Indirect Representation

Indirect Representation

Indirect representations are surrogates for something else. Indirect representations are “third-person” representations. They represent things from the third person “outside-in” perspective of an external observer. The referent of an indirect representation is whatever the representation represents. Indirect representations take many forms. Among the most developed are first order predicate logic and mathematics. Other types of indirect representations include computer programs, and various types of knowledge representations. A knowledge representation (KR) is most fundamentally a surrogate, a substitute for the thing itself, used to enable an entity to determine consequences by thinking rather than acting, i.e., by reasoning about the world rather than taking action in it [Davis et al, 1993]. As far as we know, no other species creates indirect representations at the high level of abstraction of homo sapiens. As far as we know, no other species creates persistent indirect representations, i.e., writing, or some equivalent. However, other species do communicate and the act of communication entails the use of indirect representations, so other species are capable of creating transient indirect representations. Human beings often encode indirect representations symbolically as information. However, indirect representation does not have to be symbolic. Indirect representations can be encoded in a wide variety of forms for communication.


A block diagram that shows how indirect representation represents things is shown below. This diagram represents the same relationships that were shown in the diagram of direct representation. Thing1 is related to Thing2 by relation R1. Thing1 is also related to Thing3 and Thing4 by relation R2. However, in indirect representation, all the representation is indirect. Thing1, Thing2, Thing3, Thing4, R1 and R2 are all represented indirectly. The actual representation of the thing that Thing1 refers to is located outside the representation of Thing1 itself. Thing1 and its representation are two different things. The same is true of every other thing and every relation that is represented. The advantage of this type of representation is that the definitions of things only have to be stored once, and then multiple instances of those things can refer to the same definition. This saves storage. However, the disadvantage is the representation is unencapsulated, and extra representation needs to be added if we want to make sure the representation is complete and consistent. In order to define the consistency and completeness conditions, an external intelligence must know what is to be represented and it must decide how to represent it. Contrast this with direct representation. For a direct representation, no intelligent observer is required to define consistency or completeness constraints nor do the constraints need to be represented explicitly. Instead they are represented implicitly by the ontology of the representation.