Please use this identifier to cite or link to this item:
Type: Thesis
Title: The Role of Analog Computation in Path Integrating Behaviour of the Desert Ant: A Defence of the Explanatory Credentials of Connectionism in Cognitive Science
Author: Farquharson, Robert Peter
Issue Date: 2018
School/Discipline: School of Humanities : Philosophy
Abstract: What is the relationship between the mind and the brain? Cognitive science is a unique discipline of inquiry that explores this question. Its uniqueness results from a commitment to investigating and understanding cognitive systems as information processing systems. Traditionally, the commitment to information processing has entailed two further positions regarding the nature of “information”, and how exactly it is “processed” in a medium like the brain: representationalism, and computationalism. Taken jointly, these commitments have equipped cognitive science to understand the biological mechanisms by which organisms identify and adapt to variations in their environment. That is, cognitive science is in the business of naturalistically explaining intelligence as the leveraging of internal representations in computational operations towards some goal. Much debate has surrounded the exact manner in which notions like representation and computation are to be cashed out. This debate has mostly centred around two competing accounts: classicism and connectionism. The present thesis is motivated by a desire to fully separate these two competing frameworks, because there is much confusion about whether this is possible. I take an affirmative position on that issue: connectionism is indeed an independent and unique computational framework. Both classicism and connectionism are representational and computational accounts. What separates them is the kind of representations they appeal to, and thus the species of computation they perform over those representations. Connectionism appeals to structural representations, defined over the intrinsic physical properties of the representing medium. Connectionism is thus an analog computational framework: structural representations sustain physical analogies with their task domain. The now ubiquitous ‘classical’ framework, on the other hand, operates over symbolic representations, and is thus a digital computational framework. Much ink has been spilled in assessing the theoretical differences and advantages/disadvantages of the two rival accounts. The present work is, instead, a more hands-on comparative analysis — using an illustrative case study to demonstrate the practical differences in the application of these accounts. A contemporary and influential cognitive scientist of the classical stripe, C. R. Gallistel, has criticised connectionism by (putatively) showing how it can’t explain the rudimentary cognitive phenomenon of Desert Ant path integration and dead reckoning behaviour. I take up this challenge, by 1) showing how Gallistel has misrepresented connectionism as a symbolic computational framework, and then 2) demonstrating exactly how a “true” connectionist account can interpret and model the neurobiology involved in path integration. Using the case study of Desert Ant behaviour, the true appeal and independence of the connectionist framework can be seen, separating it from its classical, symbolic rival. Abandoning symbolic computation, in favour of a structural and analog approach, secondarily shines new light on many theoretical issues that are nested within the modelling and understanding of animal navigation tasks. These have further implications for how we think of animal behaviour, computation, and intelligence more generally. Structural representations and analog computation can handle the explanatory load of the rudimentary example of animal intelligence in the Desert Ant. They also offer up robust and promising lines of inquiry for a naturalistic cognitive science to pursue.
Advisor: Opie, Jonathan
O'Brien, Gerard
Dissertation Note: Thesis (MPhil.) -- University of Adelaide, School of Humanities, 2018
Keywords: Cognitive science
artificial neural networks
mental representation
Provenance: This electronic version is made publicly available by the University of Adelaide in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. This thesis may incorporate third party material which has been used by the author pursuant to Fair Dealing exceptions. If you are the owner of any included third party copyright material you wish to be removed from this electronic version, please complete the take down form located at:
Appears in Collections:Research Theses

Files in This Item:
File Description SizeFormat 
Farquharson2018_MPhil.pdf2.95 MBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.