Stream 1: Cognitive Architectures
One of the main aims of this research project is to explore general cognitive architectures, representations and processes that can support flexible intelligence. Although several paradigms have been proposed in the past (such as connectionism and the traditional symbolic approach), they are limited in that they target either symbolic or subsymbolic computation. In contrast, flexible real-world cognition requires a seamless integration between these levels of processing. Take for example the task of changing a light bulb and the reasoning required to select an object on which to stand to gain the appropriate height. Although at some level, the task can be described as symbolic (as it requires a partitioning of the environment into nameable discrete objects) the decision itself is highly influenced by subtle task-related features that are not easily described at a symbolic level such as shape, height, stability and weight-bearing characteristics. As countless examples of human real-world reasoning and problem solving are similarly sensitive to low-level perceptual features, a central component of this research project is to investigate the types of representations and processes that afford the integration between low and high levels of processing, and to implement these ideas in subsequent revisions of the Fluid Analogies Engine (Bolland, 2005).
The Fluid Analogies Engine (FAE)
The Fluid Analogies Engine (FAE) is a general cognitive architecture for perception, problem-solving, analogy-making and creativity. It is unique in that it mixes connectionist representations and algorithms with a symbolic-like production system, dynamic binding and an episodic memory, allowing it to construct and manipulate complex hierarchical structures necessary for high-level problem-solving, but in a way that is flexible and sensitive to low-level subsymbolic features.
Click here to open the FAE applet, allowing you to play with simulations from four different domains, demonstrating various properties of the system:
(1) the Word Superiority Effect - a recreation of the classic IAC network, using the word context to facilitate the recognition of individual letters. Highlights the flexible and context-dependent nature of concepts.
(2) the Numble Domain - a recreation of the Numbo project, displaying problem-solving. In this task, you are given a set of brick numbers (e.g., 20,19,5,3,4) in which to create a target number (e.g., 102), through multiplication, division or subtraction. Each brick can be used once, at the most !
(3) the Jumble Domain - a recreation of the Jumbo project, solving anagrams. This tasks demonstrates a form of creativity, in that statistical regularities found in text drive the construction of word-like candidates (i.e. potentially "creating" novel patterns), in the search for an English word.
(4) letter string analogies - a recreation of Copycat. This task demonstrates the ability to perform analogy-making.