快猫短视频

Rome

THE end, when it came, was swift. 鈥淭he walls and towers were instantly covered by a swarm of Turks. The Greeks, now driven from the vantage ground, were overwhelmed by increasing multitudes. Amidst these multitudes the Emperor was long seen fighting and finally lost.鈥� Such was the fate of Constantinople in 1453, and with it the entire Byzantine Empire, according to Edward Gibbon in his History of the Decline and Fall of the Roman Empire. Jump forward nearly 550 years and the same battle rages on a computer screen: a swarm of blue dots pitilessly eliminates the red dots. In cyber-worlds, computer scientists are making history repeat itself. Over and over again.

Great civilisations such as the Assyrian Empire, the Mayans, the Roman Empire and its eastern offshoot the Byzantine Empire were once mighty, ruling over great tracts of land and thousands of subjects. But mysteriously, after centuries of success, they collapsed in a couple of decades or less. Historians studying such downfalls have come up with dozens of explanations from archaeological and written evidence, but they still need proving.

So now computer scientists are creating artificial societies inside their machines in an attempt to re-run the course of history. These societies are made up of thousands of independent 鈥渁gents鈥� representing people or families placed on an electronic map of the world. By programming them to behave as they think the ancient peoples did, the researchers are finding that they can 鈥済row鈥� whole civilisations. And by changing the behaviour of the agents, they can change the ways in which the civilisations expand or collapse.

Teams in the US and Italy are trying to match these simulations to archaeological evidence from particular civilisations, to find out why they declined. At the Santa Fe Institute in New Mexico, researchers led by archaeologists George Gumerman and Tim Kohler are modelling the spread and collapse of an early society of native Americans, the Anasazi. In Rome, artificial life specialist Domenico Parisi and archaeologist Mario Liverani are charting the rise and fall of the Assyrian Empire. But it鈥檚 not just the past that is of interest-both groups think their models could be applied to our present cultures, and may even help predict and solve society鈥檚 problems in the future.

Model behaviour

In 1995, Chris Langton and his team at the Santa Fe Institute developed an artificial life program called Swarm. This model recreates the patterns created by flocks of birds, traffic jams or swarms of bees. Swarm has no central authority or organisation-it proves that lots of independent individuals each following a simple set of rules can form patterns and show complex collective behaviour.

Using the same concept, computer scientist Robert Axtell and social scientist Joshua Epstein of the Brookings Institution in Washington DC developed a program called Sugarscape that models the growth of societies. In Sugarscape, dots representing people or families move around a digital landscape in search of food-sugar. Whether they live or die depends on whether they find enough food to satisfy their 鈥渕etabolic鈥� needs.

The dots, or 鈥渁gents鈥�, are given a range of abilities-such as how far they can 鈥渟ee鈥� over their virtual landscape when searching for food-and are programmed to obey certain rules. In the most basic scenario, the agents look for the richest source of sugar, and go there to eat it. But they are in competition with each other and with groups of agents programmed with different rules and abilities. By modifying the rules governing how the agents interact, Axtell and Epstein can make them either fight or cooperate. They can allow the agents to accumulate wealth, excess sugar, and measure their 鈥渉ealth鈥� by how much sugar they eat. And by introducing mating, the researchers make the agents pass on their abilities-and the rules they obey-to their offspring.

With just a few rules and conditions, the agents in Sugarscape begin to mimic aspects of real life. For example, there is a maximum number of agents that can live in any one model, which depends on the amount of sugar available. This relates to the idea that the Earth has a 鈥渃arrying capacity鈥�-the density of population it can sustain. When the level of sugar fluctuates between areas-effectively creating 鈥渟easons鈥� -the agents migrate from area to area. Axtell and Epstein have also seen the equivalents of tribal formation, trade, and even hibernation. Similarly, extinction, or the end of a civilisation, might be an outcome of the agents following a particular rule.

