Keywords: Cognitive Science, Psychology, Artificial Intelligence, Consciousness

Title: Mind and Mechanism

Author/Artist: Drew McDermott

Publisher: MIT Press

Media: Book

Reviewer: Pan
Computationalism, roughly put, is the view that mental processes, such as intelligence and perception, are essentially computational processes. In Mind and Mechanism Drew McDermott, extends the computationalist view to tackle the fundamental problem at the heart of philosophy and psychology: what is consciousness? In what is a wide-ranging and thought-provoking book, McDermott attempts to show that consciousness itself is the result of the computational processes in our brains. Simply put, he attempts to show that the same kind of processes that we can create in computer hardware and software can create the complex and wide-ranging phenomenon that is human consciousness.

This radically materialist viewpoint leaves no room for metaphysical concepts such as the soul, the spirit or other non-material phenomena. The body/spirit dualism that has been a mainstay of European philosophy for hundreds of years is at once abolished. What this means for religion is obvious, but it's a subject that the author comes back to later in the book.

At the core of McDermott's theory is an internal mental model which is aware of itself. Using this self-model, he examines what it is to have experiences, what it is to have an awareness of self and others, and how computation (i.e. figuring out) is at the heart of what it is to be human. He also shows, quite rightly, that if computation can explain away consciousness and intelligence, then it makes little difference whether the computations are performed in 'brain-stuff', software, computer hardware or bits of Lego. In this respect the 'strong AI' dream of creating a fully conscious and intelligent machine is certainly within reach, even if current technology has still got some way to go.

The sections of the book where McDermott expands on his 'self-model' and gives concrete examples of some of the computational processes which he feels are essential components of consciousness, make for interesting but necessarily technical reading. However, it must be said that his view of consciousness (and intelligence) is essentially passive. It is known that vision, for example, is a dynamic process of searching, processing and interpretation rather than a passive process of image scanning, but this is not reflected in his account.

Similarly his computationalist view of emotion is very passive, whereas it seems clear that emotions touch every aspect of mental and physical processes in an active and dynamic way. It is impossible to think without experiencing emotion, and while he works hard to give an explanation of emotions it is simplistic and unconvincing. In the words of Susan Greenfield:

...our perceptions are unified wholes, shot through with memories, hopes, prejudices, and other internalised cognitive idiosyncrasies [The Human Brain, 1997]
In part McDermott is unconvincing because he treats the semantic categories of emotion (fear, regret, love, hate etc) as discrete, separable categories whereas what we experience are broad swathes of emotion which we struggle to name. It is here that one would expect to see the difference between symbol-based approaches and connectionist (neural network, for example) approaches more fully explored, but he explicitly states that there is no dichotomy because most neural networks are modelled in software. By the same token one can argue that there is little meaningful difference between discrete (countable) and continuous (measurable) phenomena because digital technology can represent continuous phenomena such as sound and speech signals.

This reliance on discrete categories (or symbols) is a recurrent theme, and the existence of mental modules is also taken as a given, despite the fact that, as he readily admits, our knowledge of the physical brain is still very primitive. It is unclear what these modules are; innate physical structures, 'black-box' theoretical constructs? There is also little discussion of the developmental aspect of these modules and computational processes. However these possibly lie outside the scope of McDermott's project, though they are interesting and salient questions nevertheless.

While it is an ingenious and well-argued theory, and McDermott an able and inventive exponent of it, computationalism presents a curiously passive, disembodied and static view of intelligence and consciousness. It should be no surprise therefore, that there is an overlap between computationalism and evolutionary psychology.

Perhaps the most surprising section of the book is towards the end. Having dealt a blow to dualism at the beginning, McDermott sneaks it in again by the back door. In discussing the ethical and moral implications of his theory, he makes it plain that he cannot conceive of a system of ethics that does not depend on fear of a divinity to enforce it. He then asks us to make a leap of faith and accept, without proof, that God exists and has 'poured himself into the universe'.

Having created an argument that we are biological computers, creatures of flesh, blood and algorithm, he turns away from the consequences and seeks solace in the arms of a divine universe. It is as though McDermott is pulled in one direction by the force of intellect and another by his religious beliefs.

Despite this unexpected direction, the book remains a challenging and informative read, and represents a concise and well-argued case for the computationalist view of one of the hardest problems in cognitive science.

Hit the 'back' key in your browser to return to subject index page

Return to home page