Acangelosi

This user has publicly declared that they have a conflict of interest regarding the Wikipedia article Alberto Greco.

Alberto Greco

Alberto Greco is a Professor of Psychology at the University of Genoa, Italy. His research interests are in the field of cognitive psychology of categorisation and representation, and connectionist modelling.

Education

He studied philosophy at the Universita' Cattolica Sacro Cuore Milano. He worked under the guidance of Professor Agazzi, a pioneer in computational modelling, philosophy and psychology.

Research

Concept of Representation

The concept of representation in psychology. This is one of his most important research contributions. Greco distinguishes between the process of representing and what is represented. The process of representing does not consist in managing existing representations but in constructing them, creating them. These representation exist between the stimulus and the response it serves to explain. It works in two ways: in the case of symbolic representations the explanations derive from the interpretation of the symbols, in the case of non-symbolic representations from the mere existence of representations. The most innovative contribution regards the identification of different functions of the representation: replacement (reactivate past contents no longer present or anticipate future contents not yet present) and correspondence (change of internal status corresponding to change of external status). In the 1997 update he described these functions even more in detail and proposed the idea of the transition from the correspondence function to the symbolic substitution function through the simulation of the symbol grounding. He also contributed to a discussion about what neural networks can explain.

Language and action

Greco studied whether the representation of gestures has a holistic or compositional nature. To do this he created a paradigm which associated nonsense words with sensorimotor patterns. In a first investigation, there was a (compositional) condition in which a sentence was composed of three nonsense words and each of these words was associated with an aspect of a motor pattern (action, part of the body, body side: e.g., beating the left fist) and another (holistic) condition in which the entire pattern was associated with a single word. The result of those experiments seemed to support the hypothesis of a holistic representation. A subsequent study, performed similarly but with more arbitrary motor patterns, instead shown an advantage in the compositional condition, but only when the feature relevant to the composition (in this case the position of the hand) was easily discriminable and critical to distinguish between the different motor patterns. This result seems to support the idea that sensorimotor discrimination is a preliminary step for the composition. This conclusion was supported by a new experiment comparing difficult and easily discriminable (integral and separable) stimuli.

Symbol Grounding

Alberto and collaborators (Stevan Harnad and Angelo Cangelosi) proposed a connectionist model of symbol grounding in neural networks. This was subsequently used by Harnad for evolutionary models of language origins.

Categorization

In the debate on the classic models of categorization (rules, prototypes, examples) he supported the thesis that these models are not alternative but explain different aspects of this process. Greco showed that different categorization methods can be favored (by similarity, by attributes and by prototypes) through the appropriate manipulation of the stimuli to be categorised. In a recent publication with Moretti I studied the methods of analytic and holistic categorization, introducing a paradigm in which at the beginning there are biased examples (in which the salient features are not relevant), that are subsequently gradually eliminated. In addition I created a new method of analysis of the categorization process, the Active Feature Composition Task, which replaces the traditional task of classification with a task in which it is required to "build" actively the examples to be tested by combining features.