One finding is particularly curious. The researchers added rules for inheritance to Sugarscape, allowing agents to pass on whatever sugar they still owned at death to their offspring. 鈥淎gents who would have struggled through life are buffered from selection pressures,鈥� says Axtell. In other words, agents with poor genetics-such as a shorter range of vision, or a higher metabolism-can survive very well on the strength of a wealthy parent. 鈥淚nheritance . . . reduces the rate at which vision-a measure of foraging ability-increases in the population over time,鈥� says Axtell.

But do the predictions about societies match the archaeological evidence? To find out, at the Santa Fe Institute Axtell, Epstein and Gumerman are using Sugarscape, and Kohler is using Swarm, to study the Anasazi, who thrived in the southwest US-the Colorado plateau, the Mesa Verde and the Rio Grande valley-from the first to the 14th centuries AD. 鈥淭he reason we use the Anasazi,鈥� say Gumerman, 鈥渋s that we have an environmental and demographic record that is without parallel in the prehistoric world.鈥�

This allows the researchers to give their artificial Anasazi the right characteristics, or rules, and to simulate the land on which they lived very accurately. 鈥淭he Anasazi societies were quite dependent on production of maize and beans in semi-arid areas,鈥� says Kohler, 鈥渟o we pay particular attention to data on agricultural production.鈥� Each piece of land yields a certain amount of food. The scientists program in this landscape and then release agents representing Anasazi households to see how they fare.

鈥淥nce the households find a place where they can raise enough maize to make a living, they stay there,鈥� says Kohler. He plans to introduce a social factor-trade-to explain why Anasazi villages are found close to each other during periods of high productivity and why they drift apart when productivity falls. 鈥淗ouseholds tended to supplement agriculture with hunting-gathering,鈥� says Kohler, so they needed more space to eke out a living when agriculture failed.

Using archaeological data, the team has worked out the size and type of Anasazi settlement for more than 7000 known sites in the region, and has estimated the population too. So far the demographic curves of real and simulated settlements match fairly well. This suggests that the rules governing the artificial Anasazi agents are well chosen and reflect the way the real people behaved. However, despite this good correlation, the researchers haven鈥檛 yet unravelled the mystery of why the Anasazi disappeared around 1350. 鈥淚t could have been a sudden climate change, but our research is in too early a stage,鈥� says Kohler. 鈥淲e need to test the model for other social rules first.鈥� The rise of clans, war, or the practice of inheriting land could have broken up the society.

Grander scale

Although it might seem that the models could tackle many different societies, Kohler is sticking to the Anasazi for now. He doesn鈥檛 believe that the same rules and conditions would account for a complex civilisation such as the Maya. 鈥淭he Maya were a very hierarchical society based on the central control of resources,鈥� he says. 鈥淭he Anasazi were relatively simple agriculturalists,鈥� says Gumerman.

However, in Rome, Parisi and Liverani are working on a grander scale. They are attempting to simulate the rise and fall of the entire Assyrian Empire. 鈥淭he empire originated in Mesopotamia, the land between the Tigris and Euphrates rivers,鈥� says Liverani, professor of ancient history at the University of Rome. Assyrian society began to expand in the 13th century BC, and at its peak in the 7th century BC, included all of what is now Iraq, Syria, Lebanon and Israel, most of Jordan, southeast Turkey, northwest Iran and for a short time, Egypt. Yet this great civilisation collapsed in just a few years.

To try to find out why, Liverani and Parisi, who is director of research at the Institute of Psychology in Rome, are using an agent-based model to simulate the expansion of the Assyrians. In the program, developed by Parisi鈥檚 colleague, Federico Cecconi, 鈥渃ells鈥� representing the Assyrians are placed at a starting point on a map, and as long as they have enough resources, they try to occupy adjacent squares.

The main factor controlling the growth of the artificial empire is what the researchers call the 鈥渆xpansive force鈥�-a measure of the ability to spread to adjacent cells. 鈥淭he further away from the capital of the empire, the greater the cost of carrying food, weapons, messages and orders,鈥� says Parisi. 鈥淭his has a negative effect on the expansive force.鈥� Every cell has an expansive force of between 0 and 1. 鈥淭he maximum force is at the origin,鈥� says Parisi. 鈥淔rom there it decreases slightly, say to 0.98, then 0.96, from the inner to outer cells.鈥� The rate of decrease determines the extent of the empire.

