Originally developed for use as a psychological model of human memory, semantic nets are used in AI to describe relationships between concepts, objects, or events. The relationships between these elements (also called nodes) are represented by links called arcs. These arcs can describe a multitude of relationships between the elements, depending on the implementation of the semantic net, such as "A is a B", "A is part of B", "A belongs to B", etc.

[diagram of semantic net with labels to illustrate terminology]

Common properties are often shared between two objects connected by an arc. For example, if a semantic net describes a relationship "Dogs are a type of mammal", the properties of dogs can be inferred from the properties of animals, such as: Mammals are living things, mammals are warm-blooded, mammals take care of their young, etc. Such relations thus establish a property inheritance hierarchy in the semantic net. Elements lower in the hierarchy can inherit properties from elements higher up in the hierarchy, saving space since information is not duplicated among the elements in the hierarchy.