Our mission is to help computational modelers develop, document, and share their computational models in accordance with community standards and good open science and software engineering practices. Model authors can publish their model source code in the Computational Model Library with narrative documentation as well as metadata that supports open science and emerging norms that facilitate software citation, computational reproducibility / frictionless reuse, and interoperability. Model authors can also request private peer review of their computational models. Models that pass peer review receive a DOI once published.
All users of models published in the library must cite model authors when they use and benefit from their code.
Please check out our model publishing tutorial and feel free to contact us if you have any questions or concerns about publishing your model(s) in the Computational Model Library.
Displaying 10 of 535 results from clear search
An agent-based model of a sports league, a public Elo forecaster and a bookmaker. Teams have hidden true strengths that drift between seasons; the forecaster updates Elo ratings after every game; the bookmaker prices each game from the Elo forecast plus a margin (overround), spread either proportionally or loaded onto longshot prices. The model asks where the favourite-longshot pattern in betting prices comes from. With a forecaster whose ratings lag true strength, longshots win more often than their de-vigged price implies; a bookmaker that loads its margin onto longshots cancels that bias, so favourite and longshot bets both lose roughly the margin. Outputs: a calibration table (de-vigged price band vs. observed win rate) and the average return of flat bets on favourites and longshots. Python 3 standard library only; all parameters are command-line arguments; the README documents the model with a short ODD protocol and an example run. Companion empirical study and open data (CC BY 4.0): Bet Better market calibration study. Teaching model, not a betting system.
Shared vocabularies are not designed by any central authority; they emerge from the bottom up as communities of speakers repeatedly try to understand one another. We study this process with an agent-based Naming Game in which a population that begins with no common vocabulary self-organises a shared language through local pairwise interaction, and we ask how the social network governs the outcome under communication noise and under selection on communicative strategy. The study combines baseline experiments with an extended suite built specifically to meet the standards of replication and statistical inference: every reported quantity is averaged over many independent seeded runs with confidence intervals, and the central claim is tested with a formal interaction statistic. Convergence time scales with population size as a power law with exponent close to four-thirds, matching analytical results for the Naming Game on networks. Restricted networks sustain many coexisting regional dialects, which we quantify with a cluster-based metric and show to be metastable, decaying to a single language as the interaction budget grows. Our central methodological contribution concerns noise robustness. A naive comparison suggests that denser networks simply tolerate more noise, but by comparing topologies at matched mean degree and by sweeping degree directly we separate two effects: connectivity raises robustness with sharply diminishing returns, while at identical degree locally clustered topologies remain significantly less robust than well-mixed ones. The effect of topology therefore survives the removal of the degree confound. Finally, an evolutionary extension shows that selection for communicative success drives inventiveness towards a low level that a frequency-dependent reward for useful innovation raises only slightly, so conservatism is robust rather than an artefact of the fitness function.
A NetLogo 3D 7.0 model of a market with several floors. Buyers walk, keep their distance from one another, choose among the stalls they can see and change floors. Social proof acts on where buyers go (attraction) and on whether they buy (conversion), and crowding lowers the chance of a purchase. The model tests whether social proof is sufficient to congest such a market and separates the roles of influence, layout and local density. The archive contains the model with every BehaviorSpace experiment of the companion paper, all simulation outputs, the pre-analysis plan, an ODD+D description and the Python code that reproduces every reported number, table and figure.
The model’s primary aim was not to be entirely realistic; instead, it was designed to point to new questions, test the usefulness of the tool, and lay the groundwork for new insights. As such, this model is targeted at ecologists, ornithologists, and urban planners who want to test and simulate how different birds face the challenge of a densely urbanized city that aims to enhance its natural connectivity and foster a wide range of biodiversity. We aimed to identify bird movement at the city-wide scale, encompassing representatives from the entire avian community, while examining differences in movement between totally random models and an informed model, defined by expert criteria.
