Describing this structure is an interdisciplinary task I commend to philosophers. My aim for now is not to do it -- that's a long term project -- but to describe the task. This requires combined efforts from several disciplines including, besides philosophy: psychology, linguistics, artificial intelligence, ethology and social anthropology.
Clearly there is not just one sort of mind. Besides obvious individual differences between adults there are differences between adults, children of various ages and infants. There are cross-cultural differences. There are also differences between humans, chimpanzees, dogs, mice and other animals. And there are differences between all those and machines. Machines too are not all alike, even when made on the same production line, for identical computers can have very different characteristics if fed different programs. Besides all these existing animals and artefacts, we can also talk about theoretically possible systems.
A common approach to this space of possible 'behaving systems', to coin a neutral phase, is to seek a single sharp division, between those with minds, consciousness, souls, thoughts, or whatever, and those without. Where to draw the line then becomes a major problem, with protagonists of the uniqueness of man, or of living things, or champions of machine mentality, all disputing the location of the boundary, all offering different criteria for allocating things to one side or the other.
The passion accompanying such debates suggests that more than a search for truth motivates the disputants. To a dispassionate observer such debates can seem sterile.
Both sides assume that there is some well-defined concept of 'mind', 'consciousness', or whatever, whose boundaries are to be discovered, not created. But these are complex and subtle concepts of ordinary language, not designed for scientific classificatory precision. When using them of our fellow men, or animals, we don't first check that certain defining conditions for having a mind or being conscious are satisfied. Rather we take it for granted that concepts are applicable, and then we make distinctions between quick and slow minds, conscious and unconscious states, feeling of various sorts, etc. Equally we take it for granted (most of the time) that such concepts and distinctions cannot be applied to trees, lakes, stones, clouds. (However, not all cultures agree on this.) But we don't discriminate on the basis of any precise shared definition of the essence of mind, consciousness, or whatever. For there is no such precise shared definition.
One traditional way to seek an essence is through introspection. However, nothing learnt in this way about the nature of mind or consciousness could help us distinguish other beings with and without consciousness.
Another approach is to seek behavioural definitions of mental concepts: but these founder on the objection that behaviour merely provides evidence or symptoms and does not constitute what are essentially internal states.
The only alternative until recently has appeared to be to locate mind in brain matter - but this ignores important category distinctions: although neuronal states, events or processes may correlate with my being conscious, they are not themselves consciousness. Consciousness is not anything material.
Yet any other attempt to identify a referent for 'mind', 'consciousness', 'pain' etc. has, until recently, looked like an attempt to populate the world with mysterious, inaccessible metaphysically unjustified entities.
What is different now is that Computing Science has provided us with the concept of a virtual machine, within which computational states and processes can occur. A virtual machine has much in common with the kind of formal system studied by mathematicians or logicians. It is an abstract structure which can undergo various changes of state. A virtual machine can be embodied in a physical machine without being that machine. The same virtual machine can be embodied in different physical machines. Different virtual machines can be embodied in the same physical machine. Different virtual machines can have very different abilities. Work in Artificial Intelligence has shown that some virtual machines can produce behaviour which previously had been associated only with minds of living things, such as producing or understanding language, solving problems, making and executing plans, learning new strategies, playing games. By studying the space of possible virtual machines we can replace sterile old boundary drawing disputes with a new, more fruitful, more objective investigation.
First we must abandon the idea that there is one major boundary between things with and without minds. Instead, informed by the variety of types of computational mechanisms already explored, we must acknowledge that there are many discontinuities, or divisions within the space of possible systems: the space is not a continuum, nor is it a dichotomy.
Secondly, we can combine the advantages of both behaviourist and mentalist approaches to the study of the mind. The main strength of behaviourism, in all its forms, is that minds are not static things - it's what they do that is so important. But emboldened by the computational analogy we can see that some doings are external, and some internal: operations within a virtual machine. It is even quite possible for the internal processes to be too rich to be revealed by external behaviour, so that in an important sense external observers cannot know exactly what is going on. For instance, a computer program may be able to print out 'tracing' information reporting some of its internal 'states, but the attempt to trace the internal processes which produce trace printing can lead to an infinite regress. A more interesting example is a computing system with television camera performing complex and detailed analyses on large arrays of visual data, but with limited capacity 'output channels' so that any attempt to report current visual processing will inevitably get further and further behind. Here perhaps is the root of the sense of a rich but inaccessible inner experience which has been the source of so much philosophical argument.