Limits to growth

Neighbouring states, which can be added to the simulation, can also slow the advance. 鈥淚f in our model the Egyptians, say, are occupying a square when the Assyrians get there, the expansion is checked,鈥� says Parisi. 鈥淲hat we call the `political penetrability鈥� of the square is limited.鈥� If two civilisations meet at an empty square, the one with the stronger expansive force gets the land. And just as in Sugarscape, when the two sides in Parisi鈥檚 model clash, they may go to war. The benefit of conquering an enemy town is the wealth of new resources. 鈥淭his boosts the expansion force locally,鈥� says Liverani.

Natural obstacles such as mountains, deserts, rivers and seas also stand in the way of a civilisation. In Parisi鈥檚 model, each square on the map has a 鈥済eographical penetrability鈥� depending on the terrain. Mountains and deserts are difficult to occupy, owing to a lack of resources. And if the civilisation doesn鈥檛 have seafaring skills in the model, then the oceans are an obstacle. 鈥淟ike other Oriental empires, the Assyrians never conquered the seas,鈥� says Liverani.

鈥淭he goal is to match the simulated expansion with the historical maps,鈥� says Parisi. Liverani hopes this will explain how and why the empire developed. 鈥淪ince the geographical spread is already known,鈥� he says, 鈥渨e want to ascertain the relative influence of the factors that caused this spread.鈥�

Although the researchers haven鈥檛 yet managed to model the rapid collapse between 612 BC and 609 BC, Liverani thinks the program could be used to test a theory he has about the downfall of the civilisation. 鈥淚t could have been caused by a sudden collapse in the expansive force,鈥� he says, brought about by a breakdown in Assyrian society. The economy was centralised, run by a palace bureaucracy, with governors running the provinces. Liverani thinks that the downfall might have been caused by the palace officials hoarding more and more of the food and resources for themselves. 鈥淲hen the dignitaries鈥� share grew over a certain limit, there was a crisis,鈥� he says.

Getting greedy

Such a situation could be generated by turning on 鈥済reedy鈥� behaviour in cells already rich with resources. 鈥淭he French philosophers believed that the seeds of decadence were internal, inside the state machinery,鈥� says Liverani. By introducing a factor for such decadence at the beginning of the empire, Liverani thinks he might see it gradually take over in certain cells, and eventually cause a collapse when it reaches a threshold value.

Parisi thinks that by switching on and off the abilities and rules of his artificial civilisations, his model could be used to simulate many different empires or societies. 鈥淥ne could insert different geographical, demographic and political data and see what happens,鈥� he says. 鈥淲e could try our model on the Romans simply by changing the parameters.鈥� In fact, by putting several different civilisations on the same map, Parisi thinks it might even be possible to simulate the course of history across the whole of Europe and the Middle East. And ultimately he reckons the model could reproduce the entire expansion of humans over the past 100 000 years.

Both the Italian and American teams think their simulations will eventually help us learn more about our recent history too, and perhaps look into the future, to predict problems and come up with solutions before they arise. Axtell thinks agent-based models could be used to 鈥渇irm-up鈥� the study of aspects of politics. 鈥淪ay you have a theory of how states form and interact,鈥� he says. 鈥淵ou can test whether the theory is any good by creating a model with states occupying different regions, formulating the theory in terms of rules for the individual states, and then running the system forward in time.鈥�

The models might also provide cultural and social insights. 鈥淲e are interested in how cultures change,鈥� says Gumerman, 鈥渢he role of the environment, new technologies and social structures.鈥� Axtell thinks that eventually, the models will be used for serious applied social or policy studies. Before embarking on any new trade policies, say, President Clinton鈥檚 successor could run agent-based models and see their likely effects.

鈥淚 think modelling may be relevant to the contemporary and future world for predicting global movements of people, goods and cultural patterns,鈥� says Parisi-in other words for predicting the results of the clashes of cultures that will characterise the next century. So when the modellers get their rules exactly right, both the past and future could be there for us to behold.