ABSTRACT
Many different approaches to representing concepts has been suggested, e.g., rule- and exemplar-based descriptions and neural networks. This diversity has led to disagreement about which type of representation is the most appropriate. However, the general opinion seems to be that a single one of these is sufficient to capture most relevant aspects of a concept. This state of affairs might be satisfying if we only wanted to use and learn concepts in restricted domains. However, it is not sufficient when dealing with autonomous agents, natural or artificial, acting in the real world, since they need concepts to serve multiple functions. But which are these functions and how do they influence the representation of concepts?
