Simulation upholds a theory of natural selection
If natural selection means that only the fittest individuals survive to pass their genes on to the next generation, then selfless behavior should not exist. Yet dolphins are known to support their injured brethren, and some species of monkeys will scream to warn others of danger, even though doing so makes them an easier target.
Biologists have a theory to explain such altruistic behavior: Animals will help one another if they have strong genetic ties, since doing so preserves genes they have in common. Known as Hamilton’s rule of kin selection, the theory predicts that animals are prepared to make bigger sacrifices for close relatives — who share more genes — than for distant cousins.
But testing this theory in the real world would require careful observation of whole populations of animals over hundreds of generations, which would take more time than most scientists have. So Swiss researchers built a population of 1,600 virtual robots with hundreds of thousands of offspring that evolved over 500 generations.
The result: Groups of robots that were closely related acted more altruistically than groups that were made up of strangers. The stronger the genetic ties within the group, the more sacrifices the robots made.
Hamilton’s rule “really seems to be a fundamental principle of a natural system,” said Markus Waibel, a roboticist at the Federal Polytechnic School of Lausanne in Switzerland, who worked on the study published Tuesday by the journal PLoS Biology.
Waibel and his colleagues programmed the virtual robots with 33 “genes” that controlled their behavior. The robots were put into groups of eight and set loose to forage for food tokens. After a robot found and retrieved a token, it could decide whether to keep it or share the spoils with the other seven members of its group.
In some of the groups, all eight robots were exact clones that shared all 33 genes. In other groups, the robots shared about half of their genes, as if they were siblings. Some groups were made up of robots with no genetic relationship at all.
After each round of token-gathering, the researchers took the genetic code from the “fittest,” or most successful, robots and recombined it with that of other successful robots to mimic sexual reproduction. Random mutations were sprinkled in to enhance realism.
The researchers found that Hamilton’s rule did hold true. The robots’ inclination toward altruism depended on the costs and benefits of selfless behavior. When the cost and benefits were roughly equivalent, the robots were about twice as likely to help their identical twins as they were to help their other siblings, and not at all likely to help complete genetic strangers.
But when the cost-to-benefit ratio was high — in other words, when it was too much of a pain to be nice — the robots would only help one another if they were identical twins. On the other hand, if it didn’t hurt much to share some of their bounty, the robots would help one another even if they were distant relatives.
Waibel said he was surprised that the robots followed Hamilton’s rule so closely, since the theory doesn’t take into account certain complicating factors — such as the idea that one gene can affect many traits, as was the case with the robots’ virtual genome.
Troy Day, a mathematical biologist at Canada’s Queen’s University in Kingston, Ontario, who wasn’t involved in the study, said the simulated robots provided convincing data to support a genetic basis for altruistic behavior.
“It’s a creative way to explore these things experimentally,” he said.
