In very general terms, encoding means representing information in some fashion that can be used by later information processing. That is, an encoding is a particular form of representation. In artificial neural networks, encoding means choosing some form of code to use to represent stimuli as inputs for a network. In classical cognitive science, encoding means choosing some representational form to store information for later manipulation.
In classical cognitive science, encoding leads directly to the structure/process distinction (Dawson, 1998). That is, when one chooses a particular encoding -- a particular representational format -- one at the same time chooses a set of operations that can be easily used to manipulate this format. As a result, the choice of one encoding makes some problems easy to solve, and others difficult to solve. Simon (1981) presents some examples of how changing an encoding can lead to making problem-solving easier. He goes on to note that "all mathematical derivation can be viewed simply as change in representation, making evident what was previously true but obscure" (p. 153).
- Dawson, M. R. W. (1998). Understanding Cognitive Science. Oxford, UK: Blackwell.
- Simon, H. A. (1981). The Sciences of the Artificial (2nd Edition). Cambridge, Mass.: MIT Press.
(Revised April 2010)