answer: D

Conflict resolution picks one rule from the conflict set using strategies such as recency, specificity, priority or refractoriness. Matching produces the set; resolution chooses within it. Truth maintenance handles retraction and acquisition handles authoring. Because the choice can change which conclusion is reached, the strategy is part of the system's semantics and not merely an efficiency device. A limit that survives making every external component ideal is an intrinsic one, and worth knowing for that reason.

Answer: A

Recency fires the rule matching the most recently asserted working-memory element, which keeps attention on the current line of reasoning rather than wandering. Specificity prefers the more detailed rule, refractoriness stops a rule firing twice on the same data, and explicit priorities are set by hand. Most engines combine several of these in a fixed order, so knowing what each contributes is needed to predict behaviour. Working one numerical case through both cases side by side fixes the distinction better than any rule of thumb.

Answer: B

Refractoriness removes an instantiation from the conflict set once it has fired, so the same rule with the same bindings cannot loop forever. Without it, a rule whose action does not change its own conditions would fire endlessly. Recency, specificity and file order address different questions. It is the least visible of the strategies and the one whose absence is noticed immediately.