Cooperation and Reputation in Dynamic Networks

Project summary

Cooperation is crucial to the success of any human organisation. But cooperation is rarely free – while cooperative agents contribute to the collective welfare, they bear the risk of exploitation by defectors at a personal cost. The pressure of natural selection further complicates the matter, as high fitness is favoured among self-interested agents. A challenge thus arises: how can a high level of cooperation be sustained despite possible exploitations?

One possible solution points to the spatial structure of the population, introduced by Nowak and May.[1] Instead of the well-mixed case, spatial structures make agents interact with their nearest neighbours, which could lead to clustering of strategy types, increasing the payoffs for cooperative agents relative to those of defectors. Subsequent developments in the field have shifted from the regular grid paradigm to complex networks, a more realistic model of human interactions. Despite positive theoretical predictions for cooperation based on direct reciprocity in complex networks, experiments [2, 3, 4] have shown little evidence of network structures promoting cooperation. Conversely, other research [5, 6, 7] highlights the importance of the ability of link rewiring (i.e. making and breaking connections) as a form of indirect reciprocity, which leads to dynamic networks. Few attempts have been made, however, to test these models empirically.

Only recently, Rand et al. [8] ran a large-scale behavioural game experiment online, and demonstrated that dynamic social networks can indeed support system-wide human cooperation. This project will be built upon their approach, and will further the empirical investigation into <i>reputation-based</i> human cooperative behavior in dynamic networks. The incorporation of reputation is a natural extension to existing models, assuming that networked agents possess cognitive abilities. More specifically, this project aims to investigate the following questions: what effects does reputation have on cooperation among a network of agents? Which notion(s) of reputation are most effective in promoting cooperation? And finally, what effects does reputation have on the network structure of a population?

This project set out to experimentally study the effects of reputation on cooperative activity in dynamic networks. In addition to the effects of reputational knowledge (i.e. information on who did what in the past), the project also investigated those of social knowledge (i.e. information on who is connected to whom in the group) as well as their combined effects on cooperation in the same context.

The implementation of the project involved running a series of behavioural experiments with human subjects on the Internet, and the experimental design consists of four treatments: In the baseline (B) treatment, subjects only have access to local reputational knowledge: a list of their current neighbours with the last five actions chosen by each one of them and a list of the nonneighbours without any information on their past actions. They also have access to local social knowledge only, so they have no information on the structure of the network beyond their neighbours. In the reputation (R) treatment, they have access to global reputational knowledge, so they see a list of the last five actions for every other subject, but they are still limited to local social knowledge. In the network (N) treatment, they have access to global social knowledge, so they see a network figure that shows the connections among all of the subjects in the group, but they only have access to local reputational knowledge. Finally, in the reputation and network (RN) treatment, they have access to global reputational and social knowledge by seeing the whole network and the last five actions for all other subjects. The network figure is interactive, allowing subjects to hover over a node to highlight its neighbours and to drag nodes around to rearrange the network visualization.

Here, we systematically vary the knowledge available to subjects about reputation and the network to investigate experimentally their roles in determining cooperation in dynamic networks. Our results show that global reputational knowledge is crucial to sustaining a high level of cooperation and welfare. Cooperation is associated with the emergence of dense and clustered networks with highly cooperative hubs. Global social knowledge has no effect on the aggregate level of cooperation. A community analysis shows that the addition of global social knowledge to global reputational knowledge affects the distribution of cooperative activity: cooperators form a separate community that achieves a higher cooperation level than the community of defectors. Members of the community of cooperators achieve a higher payoff from interactions within the community than members of the less cooperative community.

For further details of our research findings, please find details of the publication here.



[1] Nowak, M.A., May, R.M. (1992), ‘Evolutionary games and spatial chaos’, Nature 437

[2] Grujic, J., Fosco, C., Araujo, L., Cuesta, J.A., Sanchez, A. (2010) ‘Social experiments in the mesoscale: humans playing a spatial prisoner’s dilemma’, PLoS ONE 5:e13749

[3] Cassar, A. (2007), ‘Coordination and cooperation in local, random and small world networks: experimental evidence’, Games Econ Behav 58

[4] Traulsen, A., Semmann, D., Sommerfeld, R.D., Krambeck, H-J., Milinski, M. (2010), ‘Human strategy updating in evolutionary games’, Proc Natl Acad Sci USA 107

[5] Ebel, H., Bornholdt, S. (2002), ‘Coevolutionary games on networks’, Phys Rev E Stat Nonlinear Soft Matter Phys 66

[6] Santos, F.C, Pacheco J.M, Lenaerts T. (2006), ‘Cooperation prevails when individuals adjust their social ties’, PLOS Computational Biology 2

[7] Hanaki, N., Peterhansl, A., Dodds, P.S., Watts, DJ. (2007), ‘Cooperation in evolving social networks’, Management Science 53

[8] Rand, D. G., Arbesman, S., Christakis, N. (2011), ‘Dynamic social networks promote cooperation in experiments with humans’, PNAS 108:48

Lead investigator

Chang Yan, Department of Computer Science and Balliol College, University of Oxford

Research team

Dr Edoardo Gallo, Department of Economics and Queens’ College, University of Cambridge

Contact details for enquiries

Please email the lead investigator, Chang Yan, for all queries regarding this project.