We can attempt a two level exploration of the space of possible minds, one descriptive the other explanatory, though with some overlap between them.
The descriptive task is to survey and classify the kinds of things different sorts of minds (or if you prefer behaving systems) can do. This is a classification of different sorts of abilities, capacities or behavioural dispositions - remembering that some of the behaviour may be internal, for instance recognizing a face, solving a problem, appreciating a poem. Different sorts of minds can then be described in terms of what they can and can't do.
The explanatory task includes surveying different sorts of virtual machines and showing how their properties may explain the abilities and inabilities referred to in the descriptive study.
These explorations can be expected to reveal a very richly structured space - not one-dimensional, like a spectrum, not any kind of continuum. There will be not two but many extremes. For instance one extreme will be simple servomechanisms like thermostats or mechanical speed governors on engines. Another kind of extreme may be exemplified by the simplest organisms.
Among the important divisions between different sorts of virtual machines are the following.
These are merely examples of some of the more obvious discontinuities in the space of possible explanatory mechanisms - virtual machines. Although the descriptions are general and vague, it is already clear how we can design machines which illustrate both sides of each of these distinctions. We don't yet have a full understanding of all the different ways of doing this, nor what their implications are. Moreover, many more detailed distinctions are being explored by computer scientists - distinctions between sorts of languages, sorts of operating systems, sorts of algorithms, sorts of data-structures. Eventually we should have a far clearer grasp of the structure of this space, with some sort of global, generative, description of its contents.
In terms of such mechanisms, we can begin to account for different abilities found in human beings and other animals, as well as constructing machines which display such abilities. What we still need to do is explore which combinations of mechanisms are required to account for the characteristically human abilities which have puzzled philosophers and psychologists and provide much of the motivation for research in Al. A tentative list of such characteristics in need of explanation follows:
(The order is not significant)
(a) the ability to cope with varied objects in a domain
'Object' here is a neutral term, covering such diverse things as physical objects, spoken or written sentences, stories, images, scenes, mathematical problems, social situations, programs, etc. 'Coping' includes such diverse things as perceiving, producing, using, acting in relation to, predicting, etc.
Important special cases include the creation of new domains, and the novel combination of information about several different domains to solve a problem. The more complex examples overlap with what we ordinarily refer to as 'creativity'.
Although there is no artificial computing system which combines more than a few fragmentary versions of these features, and there is no chance of combining all in the foreseeable future, work in AI suggests that provided suitable hardware and software architectures are used, most or all of these features can be explained in computational terms. (This is by no means established, however). There is still a lot more to be done to discover precisely what sorts of computational and representational mechanisms are capable of accounting for what sorts of abilities.
Instead of arguing fruitlessly about where to draw major boundaries to correspond to concepts of ordinary language like 'mind' and 'conscious' we should analyse the detailed implications of the many intricate similarities and differences between different systems. To adapt an example of Wittgenstein's: there are many ways in which the rules of a game like chess might be modified, some major some minor. However, to argue about which modifications would cause the essence of chess to be lost would be a waste of time, for there is no such thing as the essence. What is more interesting is what the detailed effects of different modifications would be on possible board states, possible strategies, the difficulty of the game etc. Similarly, instead of fruitless attempts to divide the world into things with and things without the essence of mind, or consciousness, we should examine the many detailed similarities and differences between systems.
This is a multi-disciplinary exercise. Psychologists and ethologists can help by documenting the characteristics of different types of systems to be found in nature, including the many detailed differences between humans of different ages, and the results of various types of brain damage, which produce systems not normally found in nature.
Anthropologists can help by drawing attention to different sorts of minds produced by different cultural contexts. Linguists and other students of the structures perceived and produced by human minds can help to pin down more precisely what needs to be explained. Computer scientists can help by proposing and investigating detailed mechanisms capable of accounting for the many kinds of features of human minds, animal minds, robot minds. Philosophers can help in a number of ways. They can analyse the many complex implicit assumptions underlying ordinary concepts and thereby help to indicate what exactly it is that we need to explain: for instance those who start from an over-simplified analysis of emotion concepts will over-simplify the explanatory task. More generally, a philosophical stance is needed to criticize conceptual confusions and invalid arguments, and to assess the significance of all the other work. For example, does a computational model of mind really degrade us, as some suggest, or does it reveal unsuspected richness and diversity?
