MacroscopicExplanations

Introduction

"The explanations of animal societies offered by biologists are essentially reductionist. That is, they attempt to explain the structure of societies as a consequence of the properties of the individuals which compose them. By no means all sociological or anthropological theories are of this kind. Theories in economics are reductionist... But many sociological theories are not reductionist in even this limited sense. The properties of the individual are seen as produced by society, and even as serving the purposes of that society, and not the other way around." - JohnMaynardSmith

The quote comes from the introduction to a book on problems in sociobiology. And it neatly stakes out the territory that a science such as biology lays claim to when it attempts to explain social behaviour. The contrasting attitudes in Maynard-Smith's quote can also be seen as two views on agency. The reductionist sees agency as the emergent effect of multiple microscopic components. But Smith's sociologist sees the mind moulded by its environment and those macroscopic entities of which it is itself a sub-component. In fact, social scientists are aware of the tension between these two interpretations of the relationship between the individual's behaviour, and the social structure which he or she inhabits. They label these positions structure (or holistic or top-down) and action (or individualistic or bottom-up).

What is interesting about social science is that it looks very like cognitive science played in the mirror. Cognitive science concentrates on two levels: the first is the causally mechanistic, physical level of the brain; the second, the psychological, intentional level of the mind. The task facing engineers of artificial intelligence and cognitive philosophers is to make these two levels compatible; to explain the relationship between them in a way which fits the mind into natural science, but does not ride roughshod over our intuitive feelings about it. The mind is seen as emergent out of, or supervenient upon, or in some way above the mere mechanistic level.

By contrast, a social scientist, or at least one who falls into the holistic or structural camp, also sees two levels - one of which is mechanistically causal, the other of which is intentional. But here, it is the higher, macroscopic level which obeys causally mechanistic rules, and the low level which is intentional. Within the field, there are also fierce debates between those who see the mechanistic level as being causally responsible for the intentional, and those who wish to defend intentionality from macro-reduction to these principles. Emile Durkheim, for example, asks us to accept that

"society is not a mere sum of individuals. Rather, the system formed > by their association represents a specific reality which has it's own > characteristics."

This superstructure so constrains the behavioural opportunities of the individual that it is seen as essentially guiding (or causing) that behaviour.

Within social sciences, there has been a robust defence against this way of thinking. From J.S. Mill in the 19th century to modern individualists, many are unhappy with the possibility of macroscopic structures being causal of microscopic properties. Two possible refutations seem to run as follows:

The first is to argue that these high level structures have no goals of their own. They merely inherit their seeming intentionality from the intentional humans who comprise them.

But against this, the holist believes that structures may be acting according to their own agendas. No one wants an economic recession or a bout of inflation; perhaps no one intends a moral norm to collapse. Yet these things happen. How is this to be squared with the idea that the intention to behave has come from the low level? One solution is the great man theory. That somewhere, there is someone who secretly did want the change and, this time, has got his way. Cognitive scientists may recognise this as a parallel to the idea of the grandmother sensor[^1] or undischarged homunculus.

A second strategy individualists use against structuralists is to ask directly: how does it work? How can this high level system cause these low level people to do things? The answer demanded is a low level causal one: a story of who did what to whom. When presented with such, they can then argue that, this is obviously, only a story about individuals and their behaviour. In other words, this is a reductionist argument of the kind which those cognitive scientists who wish to preserve the mind from being mere brain behaviour, are continuously fighting against.

These are rough parallels, but nevertheless, our first glance into the distorting mirror of social science, has been enlightening. Because the mirror reverses the levels that we normally consider intentional and non-intentional, it might help us understand the opposing viewpoint to our usual one.

From our intuitions about folk psychology, we tend to reject the dogma of the holistic social scientist. But from our intuitions as cognitive researchers, about both homuncular decomposition, and reductive materialism, we are tempted to deny the arguments which the individualist social scientists use to deny high level structures. Any argument which seems to deny high level structure in favour of individual action, is dangerously close to one which would collapse the psychological into the neurochemical. By contrast, any argument which could pump enough hot air into the mind to keep it afloat above the brain, could probably launch a few leviathan like macroscopic entities.

There are, of course, further, arguments which separate out these two situations. Interlevel relationships come in many subtle flavours and it is possible to argue that the relation between neural and psychological is of a different kind than the relation between mental and structural.

But in this paper I intend to accept the possibility of these social structures and argue that not only can they be useful for cognitive scientists, but that we are already beginning to use them. First though, I will explore a little further what they are, and what claims they have to existing in the first place.

