Simply Complexity (16 page)

Setting aside “Complex System” cars, it turns out that there are many potential applications for such online systems management. Next generation aircraft will have so many interacting parts that the pilot will have no hope of assimilating all the available information. In short, the plane will be “out of control” since no single human or computer will be able to respond quick enough if things go wrong. So how should this system be controlled? One approach is to use a collection of competing miniflaps located along the wing, and then manage them in real-time according to David’s scheme. Indeed this is exactly the approach being pursued by Ilan Kroo at Stanford University. In human biological systems, doctors increasingly find themselves having to manage complicated conditions to do with overall levels of physiological activity in the body. Worse still, the precise nature and level of such activity may be unknown to the doctor at any one time. For example, the precise level of immunity in the immune system
cannot easily be measured, nor can the level of heart activity or mental behavior. This in turn provides a possible connection to so-called dynamic diseases such as epilepsy and their online management. Likewise, a seizure involves a sudden change in the activity of millions of neurons. Feedback control of seizures would require an implantable device that could predict seizure occurrence and then deliver a stimulus to abort it. This sounds a very intrusive procedure. But David’s work raises hopes that it might be possible to develop a “brain defibrillator” which delivers brief but effective electrical stimuli over a small part of the brain, rather than intrusive control over each and every one of the constituent “agents” (i.e. neurons).

Another possible biological application is the treatment of cancer tumors. Inside each tumor there is an ongoing competition among cancerous and normal cells for two limited resources: nutrients in the blood supply and space to grow. A fully invasive procedure to remove a tumor might be so disruptive that it would actually promote mutation – for example, cutting out only ninety percent of a tumor might actually be even worse than leaving it untouched. By understanding how the overall population behaves one could apply David Smith’s “population engineering” to a small group of the cells. Without being one hundred percent accurate – which is anyway impossible to do – David’s work suggests that one might then conceivably be able to steer the tumor toward safer territory. Even in the immune system, where the body self-regulates itself as a result of the interaction of hundreds of different biological objects, one might be able to tweak one part of the system so that it modifies the overall behavior. For example, it might be possible to control autoimmune diseases such as arthritis, where the body attacks itself, by adding a vaccination against something entirely different. Given that all these entities or objects are interconnected, then by biasing the immune system population the overall system could be steered off in a given desired direction. Such minor and indirect intervention could prove to be enough if it is carried out at the right place and the right time.

There are interesting applications in the area of finance. First, an institution such as the Bank of England or the U.S. Federal
Reserve – who are not out to predict financial market movements
per se
– could, if necessary, step in to apply a small amount of influence to a small section of the trading population. They could thereby steer the system – in particular, the market index, or exchange rate – out of trouble. This is of course known to be possible if the intrusion is allowed to be massive – after all, if one makes trading illegal then the market will stop moving since there are now no trades being made. However, the implication of David’s work is that this can be achieved without over-intrusive intervention – and hence at relatively low cost. Second, suppose that there is a fund manager holding lots of stock. These stock effectively compete with each other for their share of the fund’s portfolio. If the fund manager sees that the value of the overall portfolio is momentarily heading downward, she could of course sell all the different stock and buy new ones. However, this is very expensive (i.e. very intrusive) because of transaction costs. David’s work shows that she might be able to tweak this portfolio in real-time, by buying/selling a very small amount of stock in order to steer it clear of danger.

Next-generation “smart” technologies can also benefit. Consider a population of autonomous agents competing for a limited resource, as discussed in earlier chapters – examples of which include a cluster of extra-terrestrial craft, a cluster of antiexplosives robots checking out a building, or even a cluster of nanobots within the human body. Reprogramming these agents might be impractical or impossible and, as such, some other form of control would be necessary. David’s work shows that this control could come in the form of a “vaccination” by injecting additional agents into the system. If the composition of this vaccination is chosen wisely, the competitive nature of the agents is such that their subsequent interactions with the rest of the population generate an overall steering effect via feedback. This approach could also benefit systems where the sheer number of participants renders conventional control impractical. Thinking further into future technology such control philosophies might even be appropriate for whole surfaces of materials which are coated with interacting active agents – so-called “smart surfaces” – or even new designs of “smart matter”.

It may even help out with understanding and eventually controlling global warming. Typically there is no feedback between the actions of humans and the weather. The fact that many people will go and lie in the sun does not affect the chances of it being sunny. However, the actions of our society as a whole do indeed seem to be changing the weather over the longer-term. The reason is that human consumption produces waste products such as greenhouse gases which rise into the atmosphere and might eventually affect global weather conditions. The weather itself results from a complicated interaction between air and water – or more specifically, the changing temperatures of the oceans, air and land masses – so it is likely that this will be altered over the longer-term by our collective actions. Global warming. David’s work suggests that we might “undo” these effects, or at least moderate them, without full control of the climatic system or knowledge of what its individual components are doing. For example, an upcoming flood, hurricane or drought might be weakened by some form of atmospheric intervention – possibly by injecting a harmless gas of particles into the air to either promote some pre-emptive rainfall or alter current cloud formations.

But for the final word on this let’s turn to Hollywood. In particular, you might have already picked up on a similarity between this research and the movie
Minority Report
. The movie’s storyline features a collection of “precogs” – which is shorthand for precognitives – who make fairly blurry predictions about the future. While the movie only has three precogs, it turns out that David Smith’s mathematical model for constructing future corridors is analogous to a population of many such precogs. When this population of precogs don’t agree – which in the movie was the origin of the secretive “Minority Report” – the corridors are very wide and have no definite direction. Hence the future becomes unpredictable. By contrast, when the population of precogs do agree – which in the movie was officially always the case – the corridors are narrow and have a definite direction. Hence the future becomes predictable.

