A Prototype-Exemplars Hybrid Cognitive Model of “Phenomenon of Typicality” in Categorization: A Case Study in Biological Classification

In this paper we introduce a new machine learning system developed to investigate the phenomenon of typicality in categorization processes of human mind. Experimental researches in cognitive psychology produced a lot of evidences that categorization is a process based on similarity between observed instances and representative instances, which constitutes the concept acquired by human mind. Conversely, the main theories developed till now, prototypes theory and exemplars theory, are unable to fully explain the experimental findings about the “phenomenon of typicality”. The classifier algorithm, introduced here, is characterized by an adaptive concepts description based on intermediate abstracted items varying from prototypes to exemplars and so it is able to subsume prototypes and exemplars theories. We test it on a well known biological classification problem. The experimental results show this classifier can be considered as a typicality theory