Cognitivist and Emergent Cognition - An Alternative Perspective Michael James Gratton School of Computer Science and Engineering, The University of New South Wales, Australia mikeg@cse.unsw.edu.au Abstract. A new perspective on classifying cognitive systems is presented in which a distinction is made based on how the world is represented, rather than the typical distinction between cognitivist and emergent approaches. It is argued that the typical classification in essence distinguishes between systems by their implementation, rather than by their properties. The alternative presented here instead focuses on how the system represents the world (if at all) and whether these represen- tations are intelligible to the designer or the system itself. From this novel angle, existing systems are better classified and importantly a gap in existing cognitive systems research becomes evident. An outline of a well-founded cognitive system that fills this space is put forward, one which cognitive robotics is ideally situated to explore. 1 The Embodiment Divide Cognitive robotics research aims to produce robotics systems that approach the level of cognitive ability that we as humans display, exhibiting capabilities such as perception, learning, predicting and taking action, in an autonomous manner. In cognitivist systems such as Shakey [1], cognition is a modular computational process underwritten by the physical symbol system hypothesis [2]. Alternatives such as the subsumption architecture [3], and those based on neural networks, dynamical systems and others were developed later as a response to perceived shortcomings of the classical cognitivist approach. These have been collectively termed emergent systems in surveys of the state of the art in cognitive systems (e.g.: [4]). Cognitive systems are now commonly classified along those lines: cog- nitivist versus everything else. While this distinction may be useful in comparing the implementation approaches of cognitive systems, it does not provide any fur- ther depth of insight into the differences between them. This problem manifests itself as ambiguities in classifying systems by charac- teristics which should properly provide clear lines of distinction. For example, in the survey above, a system’s “representational framework” is one such charac- teristic, with cognitivist systems being described as having “patterns of symbol tokens” as a representational framework while emergent systems instead have a “global system state”. It can be argued however that a collection of patterns of symbol tokens taken together also constitutes a kind of global state, thus this K.-U. K¨ uhnberger, S. Rudolph, and P. Wang (Eds.): AGI 2013, LNAI 7999, pp. 192–195, 2013. c  Springer-Verlag Berlin Heidelberg 2013