Hybrid attacks coordinate the exploitation of vulnerabilities across domains to undermine trust in authorities and cause social unrest. Whilst such attacks have primarily been seen in active conflict zones, there is growing concern about the potential harm that can be caused by hybrid attacks more generally and a desire to discover how better to identify and react to them. In addressing such threats, it is important to be able to identify and understand an adversary’s behaviour. Game theory is the approach predominantly used in security and defence literature for this purpose. However, the underlying rationality assumption, the equilibrium concept of game theory, as well as the need to make simplifying assumptions can limit its use in the study of emerging threats. To study hybrid threats, we present a novel agent-based model in which, for the first time, agents use reinforcement learning to inform their decisions. This model allows us to investigate the behavioural strategies of threat agents with hybrid attack capabilities as well as their broader impact on the behaviours and opinions of other agents.

The primary purpose of this model is to explain the dynamic processes within university-centered collaboration networks, with a particular focus on the complex transformation of academic knowledge into practical projects. Based on investigations of actual research projects and a thorough literature review, the model integrates multiple drivers and influencing factors to explore how these factors affect the formation and evolution of collaboration networks under different parameter scenarios. The model places special emphasis on the impact of disciplinary attributes, knowledge exchange, and interdisciplinary collaboration on the dynamics of collaboration networks, as well as the complex mechanisms of network structure, system efficiency, and interdisciplinary interactions during project formation.
Specifically, the model aims to:
- Simulate how university research departments drive the formation of research projects through knowledge creation.
- Investigate how the dynamics of collaboration networks influence the transformation of innovative hypotheses into matured projects.
- Examine the critical roles of knowledge exchange and interdisciplinary collaboration in knowledge production and project formation.
- Provide both quantitative and qualitative insights into the interactions among academia, industry, and project outputs.

To our knowledge, this is the first agent-based simulation of continuous-time PGGs (where participants can change contributions at any time) which are much harder to realise within both laboratory and simulation environments.

Work related to this simulation has been published in the following journal article:
Vu, Tuong Manh, Wagner, Christian and Siebers, Peer-Olaf (2019) ‘ABOOMS: Overcoming the Hurdles of Continuous-Time Public Goods Games with a Simulation-Based Approach’ Journal of Artificial Societies and Social Simulation 22 (2) 7 http://jasss.soc.surrey.ac.uk/22/2/7.html. doi: 10.18564/jasss.3995

Abstract:
…

This model looks at implications of author/referee interaction for quality and efficiency of peer review. It allows to investigate the importance of various reciprocity motives to ensure cooperation. Peer review is modelled as a process based on knowledge asymmetries and subject to evaluation bias. The model includes various simulation scenarios to test different interaction conditions and author and referee behaviour and various indexes that measure quality and efficiency of evaluation […]

The Soy2Grow ABM aims to simulate the adoption of soybean production in Flanders, Belgium. The model primarily considers two types of agents as farmers: 1) arable and 2) dairy farmers. Each farmer, based on its type, assesses the feasibility of adopting soybean cultivation. The feasibility assessment depends on many interrelated factors, including price, production costs, yield, disease, drought (i.e., environmental stress), social pressure, group formations, learning and skills, risk-taking, subsidies, target profit margins, tolerance to bad experiences, etc. Moreover, after adopting soybean production, agents will reassess their performance. If their performance is unsatisfactory, an agent may opt out of soy production. Therefore, one of the main outcomes to look for in the model is the number of adopters over time.

