The great conjuring trick of all time

Undoubtedly, one of the main unsolved intellectual puzzles of our time is how thought works. With the advent of the computer to serve as a metaphor for mind and as a virtual machine on which to try and realise a mind, the study of how thinking works is paramount. As David Gelernter states "it encompasses philosophers of mind, cognitive psychologists, neurophysiologists and legions of frantically intense computer scientists bent on carrying off the greatest conjuring trick of all time, building minds out of computers''.

According to Gelernter the whole enterprise of understanding thought has had limited success because it has only considered thought to be reasoning. He argues that this has to a large extent been understood by philosophers and psychologists and can be "faked'' by computers. However, there is another part to thought which has nothing to do with reasoning and is based around intuition, insight, creativity and metaphoric thinking. This is thought that is centred on the drawing of analogies where two superficially unrelated thoughts are brought together and bring about a new metaphor.

It is this kind of thought that drives creativity, which wells up from within, and takes hold of the artist in their deepest creative moments. It is this process of analogical thinking that artificial intelligence and cognitive science does not understand because it considers the mind to be based around formal logical reason. Gelernter wants to construct a new folk psychology that sees the human mind as a spectrum going from high focus logical thinking at the high end to low focus analogical thinking at the other. High focus thought is analytic and penetrating and deals in abstractions. Almost all attempts to simulate thought on a computer have used this narrow high focus band at the top of the spectrum. As we move down the spectrum thought becomes less penetrating and spreads out. The connections between thoughts are less based upon logical properties than on emotional bonds. The seemingly random jumps that occur between thoughts are based upon "affect linking'' and come about when "two recollections engender the same emotion, and they only happen towards the low focus end of the spectrum''. It is affect linking that causes creativity, metaphor and in some cases, spiritual mind-states to emerge.

For Gelernter, the transition from low to high focus thought can be seen over different time scales. Over the course of a day focus moves from high to low, throughout childhood focus moves from low to high and throughout human evolution the focus of thought moves from low to high. Children have short attention spans and seem to jump from one idea to another without any apparent connection. Dreams work like this as well, as do ancient texts, particularly the Bible, from which examples are drawn.

Low-focus thought is specifically associated with creativity. Creativity depends on restructuring and finding unexpected analogies. Unexpected connections are made between different memories because these memories share very similar emotional content. This means that thought at this level is not goal-directed but "generates its own agenda''. Gelernter sees the mental modelling of cognitive science as being peripheral to the mainstream low-focus thought. Mental models are seen as "the interruption, and traffic with memory is the steady, defining beat''. Emotion is the glue of low-focus thought. Furthermore, emotion cannot exist without a body. Emotional states are also bodily states which means you cannot have thought without a brain and a body. This goes directly against the entrenched notion in cognitive science that mind is to brain as software is to computer.

What then are the implications of this theory of mind and creativity for mainstream cognitive science? Gelernter states "my aim isn't merely to bulldoze cognitive science and wash my hands of the whole mess; rather, to point the computer based study of mind in a new and in some ways fundamentally different direction''.

It is not at all clear that the binding features of thought move along a spectrum from high focus rule-based logic to low focus creative emotion. If we consider mathematics or chess for example, which by Gelernter's definition are extremely high focus, are no creative insights made when we are thinking very analytically? At the other end of the spectrum, dreams which must be the most extreme example of low focus thought, do seem to follow a kind of logic. Children, who are supposed to be good examples of low focus thought are also synchronously capable of extremely logical, penetrating thought. The point is, the very nature of thought cannot be reduced or separated into the component parts of logic and emotion. In fact, the kind of high focus thought that Gelernter describes, sounds much more like the goal directed rule based "thought'' of traditional symbolic artificial intelligence, designed for computers, rather than human thought. The inseparable nature of emotion, intellect and will is what characterises human thinking. Recent research has shown an extensive interconnection in various structures and processes that must be involved in the actual operation of thought, feeling and will.

Gelernter has raised an important issue in cognitive science but his solution falls within a paradigm that has been grappling with this problem for a long time and has not found a satisfactory solution. In focussing on the way in which emotion has been neglected in modelling human thought he is correct, but by proposing an "emotion function'' which takes memories as input and outputs an emotion, he is remaining within a reductive paradigm that is ultimately highly problematic. Our experience of the world is unified and value laden. We do not have an experience and then attach an emotional value to it.

It is the case that in trying to understand creative thought cognitive science has relied on mental models of creative processes that are driven by logical problem solving, or as Gelernter would say high focus thought, and has to a great extent neglected the emotional and bodily aspects of thought. It is only in the past few years that there has been a move away from modelling disembodied thought processes to modelling embodied agents that can only "know'' about the world by being in it. This research is mainly concerned with how sensory motor neuronal architectures can adapt to changing physical conditions in the environment, but is also concerned with how values, preferences and emotions are an integral part of thought and action.

Creativity has been described in terms of the exploration and transformation of conceptual space. Most ways of exploring a space are uninteresting and aesthetically unappealing. If the exploration of a conceptual space is not driven by a bias towards logic or emotion but is based upon a unified notion of thought, then there must be some higher order cognitive processes at work.

However, to concentrate on the emotional content of creative thought is most certainly a promising way to proceed within the current paradigm. The notion that creative processes can be reduced down to their component parts within a problem domain and then have causal-logic links established between them, has been shown to have limitations when implemented in symbol-based computer programs. Although the output of such programs may be considered adequate it is certainly wide of the mark when compared with human creativity.

This book highlights a huge problem in cognitive science, and suggests a way forward within the traditional paradigm. David Gelernter has written an informed intelligent, witty and provocative book that should be essential reading for anybody interested in computers, creativity and cognition. He covers key philosophical issues in an interesting and lucid way. There is a paradigm shift going on in cognitive science at the moment and this book highlights one of the reasons for it.

Paul Hodgson is a musician and composer and a research fellow at the University of Sussex School of Cognitive and Computing Sciences.