Can the future be known? II

Introduction

This is a potent question, this 'Can the future be known?'. It bears directly on the greenhouse issue, because it asks if we can really know the consequences of the increased abundance of greenhouse gases in the atmosphere. It is also a fair question, because we have much commonplace evidence before us of the uncertainty, the unknowability, the inscrutability of the future.

How can the boffins say that the increase in greenhouse gases now will lead to profound climate change in 2020, when they can't even predict tomorrow's weather?

How can they say that climate change will have dire effects on the whole world's ecosystems, when they failed to predict the spread of even a single species, the northern sea star, from Japan to Australia in ships' ballast water?

How can they say that climate change will have major effects on the world economy, when they couldn't even predict the Asian currency crisis?

How can they, how dare they, attempt to predict such grand futures, when they can't even predict the winner of the Melbourne Cup?

Today, we will ask them. Like delinquents before the school head, we will ask the boffins to 'please explain'. For a change, we will put them under the microscope. We will poke and probe them, and make them squirm and wiggle. We will make them, in a light hearted way, give us some answers that we understand.

And we won't be satisfied with the usual two handed scientist's explanation: on the one hand ... , on the other hand.

But before we begin our inquisition, we need some context. We need a richer understanding of just what scientists think they are doing when they say they are predicting things, when they say they know something. We need to go down into science's cellar where they keep all their old stuff and fossick around a bit. We need to dust off and examine some of their assumptions and precepts, the things they have assimilated so thoroughly in becoming scientists that they have long forgotten to explain to the rest of us. In short, we need to do a little philosophising.

Realms of possibility

We need to start with a notion of possibility[1]. We start here because the idea of knowing the future begs the question of whether the future is open. This is because, if the future is determined, there are no many possible futures, but only one actual future that we need to know. Knowing this single actual future is a very different scientific (and philosophical) project from knowing about many possible futures.

Without undertaking a long philosophical excursion, I will assert that our commonsense notion of a more or less open future is the right one. I say that our idea of the openness of the future is encapsulated in the verb 'can', a verb, as Daniel Dennett[2] says 'we can hardly do without'. It fairly captures the 'certain looseness' of the world that David Hume spoke about, the looseness 'that prevents the possible from shrinking tightly around the actual'.

As well as a notion of possibility, which I will return to soon, we also need a notion of the right level of explanation, the level at which our knowing best accounts for the phenomena, the level at which we may best communicate our understanding, the level at which the explanation best connects with all other scientific explanations. As we will see later, these levels are one and the same.

Now we don't merely have possibility, to pick up my thread again, we have realms of possibility. By this I mean we have different sorts of possibility. We need this idea because we need to distinguish, in some sensible way, the differences between differently possible futures.

'There seem to be at least four different kinds or grades of possibility: logical, physical, biological, and historical, nested in that order.' (Dennett). The most embracing of these is simple logical possibility: that something is, well, possible. That is, describable without contradiction. A logical possibility might be that the concentration of greenhouse gases in the atmosphere might start to decline. It is possible, since their concentrations are the results of emanations from sources and sequestering in sinks. Indeed we believe that their concentrations have fluctuated in the past. Thus we can construct a more or less logically plausible story about how this may come to pass. This will be a far more compelling story than the one which asserts that the concentrations will decline, but yet still remain the same. That is not even logically possible.

Another logically possible story is that the amount of the greenhouse gas, carbon dioxide, might increase until the mass of its carbon exceeded some mass which we know to be greater than the mass of carbon existing in the world. But this would be physically impossible, since it would violate the physical law of the conservation of matter. The mass of the world's carbon store is fixed[3], and physics provides no way for it to be increased.

Beyond that we might envisage a physically plausible story about the sequestration of CFCs from greenhouse gases through the mediation of plants, but this is a biological impossibility, since CFCs are novel organic compounds and living things have evolved no biochemical pathways to deal with them.

This brings us to our fourth level of possibility: historical possibility. We may create a scenario where greenhouse gases are once again mopped up through the photosynthesis of primitive procaryotes. I say once again, because this is how the high levels of carbon dioxide present in the primordial atmosphere were thought to have been reduced about 3.5 billion years ago, and such organisms are still found in the oceans. But that was then. Now the procaryotes share a more complex world ecosystem with the higher plants and animals, and their opportunity to repeat their historic feat would be thwarted by a host of ecological interventions. The time when that could have occurred is now past.