Living in Structure

Natural science provides views at multiple levels of abstraction upon the physical world. Quantum mechanics, classical atomic physics and chemistry are familiar examples. Each level provides a powerful, simple theory in terms of the relationships between different entities or solids. Of these, some are seen as candidates for being real, to use Dennett's terminology: illata; while others, although useful are considered mere abstractions or abstracta. [@intStance]

There are two fundamental philosophical questions which can be asked about these illata. The first is, "Do they really exist?" and the second is "How can we know them?". To introduce some terminology which will be familiar to some and mysterious to others, the first of these is a question of ontology, and the second epistemology. Martin Hollis [@philSocSci] thus summarizes the structural or holistic claim, attributed to Durkheim, like this:

• "an ontology of 'social facts', forming an order external to individual consciousness and not explicable by reference to human nature.
• a methodology wherein social facts are explained by their function 'in relation to some social end.'
• functional mechanisms working through the medium of the 'collective consciousness' and connecting social ends to the overall level of social integration needed if a society is to flourish.
• an epistemology, so far undisclosed, which warrants our > subscribing to these components."

These "social facts" or illata, were traditionally seen as economic or political groupings such as class, or state. And they were believed to have particular interactions with each other which could be captured by scientific rules, regardless of the individuals who composed them. The relationship between high level structure and component may be subtle and convoluted. For example, at one point, Durkheim makes a claim which, though counterintuitive, illustrates the sophistication of the relationship he is promoting. The State is a powerful, self supporting, system[^2] and criminals are not its enemies, but an essential component, which helps hold it together. (Presumably because common outrage at criminal acts encourages people to defend the state against them.) Were people to abstain from acts currently considered criminal, the State would begin to redefine other acts as taboo, to restore the balance within itself, just as living bodies balance the proportions of their components.

Another idea of this genre which deserves note, is that of ideology which is defined as false consciousness. For some holists the high economic level of description is seen as the true level of illata; and a viewpoint of an individual who sees the world only in terms of individuals and their personal beliefs and desires is considered misguided. Individual consciousness is the intersection of falsehoods invented by various high level illata such as Church and State as part of their homeostatic behaviour.

Here again, is a distorted inversion of a familiar idea. For some cognitive scientists, consciousness is very nearly treated as a falsehood; a narrative spun by the brain, possibly out of multiple contradictory threads of perception.

Some Recent Parallels

Not surprisingly, extreme holistic social science is often rejected, and many of the theories which were created in the nineteenth century and flourished in the early part of this century, are discredited. But has the idea of social structure really died out? One interesting recent area of study is Game Theory, which has been explored by von Neumann, and others over the last half century. Extraordinarily, game theory is held up as an individualist analytic tool and a rival to structure based explanations. It does start with the idea of an individual, who is rational. In other words, the individual will predictably chose to behave in a way which, as she perceives things, will maximise her gain in a particular situation. But then such situations are classified as being instances of one of several abstract games such as the Prisoner's Dilemma or Chicken. Thus the game provides a framework which is itself a kind of law.

If agent X is rational then agent X will perform action A.

Some social scientists see game theory as being in opposition to traditional holism, but the parallels seem much stronger. There is not such a great difference between saying that agent X is in a situation defined as a Prisoners Dilemma and should therefore do A; and saying that agent X is in a situation defined as a Class War and should therefore do A. Both allow an observer to tell the same kind of explanatory stories about why X performed A (because A was the rational choice) and both can make the same kind of predictions (assuming X is rational he will perform A). [^3] Both are theories that there are situations or relationships which agents can find themselves in, whose very nature makes a particular actor's behaviour meaningful, or more predictable.

Thinking in terms of high-level structures can occur when borrowing explanations from biological sciences. Evolutionary theory is already shot through with intuitions that are top-down, and often criticised for being based on tautology. Concepts such as fitness, are only defined in terms of a circle of high level entities. No genotype is fit merely by virtue of its internal structure. Rather it is fit with relation to the phenotype, and other genotypes and environment. Co-evolution or the evolutionary arms race, is a high level structure introduced to provide explanations of current properties.

Evolutionary roboticists, who attempt to evolve solutions to problems, find themselves thinking in terms of the evolutionary \"niche\"; and trying to produce behaviour by designing the environment and fitness function within which a particular control system will evolve. Hence there is some acknowledgement of a world which is prior to, and causally responsible for, the behaviour of the individual. This is particularly evident when one hears calls for a metric that will compare and classify environments according to the behaviour they engender. As they attempt to evolve new agents, they are relying heavily on the perceived causal power of an illata like co-evolution. Even Durkheim's ideas about criminality have found a parallel in a discussion of the virtues of parasitism.

