What Do Agent-Based and Equation-Based Modelling Tell Us About Social Conventions: The Clash Between ABM and EBM in a Congestion Game Framework

Federico Cecconi,Giulia Andrighetto,Rosaria Conte,Marco Campenni

doi:10.18564/jasss.1585

Abstract

In this work simulation-based and analytical results on the emergence steady states in traffic-like interactions are presented and discussed. The objective of the paper is twofold: i) investigating the role of social conventions in coordination problem situations, and more specifically in congestion games; ii) comparing simulation-based and analytical results to figure out what these methodologies can tell us on the subject matter. Our main issue is that Agent-Based Modelling (ABM) and the Equation-Based Modelling (EBM) are not alternative, but in some circumstances complementary, and suggest some features distinguishing these two ways of modeling that go beyond the practical considerations provided by Parunak H.V.D., Robert Savit and Rick L. Riolo. Our model is based on the interaction of strategies of heterogeneous agents who have to cross a junction. In each junction there are only four inputs, each of which is passable only in the direction of the intersection and can be occupied only by an agent one at a time. The results generated by ABM simulations provide structured data for developing the analytical model through which generalizing the simulation results and make predictions. ABM simulations are artifacts that generate empirical data on the basis of the variables, properties, local rules and critical factors the modeler decides to implement into the model; in this way simulations allow generating controlled data, useful to test the theory and reduce the complexity, while EBM allows to close them, making thus possible to falsify them.

Highlights

1.1 Ten years ago, during the first Multi-Agent Systems and Agent-Based Simulation Workshop (MABS '98), Parunak H.V.D., Robert Savit and Rick L
1.6 In this paper, we suggest a different solution to the subject matter, arguing that in specific coordination problem situations, in 4-strategy game structures, a conventional equilibrium is not granted to emerge
7.2 Regarding section (a), in our Agent-Based Modelling (ABM) model we investigate the role of conditioned strategies - which can be seen as social conventions in Lewis's sense - in the solution of a congestion game, in comparison and in combination with unconditioned strategies

Summary

Introduction

1.1 Ten years ago, during the first Multi-Agent Systems and Agent-Based Simulation Workshop (MABS '98), Parunak H.V.D., Robert Savit and Rick L. In our model (congestion game), we adapt the projector dynamic framework to compute a payoff matrix that starts from a payoff function with two parameters, i.e. the distribution of the strategies at time t and the combination of strategies at the same time. In this case, the simulation starts with 100 right-watchers, 100 left-watchers, 100 doves and 100 hawks and both doves and hawks survive). Left-watchers and hawks and doves: 81 times (31.64%); Figure 7 In this case, the simulation starts with 100 right-watchers, 100 left-watchers, 100 doves and 100 hawks and only rightwatchers die http://jasss.soc.surrey.ac.uk/13/1/6.html. The relation obtained for the projection dynamic lets us state that the games with only one strategy have an unstable stationary point (the derivative of the frequencies has opposite sign, see figure 9). We find a stable steady state with Conditioned strategies combined with the Compliant strategy (see figure 10)

The analytic model suggests some other hints:

Findings

Discussion