Address correspondence to

Abstract not found

This paper presents a novel approach to the study of relational categorization based on the evolution of simulated agents in a relational task. In contrast to most previous models of relational categorization, which begin by assuming abstract representations and role-filler binding mechanisms, we begin by studying relational behavior in embodied dynamical agents, which results in a wider range of possibilities for relational mechanisms. The mathematical tools of dynamical systems theory are used to analyze the relational mechanism of the best evolved agent, and we then identify some of the insights offered by this analysis.

Available online at www.sciencedirect.com

We explore one aspect of meaning, the identification of matching concepts across systems (e.g. people, theories, or cultures). We present a computational algorithm called ABSURDIST (Aligning Between Systems Using Relations Derived Inside Systems for Translation) that uses only within-system similarity relations to find between -system translations. While illustrating the sufficiency of within-system relations to account for translating between systems, simulations of ABSURDIST also indicate synergistic interactions between intrinsic, within-system information and extrinsic information.

We evaluate three theories of categorization in the domain of artifacts. Two theories are versions of psychological essentialism; they posit that artifact categorization is a matter of judging membership in a kind by appealing to a belief about the true, underlying nature of the object. The first version holds that the essence can be identified with the intended function of objects. The second holds that the essence can be identified with the creator's intended kind membership. The third theory is called "minimalism" (Strevens, 2001a). It states that judgments of kind membership are based on beliefs about causal laws, not beliefs about essences. We conclude that each theory makes unnecessary assumptions in explaining how people make everyday classifications and inductions with artifacts. Essentialist theories go wrong in assuming that the belief that artifacts have essences is critical to categorization. All theories go wrong in assuming that artifacts are treated as if they belong to stable, fixed kinds. Theories of artifact categorization must contend with the fact that artifact categories are not stable, but rather depend on the categorization task at hand. 3 Psychological essentialism is the hypothesis that object categorization is a matter of assigning kind membership on the basis of a belief about the true, underlying nature of the object. Most of the discussion of psychological essentialism has concerned judgments about naturally occurring entities and their classification into natural kinds. Strevens (2001a) and Rips (2001) both make convincing cases against an essentialist view of everyday categorization for naturally occurring entities. Strevens argues instead for a minimalist view. The minimalist view assumes that categorization is a matter of judging kind me...

Abstract not found

In this paper we proposed a new classification of analogical mechanisms of representational change and gathered evidence of the operation of one of the new ones that we proposed: recategorization of events. We carried out two experiments to assess whether an analogy can trigger the recategorization of a target analog (TA). More specifically, the experiments were designed to test whether a TA not initially regarded as a member of a schema relational category can be perceived as belonging to such category as a result of being paired with a base analog (BA) consisting of a typical exemplar of that

Although many researchers have suggested that compositional concepts should be sensorimotor grounded, the method to accomplish this remains unclear. This paper introduces a second-order neural network with parametric biases (sNNPB) that learns compositional structures based on sensorimotor time series data. The data was produced by a simulated robot that executed distinct object interactions (moveto and orient-toward). We show that various sNNPB setups can learn to compositionally imitate object-interactions beyond the interactions that were specifically trained, which was not possible with previous neural network (NN) architectures, including recurrent neural networks (RNNs). We also show that these imitation capabilities are accomplished by developing a self-organized, geometrically-arranged compositional concept structure in the PB values and task-oriented, Braitenberglike sensory encodings in hidden sensory layers. Since second-order connections were necessary to accomplish the task, we hypothesize that such connections may be essential to drive the learning of both sensorimotor-grounded compositional