We have just seen a collection of competitive, decision-making objects, such as people, self-organize themselves into crowds without the need for any “invisible hand” or central controller – as if by magic. What made this even more amazing was that it was entirely unintentional. Nobody should want to be part of a crowd if they are competing with each other. Yet crowds emerged – and they did so because everybody was being fed the same global information and they were all competing for the same limited resource, such as space on a particular road or a favorable price in a particular financial market.

This makes sense for Complex Systems like financial markets and traffic where people don’t generally know each other and also don’t know how to contact each other. But people tend to be social animals – and in many examples of human Complex Systems, individuals may well try to make private contacts and form alliances or coalitions of some kind. In other words, they begin to interact directly with some other sub-group. In short, they form a
network
. And the same holds for the animal and insect kingdoms.

The added feature of local contact and communication, and hence interaction between individuals, has led to the topic of networks becoming important in the study of Complexity. A network tells us who is connected to who, and therefore who is interacting
with who and what their interactions are. Such a network may also play a role in feeding back information from one part of the population to another – for example, cell phone calls can instantly connect people who are geographically very far apart. In
chapter 4
we talked about feedback which comes in the form of common, or public, information – typically from the past. Here we are opening up the possibility that feedback may also come from different points in space – moreover, that different people may have different types and/or different amounts of feedback depending on who they are connected to in their social network.

Scientists have so far focused on static networks. In other words, they have focused on looking at the entire set of connections that have appeared over some particular span of time instead of focusing on when these connections actually appeared and disappeared. But there is a big problem with this approach. Lumping together connections means that you lose any information about the order in which connections appear and disappear. And yet, this information about
when
things happen is very important.

Just imagine a collection of people among whom a rumor is being passed – or worse, some kind of virus is spreading. Let’s focus on three people, and let’s suppose that there is a particularly nasty virus going around. Imagine that persons A and B are not yet infected, but person C is. If you are person A, it really matters whether person B spends the evening with person C before or after spending the evening with you. If it is after, then you are safe – but if it is before, watch out.

It is for this reason – especially for the spread of something such as information, or a rumor, or a virus – that the network that you sit in is so important. Above all, it really matters how and when you are connected to who. It is also for this reason that the network research field has become incredibly active recently – mainly with the focus on large networks involving many objects. Particular pioneers in this field include Mark Newman of the University of Michigan, Duncan Watts of Columbia University, Steve Strogatz of Cornell University, and Albert-Laszlo Barabasi of the University of Notre Dame. These peoples’ research papers are available at
http://xxx.lanl.gov
. Here I will only refer to specific results which are necessary for our Complexity story.

A network consists of a set of nodes, such as the three people described earlier. Depending on the network being studied, some or all of these nodes may be connected together by links. A network therefore gives us a visual picture of how a collection of objects are either connected or interact. Based on this reasoning, it follows that many of the things around us in our everyday lives represent examples of networks, from transportation networks and information networks through to social networks – and even voting networks. (See for example, the downloadable research paper listed in the Appendix which uses network analysis to uncover voting biases and cliques in the contest that Europeans love to hate – the Eurovision Song Contest).

Because of the wide range of possible internal interactions and behaviors that it can exhibit, a Complex System may produce a wide range of network shapes or “structures”. In particular types of Complex System, the ways in which the network connections are arranged may seem to follow a particular pattern, for example, many social, transport and information networks have particular objects (i.e. nodes) with many more connections than others – in fact, an abnormally large number. These objects act as hubs to which many other objects are then connected. For a concrete example, just think of an airline such as
Continental Airlines
which has hubs in Newark and Houston. You may even be stuck in one as we speak, though I very much hope not. In addition, we probably all know someone who has so many friends that she needs a PDA to keep all their contact details – and we also know of others that have so few friends that they can remember their contact numbers by heart.

An important scientific question concerns the extent to which Nature does or doesn’t favour centralized network structures. We can think of this in our everyday lives in terms of road planners choosing to either place a ring road around a city, or to add additional roads through the center. It turns out that biological systems such as fungi have to solve such a problem all the time. A fungus is essentially one big living network with no central brain or stomach. As such, it has to supply food to all parts of its
network all the time – very much like a major supermarket chain has to continually resupply its stores with food products. It is therefore interesting to see exactly how biology, and in particular the fungus, manages to deal with such supply-chain problems.

The network structure phenomenon which has attracted most attention is arguably the so-called “small world” effect. It is mimicked in the earlier example I gave, using three people A, B and C. Even if person A does not know person C directly, the fact that A knows B and B knows C means that A and C are indirectly connected. Suppose that A and C happen to meet for the first time and in casual conversation they find out that they both know person B. They are then both likely to conclude that “It is a small world”. Now imagine many people – in fact, think of the entire population of the world. We all tend to live a rather clustered existence: in different towns which are within different states or departments, which are in turn within different countries, and on different continents. Yet the shortest path between any two of us is still remarkably small on average. This was most famously demonstrated in 1967 by American psychologist Stanley Milgram. Milgram sent a number of letters to people in Nebraska and Kansas, asking them to forward the letters to a particular stockbroker in Boston. However, he didn’t reveal the stockbroker’s address. Instead, he asked the recipients to forward the letter to people who they knew, and who they suspected might be “closer” to the stockbroker as a result of their profession, location, or social circles. Milgram then let them get on with it, with the result that many of the letters arrived to the correct destination after very few forwardings. Six, on average. This is the origin of the term “six degrees of separation” and suggests that although objects might belong to quite different clusters (e.g. an office worker in Kansas, as opposed to a stockbroker in Boston) the average path length between them is typically very short. So although there are many people in the world, and the world is organized into clusters or communities, it is a small world in terms of who knows who.