The focus of my interest is on the organization of knowledge in a problem solver's memory. The aim is to explore the schematic forms in which a planner's knowledge is organized, which allow planning to be done in an abstract space. Plans for any larger goal that one has are often arrived at by suitably adapting and assembling well known plans to achieve smaller but more recurrent subgoals. This is specially relevant when one is working in an uncertain environment. The memory structures should allow situation specific knowledge to trigger solution specific sequences of actions. Neural and symbolic networks may offer insights into fine grained representation.
The granularity of knowledge could vary from a rule to a complete plan, as is done in case based reasoning. The level of abstraction is another important issue. Greater abstraction will lead to more compact structures, but may may miss out on some crucial detail in some cases. It is well known that the knowledge of experts vis-a-vis a novice includes both more abstracted structures as well as a larger collection of exceptions.
A related issue is that of natural language communication. My
motivation for looking at language is as a knowledge communicator, so
that we can build systems that can teach, as well as learn, using