The joint AIMES/GLP Working Group on Large-scale Behavioural Models of Land Use Change held its second webinar in April 2021 to discuss exciting new advances in social simulation and computational modelling that can support better representation of human behaviour in the land system.
First, Robert Axtell presented new work on large-scale agent based models in social science:
The U.S. private sector consisted, before the pandemic, of somewhat more than 120 million employees working in 6 million firms, making nearly $30 trillion of gross output, with another $1 trillion of output coming from firms that had no formal employees. Using data on all of these entities this research describes a model of the entire private sector, i.e., at 1-to-1 or full-scale (aka digital twin, mirror world) that closely reproduces dozens of gross statistical features of American firms and workers, things like firm sizes and ages, employee tenure and turnover, high rates of new firm entrance and endogenous firm exit, a variety of inter-firm networks, and so on. A book on this subject, Dynamics of Firms from the Botton Up: Data, Theories, and Models, is forthcoming from MIT Press.
Presentation: Agent-Based Modeling of Economic Phenomena at Very-Large (Full) Scale (PDF, 11 MB)
Papers:
- 120 Million Agents Self-Organize into 6 Million Firs: A Model of the U.S. Private Sector
- Handbook Chapter, Endogenous Firm Dynamics and Labor Flows via Heterogeneous Agents
Second, Brian Mac Namee described the use of machine learning and ABM in computer game design:
For almost every type of model we build we need to make decisions about the level of details with which we will model the real world in a virtual model. Badler (1997) refers to this as virtual fidelity. In the talk this was demonstrated in game agent models for enhancing player experience, machine learning models for behaviour prediction, and an agent-based model for modelling infectious disease transmission.
Badler, Norman I. “Real-time virtual humans.” Proceedings The Fifth Pacific Conference on Computer Graphics and Applications. IEEE, 1997.
Presentation: How Much Reality is Enough? Games, Cows and Disease (PDF, 7 MB)
Papers:
- O’Sullivan, Carol, Justine Cassell, Hannes Vilhjalmsson, J. Dingliana, Simon Dobbyn, Brian McNamee, Christopher Peters, and Thang Giang. “Levels of detail for crowds and groups.” In Computer Graphics Forum, vol. 21, no. 4, pp. 733-741. https://onlinelibrary.wiley.
com/doi/full/10.1111/1467- 8659.00631 - Ryan, C., Gúeret, C., Berry, D., Corcoran, M., Keane, M. T., & Mac Namee, B. (2021). Predicting Illness for a Sustainable Dairy Agriculture: Predicting and Explaining the Onset of Mastitis in Dairy Cows. In Proceedings of Explainable Agency in Artificial Intelligence Workshop @ AAAI 2021 https://arxiv.org/abs/2101.
02188 - Hunter, Elizabeth, Brian Mac Namee, and John Kelleher. “A Hybrid Agent-Based and Equation Based Model for the Spread of Infectious Diseases.” Journal of Artificial Societies & Social Simulation 23.4 (2020). https://arrow.tudublin.ie/
sciendoc/241/ - Hunter, Elizabeth, Brian Mac Namee, and John Kelleher. “An open-data-driven agent-based model to simulate infectious disease outbreaks.” PloS one 13.12 (2018): e0208775. https://journals.plos.org/
plosone/article?id=10.1371/ journal.pone.0208775
Third, Thomas Clemen presented a simulation platform capable of modelling vast multi-agent systems:
The MARS Group at Hamburg University of Applied Sciences, Germany, provides a software framework for creating agent-based simulation models on a very large scale. The presentation introduces some principal aspects from a computer science and system theoretical perspective. Additionally, the challenges and opportunities of interdisciplinary teams were discussed. A short-course format to develop the necessary skills was proposed.
Presentation: Virtual Humans on MARS: Concepts, AI, and Interdisciplinarity (PDF, 2 MB)
Papers:
- A cross-scale modeling framework for decision support on elephant management in Kruger National Park, South Africa
- Utilizing Spatio-Temporal Data in Multi-Agent Simulation
- Large-Scale Traffic Simulation for Smart City Planning with MARS
- Firewood Collection in South Africa: Adaptive Behavior in Social-Ecological Models
The webinar concluded with a discussion among presenters and attendees about how these advances can be used to develop a new generation of land system science models.
Become a Member
Join the Large-scale Behavioural Models of Land Use Change working group to be kept updated on the latest news, events and jobs of interest!
You can also view the recording and download materials from the first webinar.