Embodied concept formation and reasoning via neural-symbolic integration

  • Authors:

  • Min Jiang;Changle Zhou;Shuo Chen

  • Affiliations:

  • Department of Cognitive Science, Xiamen University, Xiamen, Fujian Province 361005, China;Department of Cognitive Science, Xiamen University, Xiamen, Fujian Province 361005, China;Department of Cognitive Science, Xiamen University, Xiamen, Fujian Province 361005, China

  • Venue:
  • Neurocomputing
  • Year:
  • 2010

Quantified Score

Hi-index 0.01

Visualization

Abstract

Modern cognitive science [1] indicates that concepts stem from individual experience, which more concretely means that an agent's concept system is generated by interactions between an agent's body and the environment it lives in. In this study we present an approach that will enable Artificial Brains to generate embodied conceptual systems, including a sophisticated introspection mechanism that will allow them to transcend their initial conceptual limitations. Our approach is based on extensions to formal concept analysis. We use incomplete formal contexts to represent the sensorimotor information of the ''body'' of an Artificial Brain, and then use uncertain formal concept analysis as a mathematical tool to settle various problems related to embodied concept formation. After proving some theorems, we show that 3-valued Lukasiewicz logic is the right instrument for our purpose, overcoming the shortcomings of the existing methods. We also describe how to use neural-symbolic integration to allow this sort of approach to provide not only advanced AI functionality but also approximate simulation of aspects of human brain function.