Science is a story

As I unfolded the different types of possibility, I referred to this story or that scenario. I now want to suggest that thinking about possibility in the way we have actually leads to the idea of thinking about science as a sort of story. Indeed it lets us connect science with the other (nonscientific) ways we have of dealing with possibility. If we are to argue that the scientific way of dealing with possibility, of knowing the future, is the best way, then we would like to be able to discuss these several ways within the same general framework. Thinking of science as a story allows us to do that.

Stories are generally plausible, or at least, better stories are generally more plausible than terrible stories. And plausibility has a lot to do with possibility. Good stories have a ring of truth to them, to the point where the novelist Garrison Keillor[4] says: 'A man who has no story is a man with no truth to offer.' And truth has much to do with where in the realms of possibility the story lives.

Creation myths, such as the Greek myths or the Book of Genesis, live mostly in the realm of logical possibility, being more or less plausible (in the sense that the logic more or less hangs together once a few heroic assumptions are accepted). The truths they contain are of a metaphysical or moral sort, and as predictors of the future, they leave a lot to be desired. I should not wish to die in battle to test whether or not Valkyries would transport me to Valhalla.

Science fiction and the broader genre of fantasy stories live mostly in the realm of physical possibility, save for a few excursions to the outer realms of logical possibility and sometimes beyond to cope with some of their more fantastic constructs. If they contain truths, they may be of a scientific kind in the sense that they elaborate the working out of some part of the world while suspending the operation of other parts.

The novel, whether Jane Austen or Peter Carey, lives well within the realm of biological possibility, but usually outside the historical realm, even when dignified with 'history' in the title. Novels rarely need to abandon this level of possibility, except when the plot gets tangled and a 'literary device' is needed to restore a sense of order. We could think of the novel as a future that could have been, as close to the idea of prediction.

But the novel fails to enter the realm of historical possibility. Novels present extremely plausible stories about the world, plausible in the sense that they are logically, physically and biologically possible, and also with an historical feel. But that feel is deceptive. And that deception is the key to understanding the place of the novel in our scheme, since the lie is intermingled seamlessly with profound truths.

A novel describes a fragment of a world that could have been. And, if a Jane Austen, it describes it with such power and beauty, that the truth and lie become one. But even if all novels were collected together into some monstrous supernovel - only a scientist could be so crass - it still would not tell a story about the world from which predictions could be made. This supernovel would be fragmented and disjointed, being composed of many logically, physically and biologically possible histories, but not containing the one actual history of the world.

History, on the other hand, does try to tell a story about the world which is embedded in the realm of historical possibility. And so does science. Indeed, science and history have much in common[5]: they live in the same innermost realm of historical possibility, they share many tools and techniques, and they are trying to tell the same one story. All the chapters are not written yet, and may never be, but the key to understanding this story is that, when completed, it will be seen to be a single connected narrative of what happened since the universe began.

Science goes further and tries to predict the future behaviour of the world and in doing so, tries to sketch out some future chapters, all the while insisting on the complete connectedness across all parts of the story everywhere. But in any one of these sketches, it may concentrate on some particular aspect or theme, not requesting the reader (as a novelist does) to suspend belief, but requesting the reader to see the phenomenon in this or that light, the better to understand it.

The idea of level of explanation

If science is one connected story, as history is, it is not told in the same way at all times to all listeners. Sometimes it is told at a very high level, a grand sweeping vision of the birth and death of the universe across eons of time and vast intergalactic distances, with no place for minutiae such as the development of life on a little known planet in a solar system near the edge of an unremarkable galaxy. Or else it is told at an abstract level of the abstruse mathematics of the subatomic world where matter, energy, space and time merge in a quantum soup. Each tell parts of the same story and each is connected to the other and to every other scientific story, so that we might ask which is the way that is best.

And best here means scientifically best: that which best assists our understanding, which best moves the game ahead, and which best communicates that understanding to others.

It turns out that we have some latitude here, since reasonable people (and scientists are reasonable to a woman) may disagree about just what is best. This is particularly so in the scientific domains we are dealing with, where we are concerned with the behaviour of living things and their interaction with their world - things in the middle of the scales of space and time in the universe. There can be sweeping visions here to be sure, but there is also a lot of nuts and bolts detail to handle as well. Our world is a world where details matter.

