Validation of Agent-Based Simulation through Human Computation: An Example of Crowd Simulation

Abstract

Agent-based modeling as a methodology for understanding natural phenomena is becoming increasingly popular in many disciplines of scientific research. Validation is still a significant problem for agent-based modelers and while various validation methodologies have been proposed, none have been widely adopted. Data plays a key role in the validation of any simulation system, typically large amounts of observable real world data are necessary to compare with model outputs. However, the complex nature of the studied natural systems will often make data collection difficult. This is certainly true for crowd and egress simulation, where data is limited and difficult to collect. In this paper we propose a new technique for validation of agent-based models, particularly those which relate to human behavior. This methodology adopts ideas from the field of Human Computation as a means of collecting large amounts of contextual behavioral data. The key principle is to use games as a means of framing behavioral questions to try and capture people’s natural and instinctive decisions. We outline some key design challenges for such games and present one example game in the form of Escape. Escape is an egress based game where people are tasked to escape from rooms inhabited by other people. We show some preliminary studies which highlight some interesting applications of the game in addressing validation of behavioral based crowd and egress simulation.