the dungeon has...

Mind-as-machine

Lakoff objects in chapter 19 of Women, Fire and Dangerous Things to what he calls the mind-as-machine paradigm.

He describes the mind-as-machine paradigm as claiming that ‘the mind is a computer with biological hardware and runs using programs essentially like those used in computers today. It may take input from the body and provide output to the body, but there is nonetheless a purely mental sphere of symbolic manipulation that can be characterised in terms of algorithms of the sort used in computer programs’ (p338). He argues that the results of cognitive semantics appear to undermine or even contradict this view.

Lakoff very quickly points out that he is not contradicting AI in general. For which he gives the following reasons:

A lot of AI does not claim to be of empirical relevance to the study of mind, e.g. expert systems and robotics.

The arguments that he gives do not contradict connectionism (because the role of the body in cognition fits naturally in neural networks – and is not considered a ‘disembodied algorithmic program’)
Often computational approaches only claim to be simulating some limited domain of cognition – providing useful evidence about cognition but not specifying how the mind works in reality. He says: ‘a computer can simulate the flow of water in a river. That does not mean the river itself is directed by any algorithm’ (p345).

He says that cognitive semantic studies of categorisation contradicts and mind-as machine approach that assumes/claims the following:

¨ The mind is separate from the body and consists purely of disembodied algorithms. I.e. that ‘every cognitive process is algorithmic in nature; that is, thought is purely a matter of symbol manipulation’ (p339) and that ‘human reason is completely abstract and not dependant in any way on human bodily experience’ (p340).
¨ Any objectivist approach to AI which claims the following:

1. Concepts are internal representations of an external reality – i.e. given meaning in virtue of their relation to the world
2. A Universal Conceptual Framework – meaning that there is a neutral cognitive framework in terms of which every thing can be represented (relates back to discussion on relativity).

He argues that all these positions tend to be conjoined because it is the only way to explain how computational symbols have meaning and allow for the understanding of people with different conceptual frameworks.

Lakoff says that cognitive models are not ‘internal representations of external reality’ because not only are they understood in terms of ‘embodiment’ rather than a direct connection to the external world, but also because they include things like metaphor and other ‘imaginative’ forms of cognition, which cannot be viewed to relate to the external world. Especially famous for arguing against an objective representational position is Putnam with his Twin Earth example of Twater. Lakoff says that this stems from the disembodied mind position which forces ‘an objectivist understanding of computational models’ (p342) because from a ‘gods eye view’ the link between the representation and reality must in some way be a correct link. This entails that:

¨ There must be links between the language used in the computational model (e.g. Mentalese or LOT) and the external world.
¨ These links must be accurate and ‘really’ represent the world.

This is counter to the twin earth example.

Lakoff says that under the mind-as-machine view learning another cognitive system would involve a translation into your own language of thought. He had argued in the previous chapter that there ‘is more to understanding another conceptual system than translation’ (p344). (If example is needed see p344 last paragraph).

He also argues against any form of arbitrary symbol manipulation based on a formal logic because he claims that human conceptual and linguistic systems are motivated to some degree and in different respects. The reason for this is that ‘it is easier to learn something that is motivated than something which is arbitrary’ (p346), whereas a mind-as-machine view would not agree because ‘In most algorithmic systems, having something extra in memory just uses up more, not fewer computational resources’ (p347).

Lakoff concludes that from this the mind-as-machine paradigm is ‘hopeless’ (p351), and that although the computational metaphor for the mind is important, useful and interesting, it has limitations (which have been discussed).

TrackBack URL for this entry: http://bluejoh.com/cgi-bin/mt/mt-tb.cgi/384