摘要：The present work explores the underlying thought behind symbolic logic which accepts concepts as atomic components, and we introduce a different formalism based on artificial neural networks for the formalization of logical reasoning as a cognitive process, which defines an approach we call subsymbolic logic. We apply this approach to analogical reasoning, which we argue is the proper reasoning. We also explore the cognitive aspects of this approach, especially in isolating and reproducing spontaneous but erroneous forms of reasoning (cognitive biases) which are a part of logical reasoning viewed as a cognitive process. Today, it is the dominant technique in artificial intelligence, but the philosophical aspects of such an approach remain mostly unexplored. To the best of our knowledge, this is the first such attempt at using artificial neural networks to analyse analogical reasoning.
关键词：analogical reasoning; meaning of words; artificial neural networks; neural language models; cognitive connectionism; subsymbolic logic
Loading...