Thus we can have explanations, and predictions built on them, stories and scenarios if you will, that dive into the detail of this or that species or chemical or economic factor. Or we can have high level descriptions of global processes that no one can see or feel or touch, such as ocean heat load, biodiversity or M3. Each captures some parts of the story, and each is connected to the other as part of the one big story.

You pays your money and you takes your choice. If it were only that simple, but, alas, it is not. There is, in science, just as in the arts, a defining idea that some ways of doing it are, of themselves, better. In mathematics, it is called elegance, and mathematicians strive for it almost above all things. In science, and I suspect, the arts, it doesn't really have a name, but it is the same thing, the thing that lifts art and science to a higher level. This transcendence, in science, at least, is what distinguishes greatness from mere goodness, from workaday utility. It also speaks of universality, of applying to more parts of the story than the part of immediate concern, and so is peculiarly attractive in the work of prediction, which is, after all, about creating a new piece of the story.

We see such transcendence first in the work of Newton in the seventeenth century, who so unified and simplified physics that science has never really gotten over it. No scientist before or since has been so deified for his sweeping scientific vision. Alexander Pope exuded this couplet in his epitaph:

Nature, and Nature's laws lay hid in night,
God said, Let Newton be! and all was light.

Only Einstein in this century has come close, with plays and T-shirts in his honour. We might add that Stephen Hawking, Newton's successor to the Lucasian Chair of Mathematics in Cambridge, has not had a similar effect on science despite the enormous popularity of his book, A Brief History of Time, perhaps the first best seller to languish unread by most of its purchasers.

Stories are models

Armed with these ideas of realms of possibility, science as a story and levels of explanation, we may now briefly examine the scientific tools before us today. We want to classify them, to see their similarities and differences. We want to assure ourselves that the similarities are drawn from their common connectedness with the scientific enterprise, and that their differences, perhaps, reflect deeper relationships.

We can inform this part of our analysis by returning to Newton and introducing the idea of a model.

Through Newton's laws, science got to be good at predicting the future of some bits of the universe by using models. That is, science invented a fantastic trick: by creating a simpler caricature of the real system - one stripped of those bits extraneous to the matter at hand - in its own caricatured time and space, it was able to 'run the model' into the model future (or past) and predict what would happen in the real world. Notice that I have not mentioned the M-word yet. Models do not have to be mathematical, although Newton's were: they can be mechanical, electrical, geometrical, pictorial or even narrative as well. All they have to do to qualify as models is to have the character of a caricature - being stripped down to the essential elements - and having their own world. Science found some models to be so powerful they were called physical laws.

Those systems whose future science successfully predicted with such models have in common the fact that they are all rather simple. They are made up of relatively few distinct entities - planets are all much the same when it comes to the way they orbit the sun - and their interactions tend to be limited - Newton's inverse square law of gravitational attraction ensures that, for all practical purposes, a planet is only affected by the very near or the very large, all the rest may be ignored.

But these are by no means all the interesting systems in the universe. They may, in fact, be the least interesting when compared to complex systems, and in particular, to complex adaptive systems. Complex systems are characterised by, well, complexity. Where simple systems have few entities, complex systems have many; where simple systems have few interactions, complex systems have many. The weather, the oceans, and the earth beneath our feet are all examples of complex systems.

Complex adaptive systems, such as all living things and their parts - cells, say, or immune systems - and their assemblages - societies, economies, ecosystems and so on - take this complexity further in exhibiting the peculiar characteristic of learning, of evolving, of adapting. We speak more easily of the behaviour of such systems to denote this flexibility, where we speak naturally of the dynamics of simple systems to denote their relative fixedness.

Since we are interested here in those systems where man is coercive or at least involved, we are naturally interested in complex adaptive systems.

For us then, one powerful way of classifying our tools has to do with whether they attack the world from an assumption of its underlying simplicity or from its contingent complexity. More simply, where do these tools think complexity lives?

Tools for an uncertain world

Qualitative tools like scenario planning are very like scientific novels. They stress plausibility at the innermost historical realm of possibility. They are histories of the future, stories with as much complexity and contingency as necessary for the issue at hand. They disturb many scientists for the strong reason that while they are connected back to the present, they are not connected (or better, in principle not connectable) across, as it were, to the rest of the future. For that reason some scientists are skeptical that they are really scientific. Others worry that, as stories, they fail some test of scientific rigour which all scientific tools must pass. I think a more generous view is called for, one that acknowledges the muddiness of human efforts and which accepts any tool which helps us think clearly and logically about possible futures.

