Towards a Validation Taxonomy for Agent-Based Models of Pedestrian Movement in Cities (Short Paper)

Abstract

Agent-Based Models (ABMs) are increasingly used to simulate and analyse pedestrian movement in cities and to support urban planning and policy decisions. However, their credibility depends critically on rigorous empirical validation, for which no standardised framework currently exists. This paper addresses this challenge by providing a structured synthesis of existing validation strategies, tailored to ABMs of pedestrian movement in cities, focusing on the comparison of simulated and reference behaviours. First, it reviews and systematises operational validation approaches spanning three model output levels: micro, meso, and macro. Then, it develops a validation taxonomy that links specific model outputs - individual behavioural trajectories, segment-level flows, and aggregate system indicators - to data sources, potential validation methods, and appropriate behavioural claims. Finally, it demonstrates the implementation of these approaches using illustrative examples. Here, it is argued that not only do different model outputs require distinct data sources and methodological strategies, but the generalisations a validation exercise can support are bounded by what is validated and how. The paper advances a practical validation taxonomy that, whilst setting standards in the discipline, can be flexibly adapted to diverse modelling contexts. In doing so, it supports the development of more robust, transparent, and policy-relevant agent-based models of pedestrian movement in cities.