25.09.2018 | Cooperation: Friends are allowed to forget

Focused social contacts promote cooperative behavior

In which forms of social networks does cooperation emerge? In a game theory simulation study, researchers at the Max Planck Institute for Human Development and the University of Nebraska-Lincoln found that cooperation is more likely to develop in social networks in which people focus their interactions on just a few others. The findings of the study have been published in the journal Proceedings of the Royal Society of London B.

We help friends move into a new home, borrow a colleague’s tent to go camping, and water our neighbors’ plants when they go on holiday. We are all part of complex social networks within which we interact and, ideally, cooperate with one another. Although the average network consists of 150 to 200 people, we do not interact with all of those people to the same extent. Instead, we tend to focus on a few, close contacts. Researchers at the Max Planck Institute for Human Development (MPIB) have used computer simulations to investigate this typically human network structure. Their results show that skewed contact patterns focusing on just a few people have a positive effect on cooperative behavior in networks. Such network structures also help to counteract the negative effects of forgetting – that is, not being able to remember how another network member acted in the past – on cooperative behavior.

“For psychologists, it’s fascinating to observe how cooperative behavior develops in social networks,” says Thorsten Pachur, researcher at the MPIB and coauthor of the study. “Interestingly, the structures within the social network seem to play a particularly important role here. To investigate the mechanisms involved in more detail, we used the prisoner’s dilemma game to simulate and compare social networks with connections of varying intensity.”

The prisoner’s dilemma is a simple decision-making game that behavioral psychologists, game theorists, and economists have been using for decades to investigate cooperative behavioral strategies. Two players, the prisoners, are accused of a crime and interrogated separately. They can either confess or claim innocence. If both prisoners deny committing the crime, both will receive a short sentence. If both confess, they will both receive a longer sentence. However, if one prisoner confesses and the other claims innocence, the confessor will go unpunished as a crown witness, but the other player will receive the maximum sentence. Things get very interesting when players play several rounds against each other and are able to react to their opponent’s decision in the previous round. Various strategies are possible. One of the simplest and most successful is the tit-for-tat strategy: cooperating in the first round and then imitating the opponent’s last move in all following rounds.

Because the players’ decisions can be easily translated into mathematical algorithms, behavior in the prisoner’s dilemma is often investigated using computer simulations in which virtual agents interact with each other. The researchers used a simulation of this kind to examine the influence of contact structures and of different strategies on the development of cooperation within social networks.

A virtual population of 90 agents each participated in 100 interactions, which they distributed among 10 neighboring agents, the members of their social network. Some agents played 10 rounds each with all 10 network members; others were programmed to imitate the focused contact structures of humans, playing 33 rounds with some network members and just 1 or 3 rounds with others. The simulation showed that the more unevenly the agents distributed their interactions – that is, the more focused they were on a smaller group of network members – the more established cooperative behavior became over generations of agents.

In addition, the researchers programmed a degree of “memory loss” into the agents. Sometimes it is difficult to remember whether a friend paid for the drinks last time, and whether it is now our turn to pay. In this context, we might remember that she helped us move house last year and decide accordingly. The agents also sometimes “forgot” how their fellow players had behaved in the previous round and had to decide on the basis of earlier rounds. The results show that although forgetting reduces the spread of cooperation, this effect was mitigated in social networks with more focused contact structures. “Although people can’t remember previous interactions perfectly due to their natural cognitive limitations, close social structures promote cooperative behavior in the longer term,” says Thorsten Pachur.


Public Relations Department
Kerstin Skork
Phone: 030-82406-211
skork [at] mpib-berlin [dot] mpg [dot] de

Artur Krutsch
Phone: 030-82406-284
krutsch [at] mpib-berlin [dot] mpg [dot] de

Further information: 

Stevens, J. R., Woike, J. K., Schooler, L. J., Lindner, S., & Pachur, T. (2018). Social contact patterns can buffer costs of forgetting in the evolution of cooperation. Proceedings of the Royal Society B: Biological Sciences, 285(1880). doi:10.1098/rspb.2018.0407