Simple physical models are direct descendants of Newton's approach. Here they are represented by quantitative econometric models. We could have offered you a feast of celestial mechanics, but that would have had little bearing on the issues of climate change and its consequences. And if you are in any doubt that econometric models are true inheritors of Newton's mantle, listen to the words of Leon Walras, the 19th century originator of the central dogma of economics, the 'general equilibrium theory':

We all accept the current description of the universe of astronomical phenomena based on the principle of universal gravitation. Why should the description of the universe of economic phenomena based on the principle of free competition not be accepted in the same way?[6]

or the words of his successor, Vilfredo Pareto, which also confirm its essentially Newtonian cast:

Let us go back to the equations which determine equilibrium. In seeing them somebody - it might be the writer made an observation, 'These equations do not seem new to me, I know them well, they are old friends. They are the equations of rational mechanics.' That is why pure economics is a sort of mechanics or akin to mechanics[7].

Such models search for simplicity through the discovery of transcendent principles or laws. They seek to strip away the initial complexity, perhaps stigmatising it as noise. They offer powerful, but idealised, insights into the dynamics of systems. Their predictive power emanates from the equations that capture that ideal, and so their predictive utility often lies at a high level of generality.

Whether composed of a few equations or many thousands, these models need to be taken seriously. They put men on the moon, predict our weather, drive our industrial processes, and inform our Treasury.

Complex physical models, here represented by process-based physical models, try to repeat Newton's trick but across two or more domains. In our examples, the domains are the weather, the soils, and the vegetation. They retain the essentials of a Newtonian mechanics, with all its strengths and weaknesses, within any one domain. But then they try the new trick of joining these domains together, so that the outputs of one domain become the inputs for another. Complexity may be stripped away within a domain, but it reveals itself in the interactions between domains. While the interactions are simple, with any domain either receiving or providing input to another, the models retain their Newtonian character and utility. But when domains interact transitively, each providing and receiving input from the other, then these models start to exhibit strange behaviours and we need to use them differently. We need to put aside our idealised Newtonian views of prediction and replace them with something more subtle.

Design framework modelling takes this evolution a step further. Here different models live in different domains, and the domains are allowed to interact, but no restriction is placed on the types of models within a domain. In our examples, the domains are the groundwater, the soils, the weather, the economy, the social setting, the vegetation, the livestock and so on. Newtonian models live in some domains, scenarios live in others, and guesstimates and descriptions live in yet others. Here there is no assumption of an underlying simplicity, and complexity is a contingent fact of the world, both within a domain and in the interactions between domains. Such models have been found particularly useful in the exploration of scenarios, the analysis of 'what if ...?' questions that lie at the heart of good policy development.

Our last example, agent-based modelling, is also our most intriguing. These models are state-of-the-art attempts, not to capture complexity, but to let complexity emerge. They exemplify a belief that the complex behaviours of living systems emerge from the interplay of simple behavioural rules applied at the level of the individual agents making up the system. In this they are at the other end of the spectrum from Newtonian models which seek universal simplicity by stripping away the particular complexities of any particular situation. Not surprisingly, these models find their best work in the analysis of systems built by living things, such as ecosystems, economies, social systems and so on.

Conclusion

At issue here is something more than 'You pays your money and you takes your choice'. While it is true that models are caricatures and not the world itself, it is also important to recognise that they are also tools. We can use them to understand an uncertain world. But our modelling choices need to be made with care.

The ecologist, Richard Levins[8], argued that 'no model can simultaneously optimize generality, realism, and precision'. He goes on to suggest that two out of three ain't bad, since mostly we are doing well to get but one of this triptych. He then concludes that:

Models differ in the aspect of reality preserved, in the departures from reality, and in their manipulative possibilities. They can therefore give different results. Since it is not always clear which consequences derive from the properties of nature, it is often necessary to treat the same problem with different models. A theorem is then called robust if it is a consequence of different models, and fragile if it depends on the details of the model itself. The search for robustness leads to the proposition that truth is the intersection of independent lies.

References