Training LSTM-neural networks on early warning signals of declining cooperation in simulated repeated public good games

Manfred Füllsack,Marie Kapeller,Simon Plakolb,Georg Jäger

doi:10.1016/j.mex.2020.100920

Abstract

We present results of attempts to expand and enhance the predictive power of Early Warning Signals (EWS) for Critical Transitions (Scheffer et al. 2009) through the deployment of a Long-Short-Term-Memory (LSTM) Neural Network on agent-based simulations of a Repeated Public Good Game, which due to positive feedbacks on experience and social entrainment transits abruptly from majority cooperation to majority defection and back. Our method extension is inspired by several known deficiencies of EWS and by lacking possibilities to consider micro-level interaction in the so far primarily used simulation methods. We find that•The method is applicable to agent-based simulations (as an extension of equation-based methods).•The LSTM yields signals of imminent transitions that can complement statistical indicators of EWS.•The less tensely connected part of an agent population could take a larger role in causing a tipping than the well-connected part.

Highlights

We report on attempts of training an LSTM-ANN for predicting critical transitions in time series generated with an ABM-model of a Repeated Public Good Game
This information could be moved to the end under Additional Information

Summary

Method Article

The method is applicable to agent-based simulations (as an extension of equation-based methods). The LSTM yields signals of imminent transitions that can complement statistical indicators of EWS. Subject Area Social Sciences More specific subject area: Method name: Name and reference of original method. Select one of the following subject areas: Social Sciences Describe narrower subject area. Systems Sciences Please specify a name of the method that you customized. Warning Signals for Critical Transitions If applicable, include full bibliographic details of the main reference(s) describing the original method from which the new method was derived. Include links to resources necessary to reproduce the method (e.g. data, software, hardware, reagent)

Introduction

Results on unseen data

Summary