Out of the woulds

When people are in charge of systems they sometimes make mistakes. Generals lose battles. Directors lose companies. Betamax, the arguably better video standard, lost the battle to VHS. People lose elections. Someone is always held responsible. This view is wrong, and is perhaps the most important message from John Casti's Would-be Worlds.

It is not that people cannot be held responsible, but that there is not always a "rational" thing to do. Do we go to a party by thinking carefully about whether we will meet the people we want to? If the other party-goers also think about whether we will be going, and whether they want to meet us, then we ought to take their thinking into consideration. Likewise, they will be taking our plans into consideration. This is the sort of problem that when everyone thinks seriously about it, anything could happen. It is a typical complex systems problem.

Casti has written a readable overview of how to understand complex systems, and how computers may be used to simulate what can happen. As he shows, there are lots of complex systems - financial markets, traffic flow, football games, climate systems, even parties - that lend themselves to simulation.

Would-be worlds are simulated worlds inside a computer that would be real worlds if it was not for being simulated. Their central importance is that we live in many complex systems, and we have no other way of stepping back and seeing what is going on other than by using computer simulations. Many complex systems are extremely important to us and for our survival; we desperately need to be able to make sensible decisions. The world is not a party.

With would-be worlds, we can experiment with alternatives in a way that is not possible with reality. Thus a town planner can simulate whether a proposed bridge would improve traffic flow. There is no need to build a real bridge at great expense, and possibly make a mistake. A would-be world can be re-run. It has alternatives.

We know where Venus will be next week: the laws of physics seem pretty reliable for predicting what is going to happen. But we only know where Venus will be because an expert astronomer worked it out carefully. Most of us would not know how to do it, because most of us do not really understand the laws and how to use them properly. Likewise, when would-be worlds are used, it is important to understand how to interpret them properly. Not everyone is likely to understand and apply their results correctly.

Do computer simulations give only rough hints about what can happen in the real world, or do they give us real insights? This is no idle question. If we are to rely on would-be worlds to help make important decisions, then we must think hard about their reliability. We have conventional insights about what makes for good science, and Casti discusses how we might apply these insights as analogies to help decide what makes for good would-be worlds. Should we believe a climate simulation and take its recommended, but drastic, steps to avoid global warming?

There are many things in the world that are still too complex even to simulate. Casti describes biochemical problems such as protein folding: amino acids seem to solve complex problems rather too fast. These little molecules know something we do not. Unfortunately, there are many everyday problems that seem equally ill-fated to lie beyond our grasp. For the time being, would-be worlds cannot help with all problems. Recognising their limits is going to be an important area of work.

Would-be Worlds is a great syllabus of new ideas, and anyone wanting a taster should read it. It is a revision course for any would-be computer-literate thinker. But as a primary source of ideas that the reader can get to grips with, the book is ultimately disappointing. Rather too often, Casti says he does not have space to explain things. Yes, he is covering a wide spread of topics, but the writing is selfconsciously fast-paced. He leaves his readers with facts and names, not new skills they can use. Perhaps Casti is really trying to tell us something more important than the details, something like the fundamental nature of reality and our place within it. One chapter starts to rise to the challenge, and as I read it I felt: maybe he is starting to get there. Then I turned to the next chapter ... but the book had finished. The thick part of the book I had been hoping would carry on was the bibliography.

Perhaps we can look forward to a further edition that also tells us how we can run with the ideas ourselves, in worlds running on our home computers to get to understand them, and have a last chapter that completes the story. That would be my would-be book, but it is clear there is a lot of research still to be done.

Harold Thimbleby is director of research, faculty of technology, Middlesex University, London.