Informational inference via information flow

Human judgments about information would seem to have an inferential character. This article presents an informational inference mechanism realized via computations of information flow through a high dimensional conceptual space. The conceptual space is realized via the Hyperspace Analogue to Language Algorithm (HAL), which produces vector representations of concepts compatible with those used in human information processing. We show how inference at the symbolic level can be implemented by employing Bai-wise and Seligman's theory of information flow: The real valued state spaces advocated by them are realized by HAL vectors to represent the information "state" of a word in the context of a collection of words. Examples of information flow are given to illustrate how it can be used to drive informational inference.