The project investigates and implements computational models of cognitive processes involved in creative problem solving, such as dynamic goal setting, rerepresentation of problems, and multiple representations with mutual perception and constraint.
Our model of creative problem-solving has integrated planning, perception and memory aspects, and is highly parallel, and is inspired by human cognition and brain architecture. We will also investigate
the practical applicability of the system to the support of human creativity, and to the use of web-based knowledge. This project is funded by the CreativeIT program of the National Science Foundation.
We believe that our proposed modular processing architecture lends itself naturally to creative problem solving by providing:
- different types of representations which can perceive each other, and mutually interact,
allowing large ("lateral") changes of representation
- a distributed representation of the current situation and of the currently perceived problem, as a set of data structures of different types with representations of desired constraints or goals of different types,
- a planning and metalevel planning regime which allows flexible exploration of alternatives,
restructuring, and selection of representation type,
- an episodic memory facility which allows the history of the problem solving process to be used in planning further problem solving actions.