Virtual robots have āevolvedā to cooperate ā but only with close relatives. The finding bolsters a long-standing ārule of thumbā about how cooperation has evolved, and could help resolve a bitter row among biologists.
The droids have real-life counterparts, called Alice robots, which live in the lab of of the Swiss Federal Institute of Technology in Lausanne. Each is roughly cube-shaped, 2Ā centimetres on each side. Equipped with two wheels, they motor around a small arena in search of scraps of āfoodā ā each a small chunk of plastic. When they find one, they are programmed to push it up to a white wall along one side of the arena ā if they succeed they are awarded points. The robots then have a choice determined by their software: keep the points they have scored or share them with other robots in the arena.
Floreano developed the experiment with of the University of Lausanne in Switzerland and , now at the Swiss Federal Institute of Technology in Zurich.
Advertisement
Robots in cyberspace
In order to see whether the robots would evolve to become more or less likely to share points over 500 generations, the team simulated them in a computer program.
They modified 200 versions of the simulation to control two factors: the relatedness of the robots and the cost and benefit associated with sharing points.
Relatedness is a relatively straightforward concept: in the simulations the team gave each robot a set of 33 āgenesā that could be altered to make robots distant or close relations. Cost and benefit is more difficult to define, however. Diving into a river to save your cousin, for instance, carries costs and benefits that vary with your ability to swim, the speed of the current, and the degree of relatedness to the family member, for example. So the team went for a simple but artificial construct: in some of the simulations a robot might receive 100 points from a neighbour that had only 10 points to offer ā a high benefit for a low cost ā in other simulations a robot received just 10 points when a neighbour chose to donate 10 points.
When it came time to produce a new āgenerationā of robots, it was descended from the most successful ā the highest scoring ā robots from the previous one.
The team found that, over several generations, a pattern emerged: robots became more likely to share points with another if the two robots were highly related and if the benefit associated with a cost was high. In detail, a robot would share its points only if the number of points received by the second robot, multiplied by a fraction indicating the relatedness of the two robots (with ā0ā indicating no genetic relationship and ā1ā indicating identical genetics), was greater than the number of points donated by the first robot. As a result robots with few or no genes in common were unlikely to share points, while those with many genes in common were more likely to share.
This set-up represents the first real confirmation of Hamiltonās rule, one of the most fundamental theories in modern biology. Put forward by W.Ā D. Hamilton in 1964, the rule defines when animals will, and will not, help others at a cost to themselves. Hamilton argued that the evolutionary benefits of helping another would outweigh the costs only if the animals were closely related. Specifically, he said the benefit to the other, multiplied by the relatedness of the animals, had to be greater than the cost to the helper.
Testing the rule
āIt had not been possible to test Hamiltonās rule in a quantitative way in real animals,ā Keller explains. āFor some people thatās been a major issue.ā Measuring relatedness, cost and benefit in real animals, all at once, has proved impossible. But this wasnāt a problem in the robot study, because the team controlled all three factors.
The lack of a true test of Hamiltonās rule was a key factor in a major attack on it last year by three scientists at Harvard University, led by . In a controversial paper in Nature (), they argued that Hamiltonās rule does not explain how cooperation evolved. The paper has attracted a , but Nowak is sticking to his guns.
Nowakās criticism has now been answered, argues Keller. āWe show that the rule works very well,ā he says. āBut Iām sure some people wonāt change their minds.ā
Indeed, Nowak remains unconvinced, saying that the simulationās design forces it to obey Hamiltonās rule. āIt is no surprise that Hamiltonās rule holds in a system that is designed to validate it,ā he says. āIt tells us nothing about whether Hamiltonās rule makes a correct prediction for actual biological systems.ā
Journal reference:
