October 20, 2021

I, Science

The science magazine of Imperial College

Evolution can be merciless. It selects for success and banishes failure. It is therefore natural to assume that it favours genes that help themselves and not others. Yet for some reason, cooperative behaviour has managed to evolve, with no solid explanation as to why. But thanks to computer technology, we may now be closer to the answer.

Discussions of cooperative behaviour often involve the game, ‘The Prisoner’s Dilemma’, which describes a scenario where two people arrested by the police are each offered a reduced sentence in exchange for a testimony against the other person. The game considers two types of players: cooperators who pay a cost to help others, and defectors who avoid paying the cost, benefiting from the cooperators. Everyone is better off if both cooperate, but defection is always more beneficial from the individual’s point of view.

A single play of this game may not reveal much apart from the difficult choice individuals face in this scenario. However, the power offered by computer simulations in running the game repeatedly for a hypothetical population of individuals, has already yielded insights into the factors that may have contributed to the evolution of cooperation.

Now researchers from the Max Planck Institute of Evolutionary Biology in Plön, Germany have for the first time run computer simulations that consider two of these factors simultaneously. One factor involves repeated interaction, when individuals base their decisions to either cooperate or defect on past experiences. The second factor, population structure, involves cooperators being physically close to other cooperators, and therefore more likely to cooperate. Taking a hypothetical population of 200 individuals, the researchers ran 10,000 versions of the simulation, for 500,000 generations, to conclude that a bit of population structure, and a lot of repeated interaction could have been a possible recipe for human cooperation.

“To get a good approximation of what’s going on, the simulation needs to run for a long period,” says Dr. Julian Garcia of the Max Planck Institute. “Each run can take between one day and three weeks depending on the parameters. This method was hinted at in the early 1980s, but it would have been impossible to have such computer simulations even ten years ago.”

But these simulations also have their limitations. Anthropologists and psychologists have traditionally focused on experiments and fieldwork involving small-scale human societies and primates to explain cooperative behaviour. These include behavioural observations focusing on naturally-occurring cooperation, such as food sharing or social support, while also considering kinship and other social relations.

“Computer simulations have been very useful in generating hypotheses,” explains Dr. Jeff Stevens of the University of Nebraska-Lincoln. “But one strong weakness is that they neglect important aspects of the organism’s psychology and cognition.”

Weaknesses in computer simulations include their assumption of near-perfect memory during repeated games of the prisoner’s dilemma, when human experiments have shown limitations in remembering whether a partner cooperated or defected the last time they interacted.

“A model is only as useful as what you put into it,” says Garcia. “I could include many more parameters, but it is only when we look at a limited setting that it leads to some understanding. But I think here we are aiming at striking a right balance.”

The ability to tweak models and run virtual evolution over thousands of generations at the touch of a button is clearly a tool not to be ignored. But without physical evidence to prove the factors responsible for the evolution of cooperation, simulations alone may not be enough to provide the necessary answers. It is perhaps only through a combination of modern and traditional research methods that we will ever be able to solve the conundrum of cooperation.

Image: flickr | chuckp