- 著者
-
松香 敏彦
- 出版者
- 一般社団法人 人工知能学会
- 雑誌
- 人工知能学会全国大会論文集 第29回全国大会(2015)
- 巻号頁・発行日
- pp.4B1CS1, 2015 (Released:2018-07-30)
The present paper introduces some key theories and phenomena about “ concept ” known among cognitive scientists to researchers in Artificial Intelligence. First, how natural categories are internally represented in our mind is discussed. It may sound counterintuitive, but many behavioral and modeling studies indicate that categories are represented by collection of unsorted exemplars, but not rules nor prototypes. Second, the hierarchical structure of categories is discussed. Although people generally can use taxonomic relations in inferences and reasoning, there are some evidences that category hierarchy is not stored in our memory, but is computed during inferences and reasoning. Third, two theories about symbol systems, namely amodal symbol systems (ASS) and perceptual symbol systems (PSS) are discussed. While a modular and amodal semantic memory is the main vehicle of knowledge in ASS, multi-modal perception, action, and affection in the brain ’s sensory-motor system is the key vehicle in PSS. Although, many theories on categorization and concept in cognitive science are built on the basis of ASS, its limitation and PSS ’s potential advantages are discussed.