Journal

Volume 17, Issue 1 (March 31, 2016)

5 articles

• Abstract Mechanisms and Neural Computation
  by Abel Wajnerman Paz
  J. CS. 2016, 17(1), 1-26;
  Abstract A characterization of computation and computational explanation is important in accounting for the explanatory power of many models in cognitive neuroscience. Piccinini (2015) describes computational models as both abstract and mechanistic. This approach stands in contrast to a usual way of understa... [Read more].
  Abstract A characterization of computation and computational explanation is important in accounting for the explanatory power of many models in cognitive neuroscience. Piccinini (2015) describes computational models as both abstract and mechanistic. This approach stands in contrast to a usual way of understanding mechanism which implies that explanation is impoverished by abstraction. I argue that in order to provide a useful account of computational explanation in cognitive neuroscience, Piccininiʼs proposal must be complemented by an abstraction criterion that fulfills two conditions: motivating abstractions enough to make a model computational and not motivating the omission of information that is constitutive of mechanistic explanation. These conditions are relevant because although there are computational and mechanistic descriptions of neural processes (Piccinini & Bahar 2013) mechanism must, as a normative theory, determine whether the abstractions that these models involve are well motivated. I argue that the abstraction criterion proposed by Levy and Bechtel (2013) is a promising candidate to fulfill these requirements. First, I show that this criterion can legitimize the omission from recently proposed neurocognitive models of all features that are non-computational according to Piccinini’s approach (although it also motivates some modifications of his characterization of neural computation). Second, I argue that this criterion legitimizes those models only if we interpret them as including all the information constitutive of mechanistic explanation. [Collapse]
• Insight Grammar Learning
  by Emilia Garcia-Casademont & Luc Steels
  J. CS. 2016, 17(1), 27-62;
  Abstract We report on computational experiments in which a learning agent incrementally acquires grammar from a tutoring agent through situated inter-actions. The learner is able to (i) detect impasses in routine language pro-cessing, such as missing a grammatical construction to integrate a word in the rest... [Read more].
  Abstract We report on computational experiments in which a learning agent incrementally acquires grammar from a tutoring agent through situated inter-actions. The learner is able to (i) detect impasses in routine language pro-cessing, such as missing a grammatical construction to integrate a word in the rest of the sentence structure, (ii) move to a meta-level to repair these impasses, primarily based on semantics, and (iii) then expand or restructure his grammar using insights gained from repairs. The paper proposes a cog-nitive architecture able to support this kind of insight learning and tests it with a computational implementation for a grammar learning task. [Collapse]
• Understanding Consciousness as Data Compression
  by Phil Maguire, Philippe Moser, & Rebecca Maguire
  J. CS. 2016, 17(1), 63-94;
  Abstract In this article we explore the idea that consciousness is a language-complete phenomenon, that is, one which is as difficult to formalise as the foundations of language itself. We posit that the reason consciousness resists scientific description is because the language of science is too weak; its p... [Read more].
  Abstract In this article we explore the idea that consciousness is a language-complete phenomenon, that is, one which is as difficult to formalise as the foundations of language itself. We posit that the reason consciousness resists scientific description is because the language of science is too weak; its power to render phenomena objective is exhausted by the sophistication of the brain’s architecture. However, this does not mean that there is nothing to say about consciousness. We propose that the phenomenon can be expressed in terms of data compression, a well-defined concept from theoretical computer science which acknowledges and formalises the limits of objective representation. Data compression focuses on the intersection between the uncomputable and the finite. It has a number of fundamental theoretical applications, giving rise, for example, to a universal definition of intelligence (Hutter, 2004), a universal theory of prior probability, as well as a universal theory of inductive inference (Solomonoff, 1964). Here we explore the merits of considering consciousness in such terms, showing how the data compression approach can provide new perspectives on intelligent behaviour, the combination problem, and the hard problem of subjective experience. In particular, we use the tools of algorithmic information theory to prove that integrated experience cannot be achieved by a computable process. [Collapse]
• The Delineation of ‘Throw’ Verbs in Mandarin Chinese: Behavioural and Perceptual Approaches
  by Helena Hong Gao, Haoshu Wang, & Elena Nicoladis
  J. CS. 2016, 17(1), 95-131;
  Abstract Within a semantic domain, terms that can be used in a similar way to describe a similar event are members of the same class of words, or near-synonyms. They are common in a language but difficult to distinguish from one and another. Physical action verbs such as ‘throw’ verbs are a typical example o... [Read more].
  Abstract Within a semantic domain, terms that can be used in a similar way to describe a similar event are members of the same class of words, or near-synonyms. They are common in a language but difficult to distinguish from one and another. Physical action verbs such as ‘throw’ verbs are a typical example of this. In this study we attempted to distinguish six Chinese ‘throw’ verbs (rēng, diū, pāo, tóu, shuāi, shuǎi) from each other within the framework of cognitive semantics. Two experiments were conducted with two groups of native Chinese speakers (60 participants in total) to examine their behavioural and perceptual responses to the throwing actions that can be typically described by each of the six verbs. The results show that the verbs the participants enacted revealed differences in terms of dimensional features. Further, visual input about the verb enacted, successfully elicited the participants’ responses corresponding to the semantics of each individual verb. Typical actions and differences between five dimensions were used as discriminative features of the verbs. The validity of action performance as a paradigm for verb meaning specification was verified. [Collapse]
• The Syntax of Adverbials in Ewe-English Codeswitching
  by Evershed Kwasi Amuzu
  J. CS. 2016, 17(1), 133-165;
  Abstract This paper discusses adverbial switches (single-word adverbs, adverb phrases, temporal noun phrases, prepositional phrases, and adverb clauses) in bilingual clauses in Ewe-English codeswitching, spoken in Ghana. The data analysed come mainly from two databases created in 1996 and 2002 respectively. ... [Read more].
  Abstract This paper discusses adverbial switches (single-word adverbs, adverb phrases, temporal noun phrases, prepositional phrases, and adverb clauses) in bilingual clauses in Ewe-English codeswitching, spoken in Ghana. The data analysed come mainly from two databases created in 1996 and 2002 respectively. It is found that while English adverbials occur in Ewe-based clauses, Ewe adverbials do not occur in English-based clauses. Furthermore, it is found that the English adverbials are allowed only in the four positions in which Ewe adverbials may occur. They do not occur in clause-medial adverb positions (e.g. the position in-between the subject-NP and the VP and in the VP-internal position) that are found in the English clause structure but not in the Ewe clause structure. The main significance of these findings is that although adverbial switches constitute an amorphous syntactic and semantic category, their distribution is consistent with principles outlined in Myers-Scotton’s Matrix Language Frame model that have been found attested in the distribution of English nouns, verbs, and adjectives in Ewe-English codeswitching (cf. Amuzu 1998, 2005a, 2014a, 2014b, 2015) and in codeswitching elsewhere in West Africa (Amuzu 2005b[2010], 2013a, Quarcoo 2009, Vanderpuije 2011, and Bolaji et al 2014). Also, the findings support other scholars’ studies of adverbial switches (e.g. Treffers-Daller 1994 and Hebblethwaite 2010) with respect to which of the languages in codeswitching contact may serve as the source of adverbial switches in bilingual clauses: the source is consistently the more socially dominant language, the embedded language from the perspective of the Matrix Language Frame model, in this case English.  [Collapse]