Outside the Head

For this final section, I will change the subject slightly in order to emphasise another use that we might make of macroscopic entities.

Consider Terrence Horgan's [@superduper] assertion, that:

"Materialists who back away from type-type psychophysical identity claims, but who also seek to vindicate the causal/explanatory efficacy of mental properties, are already committed to some form of compatibilism on the issue of mental quasation. Since they are stuck with this compatibilist commitment anyway, they should take seriously the possibility that the right kind of compatibility will vindicate the causal/explanatory efficacy of mental properties that do not supervene on the properties that physically realize them, and perhaps will also vindicate the causal/explanatory efficacy of mental properties that do not even supervene on what's in the head."

This is a quote with some questionable anomalies. But consider only the emphasized portion of the text. If, as Terrence Horgan seems to, we want to talk about mental entities as supervenient upon more than what's in the head, then we are going to need to consider physical contexts that have a far wider scope than individual brain structures. The brain is already too complicated an organ to really think about in detail. So considering brain and a context which could, itself, include many other intentional agents, will require a high degree of abstraction. Those abstractions will be entities within some macroscopic theory and thus be open to the sort of questioning that more traditional macro-entities receive. Macroscopic structure just is the vocabulary needed to describe and model the context within which an agent operates.

Conclusion : Taking Society Seriously

This paper is only one step in a project dedicated to taking social, or macroscopic structure seriously when thinking about minds. What, though, is the purpose of this attempt to resuscitate social or top-down explanations of agency when we seem to be making such progress through scientific and reductionist approaches? Isn't this just obfuscating the issue, or worse, trying to sneak some dodgy political or humanities talk into cognitive science?

First, there is an issue of general macroscopic causation. Can there be a flow of explanatory responsibility back from the macroscopic entity to the individual? This relationship, sometimes known as quasation or quasi-causation is still controversial.

A nice toy example that is often used to make points about levels of description is Conway's Life Game. One can argue that the Life game may be viewed either on the level of individual cells or at a higher level of gliders and other abstractions. It is not that the Life game requires to be interpreted at either one or the other of these levels. Both are contemporaneously appropriate.[^4]

Consider a particular Life universe containing one glider which at a particular time t has passed through cell C. At t we can say that cell C has become alive. The question "did C become alive because it became involved with the glider?" cuts right across carefully separated levels of description. In this case a macroscopic explanation is intuitively acceptable. It really does seem that the passing of the glider caused the behaviour of C. But Maynard-Smith's reductionist, while able to say that the glider passed through C at t because C came alive then, ought not to phrase it the other way around.

The second issue is the particular problem of social illata: higher level structures to which rational, intelligent agents belong. Does the fact that we are already intentional beings prevent these higher level structures from influencing or causing our behaviour? Or, as claimed by holists, do we derive intentionality from those structures which we make up?

There are several good reasons for studying social illata. The first, as pointed out in the beginning of the paper, is that social science provides a kind of distorted mirror image of cognitive science. Looking into it we see some of our intuitions turned upside down, and some familiar arguments stretched and squeezed into unfamiliar shapes. This may inspire us to new intuitions about the problems of relating levels.

Secondly, it appears that, through considerable interest in evolutionary theory, researchers, particularly in ALife, have already begun to call on some macro-structure to do explanatory work. The language used by ethologists when studying animal's group behaviour begins to blend into that of the sociologist.

Finally, macroscopic entities, can be seen as tools for talking about context and relationships within that context. In this case, telling explanatory stories about mental illata may require one to include macro-illata.

• [^1]: How did this macroscopic mind manage to think about grandmother when all it is is a load of neurons firing? Well, obviously, one of the neurons is the neuron which thinks about grandmother.
• [^2]: Today, we might say autopoeitic.
• [^3]: One difference may be that game theorists will deny that they are making strong ontological claims for Prisoner's Dilemma games. Prisoner's Dilemmas do not exist. They are merely abstracta, not illata and have no causal power over the players who find themselves in the relationship. By contrast, the holistic social scientist is making this claim about Class.
• [^4]: Although occasionally there are times during the evolution of the Life game, when there are NO patterns for which abstraction to an alternative levels would be appropriate. Imagine a large Life universe has just been filled randomly (a 0.5 probability that any cell will be alive. For a while, it is uncertain whether any stable structures will appear. Although the beauty of Life is that some nearly always do.