Artificial intelligence (AI)

A vague 'a computer that thinks or acts like a human', with no characteristic, scores nothing. Mark schemes reward the characteristics: a collection of data + rules, the ability to reason, and the ability to learn/adapt. State one of these rather than a general 'human-like' phrase.

The most-lost marks are on the roles of expert-system components. Candidates can name the knowledge base, rule base, inference engine and interface but cannot say what each does - especially the inference engine (it applies the rules in the rule base to the facts in the knowledge base to reason towards a conclusion). Give each component as name + role.

Asked what machine learning means, vague answers ('the computer learns') score little. The mark-scheme idea is that the program can automatically adapt its own processes and/or data (improve from more data) without being reprogrammed. Examiners noted candidates could pick the missing terms in a fill-in but could not then explain machine-learning capabilities.

The inference engine is the reasoning part of an expert system, not a store. Describing it as 'where data or rules are kept' is wrong - the knowledge base stores facts and the rule base stores rules. The inference engine applies the rule base to the knowledge base to reach a conclusion, often with a probability.

Do not treat the knowledge base and rule base as the same thing, or swap them. The knowledge base holds the facts about the subject; the rule base holds the logical IF-THEN rules that link those facts. The inference engine uses both - facts from the knowledge base, rules from the rule base - to reason.