COGNITIVE SYSTEMS, Volume 5, issue 1
Contents and Abstracts
Representations, beliefs, and hindrances in understanding pp.1-16
The implicit/explicit distinction in connectionism and symbolic artificial intelligence pp.17-35
A. Micarelli and P. Boylan,
Foreign-language tutoring systems today: Old-fashioned teaching with newfangled gadgets pp.37-56
Grammar, semantics, and logic: Towards an 'ordinary language' theory of language pp.57-80
Spatial representation, connectionism, and cognitive development pp.81-102
University of Geneva
Representations, beliefs, and hindrances in understanding
An analysis is presented of meaning as it may be supposed to be represented in the mind. This analysis is rooted in different levels of natural logic, which was in the first place developed in relation to the types of logic employed by children. More specifically, a number of "molecular transformations" are discussed: focusing, dissociation, activations, differentiation, and expansion. Cognitive representations may be characterised by a number of general properties, like prototypicality, amalgamation, egocentrism. Other important properties are the ability to create mental representations of one's own representations, of other person's representations, and of imaginary objects, notably ones with undesired properties (occultation). The natural logics on which mental representations are based, may be incomplete, to different degrees, especially in children. To cope with specific problems of daily life, different evaluations may be employed, to make up for lack of inferential rules. These may include examples, (in)convenience, previous experiences, especially if they led to success earlier, metaphors, and others. There are many other factors which may influence the mental processes in relation to problem-solving, an analysis makes it clear that there may be large discrepancies between formal systems of logic and reasoning, and actual practice. An important role is played by intentionality, for which various definitions have been proposed in the history of the study of mental processes. Brentano was the first to emphasize its importance.
LEIBNIZ-IMAG-CNRS, Grenoble, France
The implicit/explicit distinction in connectionism and symbolic artificial intelligence
Because connectionism is often opposed to symbolic AI as processing implicit rather than explicit representations, we investigate the nature of the implicit/explicit distinction. On closer examination, this conceptual distinction turns out to be relative to a given interpreter, and thus of limited use to characterize either connectionism or AI. Yet the existence of a similar and well-documented psychological difference between forms of memory leads us to redefine the distinction as one between bound and free knowledge, which points to a serious shortcoming of today's connectionist models. They cannot reuse their context-bound knowledge as easily as symbolic AI systems, and therefore lack the open-ended reflectivity of human conscious thought. However, this is not an insurmountable limitation of connectionist systems, and we outline research directions to remedy the present deficiency of neural networks.
Alessandro Micarelli (1) and Patrick Boylan (2)
Universitý di Roma Tre, Italy:
(1) Dipartimento di Discipline Scientifiche - Sez. Informatica,
(2) Dipartimento di Linguistica
Foreign-language tutoring systems today: Old-fashioned teaching with newfangled
Researchers and practitioners in the field of language instruction largely agree today that proficiency in a foreign language requires more than grammatical competence: it requires a kind of know-how they call "communicative competence", of which grammatical competence is but a subset. The present paper claims that, although this view of language learning has been prominent for over a generation (Hymes seminal paper "On communicative competence" dates back to 1966), it has had little impact on Computer Assisted Language Learning (CALL) projects. Indeed, CALL systems regularly use the most recent and technologically advanced hardware/software resources to implement the most out-dated grammar-based pedagogy; thus, the total instructional value of such systems is undercut by the very didactics they (often quite ingeniously) embody. In the present paper, a case is made for "Conversation Rebuilding", a recent trend in "communicative teaching" that has been implemented in a text-based CALL system built around a Common LISP Artificial Intelligence engine on a MacIntosh platform. It is argued that a system of this kind is likely to give more instructional value than systems using more sophisticated interfaces or architectures but which embody more traditional language-teaching methods.
Francis Y. Lin,
Somerville College, Oxford, England
Grammar, semantics, and logic: Towards an 'ordinary language' theory of language
When we master a language we not only know which sentences are grammatical, we also understand the meaning of the sentences, and are able to reason in the language. Thus, grammar, semantics, and logic should not be separate domains of inquiry; rather, they should be an integrated whole. Language, with Wittgenstein, consists of language games. The rules of language should be known to the speakers, the former should guide the latter's linguistic behavior. Most of the existing studies of language treat grammar, semantics, and logic as isolated topics, and the rules they postulate are only tacitly known to the speakers. In direct contrast, the present paper offers a new conception of grammar, semantics, and logic. It starts from words and their use, which every speaker knows, and then on this basis constructs an account of grammar, semantics, and logic. In the process several important language phenomena, e.g. grammatical competence, semantic competence, and reasoning in ordinary language, are explained, and this is done in a unified way. This paper also contributes to 'ordinary language' philosophy in that it applies it to some central problems in linguistics and philosophy, and provides it with relevant details.
University of Toronto
(now at Northeastern Univ., Boston, USA)
Spatial representation, connectionism, and cognitive development
Connectionism represents a serious challenge to the view that non-linguistic knowledge is organized as mental imagery or mental simulation. In this paper, the cognitivist models of spatial representation are first presented and then contrasted with connection explanation. Unlike mental symbols, it is argued, connectionist representations can be described as developmental processes based on vectorial superposition. Two resulting characteristics, interaction and integration between components, offer the opportunity to study neural networks as hierarchical systems of stochastic actions. At the macro-level, the networks can be described as adaptive systems that self-organize behavioral patterns supporting recognition and production of spatial configurations. At the micro-level, the cognitive components of recognition can be explained with neural mechanisms of activity selection (activation unbinding) and activity combination (activation binding) conveying spatial information.
|ESSCS||european society for the study of cognitive systems|