Taking Agent-Based Social Simulation to the Next Level Using Exascale Computing: Potential Use-Cases, Capacity Requirements and Threats

Matt Hare (The James Hutton Institute), Doug Salt (The James Hutton Institute), Ric Colasanti (University of Glasgow), Richard Milton (University College London), Mike Batty (University College London), Alison Heppenstall (University of Glasgow), Gary Polhill (The James Hutton Institute)

Abstract

Exascale computing (10 18 FLOPS) officially arrived in 2022 when Oak Ridge's Frontier achieved that performance benchmark, and other countries seek their own exascale capabilities. High-end computing is typically used by the natural sciences, but empirical Agent-Based Social Simulation (ABSS) is a social science application. Empirical ABSS has a long history, but was prominent during the Covid crisis. In future crises, policy options could be evaluated within rapid policy design windows using exascale computing. We report on a group model-building exercise, co-constructing a causal loop model, to explore visions of the potential of exascale computing in ABSS, identifying potential use cases, capabilities, capacity requirements and threats.