The main agents are farmers. Generally, factors influencing farmers’ decision-making are divided into seven main areas: 1) external environmental factors, 2) cooperation and learning (with slight differences depending on whether they are arable or dairy farmers), 3) crop-specific factors, 4) economics, 5) support frameworks, 6) behavioral factors, and 7) the role of mobile toasters (applicable only to dairy farmers).
Moreover, factors not only influence decision-making but also interact with each other. Specifically, external environmental factors (i.e., stress) will result in lower yield and quality (protein content). The reducing effect, identified during participatory workshops, can reach 50 %. Skills can grow and improve yield; however, their growth has a limit and follows different learning curves depending on how individualistic a farmer is. During participatory workshops, it was identified that, contrary to cooperative farmers, individualistic farmers may learn faster and reach their limits more quickly. Furthermore, subsidies directly affect revenues and profit margins; however, their impact may disappear when they are removed. In the case of dairy farmers, mobile toasters play an important role, adding toasting and processing costs to those producing soy for their animal feed consumption.
Last but not least, behavioral factors directly influence the final adoption decision. For example, high risk-taking farmers may adopt faster, whereas more conservative farmers may wait for their neighbors to adopt first. Farmers may evaluate their success based on their own targets and may also consider other crops rather than soy.

The aim of this model is to study the dynamic propagation of individual climate adaptive behaviours in different scenarios within the analytical framework of conservation motivation theory, focusing on the impact of social and experiential learning on the adoption of climate adaptive behaviours by coastal farmers.
Model for paper “Promoting climate resilience through learning-based behavioural change: Insights from an agent-based model of a coastal farming community in Guangxi, China” in Environmental Science & Policy, Volume 179, May 2026, 104375, https://doi.org/10.1016/j.envsci.2026.104375

This model extends the original Artifical Anasazi (AA) model to include individual agents, who vary in age and sex, and are aggregated into households. This allows more realistic simulations of population dynamics within the Long House Valley of Arizona from AD 800 to 1350 than are possible in the original model. The parts of this model that are directly derived from the AA model are based on Janssen’s 1999 Netlogo implementation of the model; the code for all extensions and adaptations in the model described here (the Artificial Long House Valley (ALHV) model) have been written by the authors. The AA model included only ideal and homogeneous “individuals” who do not participate in the population processes (e.g., birth and death)–these processes were assumed to act on entire households only. The ALHV model incorporates actual individual agents and all demographic processes affect these individuals. Individuals are aggregated into households that participate in annual agricultural and demographic cycles. Thus, the ALHV model is a combination of individual processes (birth and death) and household-level processes (e.g., finding suitable agriculture plots).

As is the case for the AA model, the ALHV model makes use of detailed archaeological and paleoenvironmental data from the Long House Valley and the adjacent areas in Arizona. It also uses the same methods as the original model (from Janssen’s Netlogo implementation) to estimate annual maize productivity of various agricultural zones within the valley. These estimates are used to determine suitable locations for households and farms during each year of the simulation.

The intention of this model is to create an universal basis on how to model change in value prioritizations within social simulation. This model illustrates the designing of heterogeneous populations within agent-based social simulations by equipping agents with Dynamic Value-based Cognitive Architectures (DVCA-model). The DVCA-model uses the psychological theories on values by Schwartz (2012) and character traits by McCrae and Costa (2008) to create an unique trait- and value prioritization system for each individual. Furthermore, the DVCA-model simulates the impact of both social persuasion and life-events (e.g. information, experience) on the value systems of individuals by introducing the innovative concept of perception thermometers. Perception thermometers, controlled by the character traits, operate as buffers between the internal value prioritizations of agents and their external interactions. By introducing the concept of perception thermometers, the DVCA-model allows to study the dynamics of individual value prioritizations under a variety of internal and external perturbations over extensive time periods. Possible applications are the use of the DVCA-model within artificial sociality, opinion dynamics, social learning modelling, behavior selection algorithms and social-economic modelling.

This model is an agent-based simulation that consists of agents who play the spatial prisioner’s dilemma game with coalition formation. The coalition dynamics are mainly influenced by how much the agents trust their leaders. The main objective is provide a simulation model to enable the analysis of the impacts that the use of trust may cause in coalition formation.

Agent-based model of intergroup conflict in divided communities.