A Dynamic Approach to Context Modeling

This paper presents the issues of dealing with context from the perspective of cognitive modeling. A dynamic theory of context is proposed which considers context as the set of all entities that influence human cognitive behavior on a particular occasion. As a consequence context is thought of as the dynamic fuzzy set of all associatively relevant memory elements at a particular instant of time. These memory elements might be both mental representations and operations. Some experimental facts about the influence of the perceptible environment as well as of the previous memory state on human problem solving are briefly presented. The dynamic nature of context influence on behavior is emphasized. A general cognitive architecture, DUAL, is presented which consists of many small agents running autonomously in parallel with variable speeds depending on their current associative relevance. A model of problem solving, AMBR, based on DUAL is discussed where problem solving emerges from the collective behavior of the agents. The possibilities of AMBR for modeling context and priming effects are considered and some simulation results are presented.