Machine-assisted agent-based modeling

Machine-assisted agent-based modeling: Opening the black box

Taghikhah, Firouzeh (University of Sydney)
Voinov, Alexey (University of Twente)
Filatova, T. (TU Delft Policy Analysis)
Polhill, J. Gareth (The James Hutton Institute)

2022

While agent-based modeling (ABM) has become one of the most powerful tools in quantitative social sciences, it remains difficult to explain their structure and performance. We propose to use artificial intelligence both to build the models from data, and to improve the way we communicate models to stakeholders. Although machine learning is actively employed for pre-processing data, here for the first time, we used it to facilitate model development of a simulation model directly from data. Our suggested framework, ML-ABM accounts for causality and feedback loops in a complex nonlinear system and at the same time keeps it transparent for stakeholders. As a result, beside the development of a behavioral ABM, we open the ‘blackbox’ of purely empirical models. With our approach, artificial intelligence in the simulation field can open a new stream in modeling practices and provide insights for future applications.

Behavioral analytics
Conceptual modeling
Interpretable artificial intelligence
Social communications
Systems thinking

http://resolver.tudelft.nl/uuid:eab482af-3ce8-436d-a45b-1d3c0ec64767

https://doi.org/10.1016/j.jocs.2022.101854

2023-07-01

1877-7503

Journal of Computational Science, 64

Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.

Institutional Repository

journal article

© 2022 Firouzeh Taghikhah, Alexey Voinov, T. Filatova, J. Gareth Polhill