Coordination and Consensus: Computational Approaches

Time: 15:00 until 17:15.

Programme

15:00 - 16:00 Elliott Wagner "The Evolutionary Dynamics of Collective Action"
16:15 - 17:15 Andreas Flache "Modelling polarization in opinion formation"

___

ABSTRACTS

Elliott Wagner "The Evolutionary Dynamics of Collective Action"

Stag hunt games are often used to model the strategic aspects of collective actions such as coalition formation and group hunting.  But the evolution of cooperation in these games remains somewhat mysterious.  This mystery is due to the fact that under all standard evolutionary and learning dynamics the cooperative equilibrium of a stag hunt game has a smaller basin of attraction than that of defection. Furthermore, many different stochastic dynamics, such as the stochastic replicator dynamic, predict that populations will spend almost all the time defecting in 2-player stag hunt games.  But in this talk I prove that this is not true of N-player stag hunts.  Populations that evolve (or learn) according to the stochastic replicator dynamic will spend almost all of their time cooperating in N-player stag hunt games where N>2.  This perhaps counterintuitive result indicates that cooperation may be more likely to evolve in strategic settings with more than two participants than it is to evolve in pairwise interactions.


Andreas Flache "Modelling polarization in opinion formation"

Many existing models of opinion formation are built on the assumption that social influence fosters consensus. In these models, interacting individuals move their opinions towards those of their interaction partners, so that disagreement is reduced. Models of such 'positive' social influence can explain polarization and persistent dissensus only  for exceptional social conditions, such as total isolation of subgroups from outside influences. I will present some mechanisms that we have proposed in recent work that can explain polarization under a wider range of conditions, and discuss corresponding social conditions. Agent-based computational models of polarization based on negative influence, on persuasive argument exchange and on strive for individualization will be presented. I will also show  evidence from experiments designed to test some of these models.