By mapping the space of possible mental mechanisms we may achieve a deeper understanding of the nature of our own minds, by seeing how they fit into a larger realm of possibilities. We may also hope to get a better understanding of the evolutionary processes which could have produced such minds. We will learn that there is neither a continuum of cases between ourselves and a thermostat or amoeba, nor an impassable gulf either.
So much for exhortation. The hard work remains to be done, far more systematically than the philosophical study of language games.
The Meta-Morphogenesis project, inspired by Turing's paper on Morphogenesis
extends this enquiry to include all(!) varieties of information-based control in
organisms between the earliest, simplest, micro-organisms (or prebiotic
structures) and the latest, most complex organisms:
Among other things, shows how long term grief refutes popular theories concerning the nature of emotions, especially theories emphasising embodiment.
Evolution "discovered" that as organisms became more complex, with increasing varieties of independently changing needs, goals, preferences, knowledge, preferences and abilities, related to expanding and increasingly varied spatial regions and temporal periods, the information processing requirements demanded increasingly *disembodied* forms of cognition -- i.e. using detachment from current percepts and actions to meet increasingly complex and ambitious needs.
(Compare how much disembodied cognition is required for an architect designing a new skyscraper using novel materials and designs, or space engineers designing a Mars Rover.)
Likewise, desires, ambitions, fervent hopes, fears, regrets, joys and sorrows, can, in humans, relate to spatially and temporally remote happenings, while unrelated local activities continue as normal.
http://www.cs.bham.ac.uk/research/projects/cogaff/misc/impossible.htmlI suspect those ancient mathematical minds build on spatial reasoning mechanisms that evolved much earlier and are shared with many other species, the main difference being that they are (I assume) unable to notice that they are doing such reasoning or to think about how it works or explain it to conspecifics. (This is also true of many human competences.)
In contrast, the explorations and achievements of biological evolution over billions of years provide a much richer space, though investigating it is a task fraught with difficulties -- requiring a huge amount of highly disciplined creative guesswork.
I suspect that some of the unexplored features discovered and used in pre-human species by biological evolution remain essential for understanding human minds, and for designing artificial minds with human-like abilities -- including abilities of great ancient mathematicians (e.g. Archimedes, Euclid, Zeno and many others). These features, concerned with creative spatial reasoning in many intelligent species, are missing from current AI systems, and also ignored (unnoticed?) by psychologists and neuroscientists. I think they are involved in intelligence of squirrels, elephants, octopuses, magpies, weaver birds, cetaceans, monkeys, apes, etc.
Shanahan follows Wittgenstein in wanting to abolish minds with "private and subjective" contents. Wittgenstein knew nothing about how to design working intelligent systems, despite his engineering background. Computers were not available to him.
Evolution, however, discovered the need for such contents (including what have
variously been referred to as sense-data, qualia, impressions,
sensory-contents, etc.) in increasingly sophisticated animals interacting with
richly varied environments offering multiple possible changes and constraints on
change, perceived from different viewpoints that yield changing information
M.P. Shanahan, 2010,
Embodiment and the inner life: Cognition and Consciousness in the Space of Possible Minds,
Contrast the deep, inspired speculations of Kenneth Craik, e.g. in The Nature of Explanation CUP, 1943.) -- written before digital computers, digital cameras, and other components of modern robots existed -- especially his deep questions about how messy brain mechanisms could represent perfectly straight lines and other geometrical structures. He considers and rejects the idea developed decades later in machine vision systems of a regular array of sensors onto which visual images are projected and array values repeatedly stored (frame-grabbers). But he is clear that there must be some brain mechanisms by which the abstractions of ancient geometry can be represented and used. 75 years later, I don't think anyone knows what they are yet.
In 2003, Ron Chrisley and I explained some aspects of privacy of contents of
experience as resulting from causal indexicality in
A. Sloman and R.L. Chrisley, 2003, Virtual machines and consciousness, Journal of Consciousness Studies, 10, 4-5, pp. 113--172, http://www.cs.bham.ac.uk/research/projects/cogaff/03.html#200302