Rigorous Agent-Based Modeling of Green Practice Diffusion: Analytical Approximations and Validation on Organizational Networks
Angelika Abramiuk-Szurlej1 ✉️, Katarzyna Sznajd-Weron2 ✉️, Mikołaj Szurlej3 ✉️ 1Faculty of Management, Wroclaw University of Science and Technology, Poland 2Faculty of Management, Wroclaw University of Science and Technology, Poland 3Faculty of Architecture, Wroclaw University of Science and Technology, Poland
Cite as: Abramiuk-Szurlej, A., Sznajd-Weron, K. and Szurlej, M. (2025, May). Rigorous Agent-Based Modeling of Green Practice Diffusion: Analytical Approximations and Validation on Organizational Networks. In SCS 2025, 2nd International Conference on Social Contexts of Science (p. 29). Wrocław University of Science and Technology, Poland.
Abstract
Agent-based modeling (ABM) is increasingly used to manage pro-environmental behavior change, especially in energy-related contexts. A key advantage of ABM is its ability to model local consumer interactions, which play a crucial role in promoting pro-environmental behavior driven by peer pressure and social norms. However, ABM is often criticized for its lack of rigorous validation and sensitivity analysis. To address these challenges, we refine an existing ABM of green product and practice diffusion, applying Pair Approximation (PA) and Monte Carlo Simulations (MCS) to real-world organizational networks. This approach provides new insights into how well analytical methods can capture diffusion dynamics in social systems. The model considers two main factors: (1) social interactions among agents, crucial for the spread of energy-related behaviors, and (2) the probability of engagement in a certain behavior. The original model assumed engagement following a logistic function. We propose a modified version where engagement probability is treated as an independent parameter not defined by any specific functional form. The new version can be seen as a general innovation diffusion model that extends beyond pairwise interactions. The model simulates agents in a social network, where each agent has neighbors defined by the network structure. Interactions between agents allow social influence to shape their decisions. Each agent can either: 1) randomly decide to adopt or reject an innovation, or 2) respond to social influence by conforming to a unanimous group of neighbors or maintaining their previous state. The system evolves through random sequential updating, simulating continuous time. In each update, a randomly chosen agent changes its decision. This process continues until the system stabilizes and reaches a steady state. We use two analytical methods