Simulating Social Norms In Societies On Supercomputer

Whether social norms survive the passing of time depends on the kind of society they are practiced in, a RIKEN researcher and a collaborator have found¹.

Many times, we might help someone in the expectation that they will return the favor and help us is some way. This kind of 'you scratch my back and I'll scratch yours' interaction involves direct reciprocation.

But other times, we might help others even when there is no obvious immediate benefit to us. While appearing more altruistic, these kinds of interactions are often based on subtler motivations. By helping you, I can gain a good reputation in the eyes of others, which will make them more favorably disposed toward me when I need help in the future.

These indirect reciprocations are governed by social norms that people use to assess actions by others. Examples of social norms might include relinquishing your seat on a train for an elderly or disabled person or tipping in a restaurant. These social norms often influence our behavior more than we realize.

Scientists are interested in modeling how social norms change over time in societies. Some social norms prove not to be beneficial to individuals over the long term and so they tend to disappear.

But the sheer number of social norms and the complexity of interactions make this extremely challenging, and consequently researchers often resort to highly simplified models.

"Indirect reciprocations are much more complex than direct ones because many more people are involved in them," explains Yohsuke Murase of the RIKEN Center for Computational Science. "This makes them much more challenging to study."

To gain a more comprehensive picture, Murase and Christian Hilbe of the Max Planck Institute for Evolutionary Biology in Germany have run large-scale simulations on Fugaku, one of the world's most powerful supercomputers. Their simulations involved more than 2,000 social norms and the interactions between them, resulting in a massive matrix made up of about 4 million entries.

"We wanted to overcome the limitations of simple models, and so we conducted a global analysis that includes a more comprehensive set of norms," says Murase. "The question we wanted to answer is does cooperation evolve or not?"

The pair found that the answer depends on the type of society-cooperation rarely evolves in well-mixed populations, but it can in populations consisting of well-defined subgroups. The findings reveal that cooperation is promoted through the interaction of social norms, reputations and population structure.

Both simple and complex simulations have their place in investigating social norms, the researchers note. Simple simulations are helpful for forming intuitions about social norms, while large simulations can test those intuitions.