The Episim framework builds upon the established transportation simulation MATSim and is capable of tracking agents’ movements within a network and thus computing infection chains. Several characteristics of the virus and the environment can be parametred, whilst the infection dynamics is computed based upon a compartment model. The spread of the virus can be mitigated by restricting the agents’ activity in certain places.

The SAFIRe model (Simulation of Agents for Fertility, Integrated Energy, Food Security, and Reforestation) is an agent-based model co-developed with rural communities in Senegal’s Groundnut Basin. Its purpose is to explore how local farming and pastoral practices affect the regeneration of Faidherbia albida trees, which are essential for maintaining soil fertility and supporting food security through improved millet production. The model supports collective reflection on how different social and ecological factors interact, particularly around firewood demand, livestock pressure, and agricultural intensification.

The model simulates a 100-hectare agricultural landscape where agents (farmers, shepherds, woodcutters, and supervisors) interact with trees, land parcels, and each other. It incorporates seasonality, crop rotation, tree growth and cutting, livestock feeding behaviors, and farmers’ engagement in sapling protection through Assisted Natural Regeneration (ANR). Two types of surveillance strategies are compared: community-led monitoring and delegated surveillance by forestry authorities. Farmer engagement evolves over time based on peer influence, meeting participation, and the success of visible tree regeneration efforts.

SAFIRe integrates participatory modeling (ComMod and ComExp) and a backcasting approach (ACARDI) to co-produce scenarios rooted in local aspirations. It was explored using the OpenMole platform, allowing stakeholders to test a wide range of future trajectories and analyze the sensitivity of key parameters (e.g., discussion frequency, time in fields). The model’s outcomes not only revealed unexpected insights—such as the hidden role of farmers in tree loss—but also led to real-world actions, including community nursery creation and behavioral shifts toward tree care. SAFIRe illustrates how agent-based modeling can become a tool for social learning and collective action in socio-ecological systems.

CapOvCWD is an agent-based model that simulates a captive cervid herd composed of adults and fawns. The model deer population is initialized using data on herd size and composition from captive facility records. Individual deer domiciliary history and annual CWD testing records inform the herd size and sample size (for CWD testing), respectively. The model can be used to iteratively estimate the facility level annual CWD detection probability. Detection probability estimates can be further refined by incorporating multiyear CWD testing data. This approach can be particularly useful for interpreting negative test results from a subset of the captive herd. Facility level detection probability estimates provide a comprehensive and standardized risk metric that reflects the likelihood of undetected CWD in the facility.

An Agent Based Model that explores the deployment of hydrogen among a regional industrial cluster in the Netherlands, consisting of 15 companies. The companies seek to decarbonize by replacing their natural gas by hydrogen.
The model integrates technical characteristics as well as company motivations to transition to hydrogen. The baseline model only considers individual investments where company can locally produce hydrogen. If they reach the backbone threshold, companies can also consider buying hydrogen through a connection to the national hydrogen network. The second scenario considers that companies can participate in a joint investment to get an electrolyzer to locally produce the hydrogen.
Two experiments look at the impact of the sectoral configuration and at the impact of subsidy conditions on the region’s hydrogen transition

We introduce a model of prediction markets that uses opinion dynamics as its underlying mechanism for price formation. We base the opinion dynamics on the Deffuant model of bounded rationality. We have used this model to show that price formation in prediction markets can be robustly explained by opinion dynamics, and that the model can also explain phase transitions depending on just two parameters.

A model to show the effects of flood risk on a housing market; the role of flood protection for risk reduction; the working of the existing public-private flood insurance partnership in the UK, and the proposed scheme ‘Flood Re’.

The model provides instruments for the simulation of interbank network evolution. There are tools for dynamic network analysis, allowing to evaluate graph topological invariants, thermodynamic network features and combinational node-based features.

Emulation is one of the simplest and most common mechanisms of social interaction. In this paper we introduce a descriptive computational model that attempts to capture the underlying dynamics of social processes led by emulation.

The wisdom of the crowd refers to the phenomenon in which a group of individuals, each making independent decisions, can collectively arrive at highly accurate solutions—often more accurate than any individual within the group. This principle relies heavily on independence: if individual opinions are unbiased and uncorrelated, their errors tend to cancel out when averaged, reducing overall bias. However, in real-world social networks, individuals are often influenced by their neighbors, introducing correlations between decisions. Such social influence can amplify biases, disrupting the benefits of independent voting. This trade-off between independence and interdependence has striking parallels to ensemble learning methods in machine learning. Bagging (bootstrap aggregating) improves classification performance by combining independently trained weak learners, reducing bias. Boosting, on the other hand, explicitly introduces sequential dependence among learners, where each learner focuses on correcting the errors of its predecessors. This process can reinforce biases present in the data even if it reduces variance. Here, we introduce a new meta-algorithm, casting, which captures this biological and computational trade-off. Casting forms partially connected groups (“castes”) of weak learners that are internally linked through boosting, while the castes themselves remain independent and are aggregated using bagging. This creates a continuum between full independence (i.e., bagging) and full dependence (i.e., boosting). This method allows for the testing of model capabilities across values of the hyperparameter which controls connectedness. We specifically investigate classification tasks, but the method can be used for regression tasks as well. Ultimately, casting can provide insights for how real systems contend with classification problems.

This is an extension of the basic Suceptible, Infected, Recovered (SIR) model. This model explores the spread of disease in two spaces, one a treatment, and one a control. Through the modeling options, one can explore how changing assumptions about the number of susceptible people, starting number of infected people, the disease’s infection probability, and average duration impacts the outcome. In addition, this version allows users to explore how public health interventions like social distancing, masking, and isolation can affect the number of people infected. The model shows that the interactions of agents, and the interventions can drastically affect the results of the model.

We used the model in our course about COVID-19: https://www.csats.psu.edu/science-of-covid19