CCNCS Seminar Details
Is there a future in architectural modelling of mind?
|Date/Time:||Wednesday 30 March 2011, 4.15pm|
This presentation will be a dry run of a talk to be given at a festschrift in honour of Professor Phil Barnard at the Cognition and Brain Sciences Unit in Cambridge, who is retiring. Architectures of mind are broad scope theories, which propose explanations across many cognitive functions, such as, language, short term memory, long term memory, affect, attention, decision-making, etc. Phil Barnard's interacting cognitive subsystems is exactly such a 'framing' architectural theory of mind. If one is going to realise such architectures in a computational model, then one needs to handle the issue of scale. Such theories are, necessarily, large. I will discuss how methods employed in computer systems engineering could be used to enable architectural theories of mind to be modelled. A particular focus will be on structural decomposition and abstraction. It is notable that architectural theories of mind seem to uniquely arise in the symbolic tradition, e.g. SOAR, ACT-R, EPIC, even, Shallici's supervisory attentional system. In contrast, the pre-eminence of cognitive neuroscience, has brought neural network modelling to the fore. However, neural network models are pretty well exclusively, specific, highly focussed, and not architectural in scale. I will discuss why this state of affairs has arisen and consider whether architectural theories can have a role in the age of cognitive neuroscience.