Agent-based Modelling in Government

Kavin Narasimhan

Agent-based Modelling (ABM) is a bottom-up approach used to replicate the behaviour of complex systems, like societies, in silico by simulating the behaviour of individual units called agents, which represent entities like people, businesses, and policymakers. The interactions between agents and their environment result in macro-level patterns or outcomes that are computationally irreducible and hard to describe using mathematical equations for reasons including agent heterogeneity, adaptive behaviours, nonlinearity, and path dependence. ABM captures these characteristics effectively to simulate the emergence of complex phenomena from micro-level assumptions and is easy to implement with the availability of right data. However, its concepts are challenging to master, and there are limitations. Building a model at the right level of description with the right amount of detail requires resources and interdisciplinary expertise; the results obtained from simulating human behaviour can range from being purely qualitative to highly quantitative and thus require human expertise to interpret the results to generate insights; validating the macro-level patterns and micro-level assumptions in ABM requires fine-grained data; and finally, ABM is computationally expensive.

Against the backdrop of the challenges and opportunities of ABM, our research seeks to explore the application of ABM in Government by foregrounding the practical experiences of modellers and model users in using ABM to aid planning and decision-making. Kavin will present emerging findings based on interviews conducted with agent-based modellers and model users for/in Government. There will be opportunities to hear about and reflect on the use of ABM and other bottom-up modelling approaches to capture heterogeneity, the relevance of models at specific stages of the policy cycle, the collaborative relationship between model developers and model users and its consequences for model application and legacy, issues of uncertainty in bottom-up modelling, and implications for quality assurance.

Speaker bio: Dr Kavin Narasimhan is an Assistant Professor at the Centre for Interdisciplinary Methodologies (CIM)Link opens in a new window at the University of Warwick. Her research focuses on computational modelling for public policy using methods like agent-based modelling, causal loop diagramming, network mapping and analysis. She is also interested in knowledge co-production using participatory research methods. Kavin's work reviewing the use of ABM in Government was supported by the UK Research and Innovation (UKRI) Economic and Social Research Council (ESRC) as part of the Policy Fellowships pilot programmeLink opens in a new window [ES/W008548/1].