Induction is a form of reasoning in which premises are viewed as supplying some satisfactory evidence as to the truth of a conclusion or fact. This is opposed to deductive reasoning, from which the truth of a conclusion must follow from premises. The induction of rules or patterns from the relationships between entities in relational databases has seen valuable application in recent years.
The strengths of inductive reasoning when applied within knowledge graphs are two fold. Knowledge graphs operate under the assumption that absent facts are unknown, rather than false. This means that induction-aided rules can help to point knowledge graphs towards missing and likely facts. Secondly, rules can be utilized to point to outliers or cast doubt on unlikely facts that are the exception rather than the rule.