This archive contains the SAFE-ABM computational model developed to evaluate targeted intervention strategies for protecting essential workers during epidemic conditions, accompanying the published article “Essential Workers at Risk: An Agent-Based Model (SAFE-ABM) with Uncertainty Quantification” in the Journal of Artificial Societies and Social Simulation, 29(4), Article 7 (2026). https://doi.org/10.18564/jasss.6276. SAFE-ABM is an agent-based model that simulates disease transmission through structured interactions across households, workplaces, and schools, while incorporating stochastic disease progression and uncertainty in transmission, recovery, and mortality parameters.
The model evaluates five intervention scenarios: Scenario 1, no restrictions; Scenario 2, school-from-home; Scenario 3, essential-workers-only movement; Scenario 4, rotational shifts among essential workers with quarantine of symptomatic active workers; and Scenario 5, rotational shifts among essential workers without quarantine. The archive includes scenario-specific simulation scripts, comparison-analysis scripts, saved derived outputs, README documentation, software requirements, and reproducibility notes.
The included comparison scripts reproduce the main numerical summaries and manuscript outputs, including median epidemic trajectories, predictive-uncertainty plots, cumulative recovery and death summaries, inflection-point summaries, Welch-test comparisons, and mutual-information sensitivity rankings. Saved derived outputs are included so that the main figures and tables can be reproduced without rerunning all large-scale simulations.
…
This model extends Schelling’s segregation model with mobility sharing. Agents of two groups occupy cells of a 32 × 32 grid and repeatedly choose activity destinations subject to a preference threshold, the minimum share of ingroup neighbours they accept. Trips are matched into shared rides between members of different groups, using one of four matching algorithms on a shareability network (maximal cross-group matching, maximum weight matching on contact time, maximum cardinality matching, or random greedy matching) with a maximum detour constraint. After each trip, an agent’s preference threshold changes with the time spent in contact with outgroup members.
Starting from complete segregation, the model examines under which conditions contact during shared trips leads to integration. Besides the main scenario, in which contact always lowers the threshold, it includes two extensions. In the first, a share of adversarial agents becomes less tolerant with contact, and this share grows or shrinks with the current level of segregation, governed by a branching dissimilarity index and a conversion rate. In the second, each contact raises or lowers the threshold with a probability that depends on the agent’s current attitude relative to a compatibility threshold.
Segregation is measured with the index of dissimilarity and the Gini coefficient. The model file includes BehaviorSpace experiments for each analysis in the associated article. The matching step uses the NetworkX library through the Python extension of NetLogo (requires Python 3 with numpy and networkx 3.x).
This model represents individuals from two different groups interacting and sharing attitudes toward both groups. Through these interactions, individuals adjust their attitudes, which can lead to different outcomes in opinion dynamics. The model can study both intra-group and inter-group opinion dynamics. Under certain conditions, agents with initially moderate attitudes might enter a negative spiral, leading to increasingly extreme intergroup attitudes. Alternatively, these moderate attitudes can foster consensus between groups, resulting in agreement on intergroup attitudes. This model serves as a tool for exploring opinion polarization, particularly in contexts where attitudes evolve through interaction rather than remaining static.
By adjusting different parameters, one can examine the effects of variations in attitude distributions; group size and relative group size; contributions of xenophobia, opinion distance, and individual attitudes towards the outgroup; as well as ratio between xenophobia and attitudinal bias to the overall intergroup bias.
The model accompanies the paper When contact backfires, and when it does not: A computational model of the opinion dynamics of intergroup attitudes.
This repository contains a Python simulation script for studying network-based linguistic diversity. The model extends the classic Greenberg Diversity Index by introducing diversity measures defined over social ties rather than only over population shares. It also includes two extensions: a competence-conditioned version and a multilayer version.
Given the importance of experiencing fulfilling and productive workdays, scholars have focused on ways to boost employees’ daily engagement and performance, suggesting this may be accomplished by increasing meaningfulness of their work tasks. However, an alternative approach may be needed in multifaceted work contexts where employees need to complete sets of varied tasks each day, including tasks that lack meaningfulness. To address this issue, we integrate the workday design perspective with research on residual engagement to examine implications of strategically sequencing tasks in terms of their meaningfulness. We theorize that because of a process we refer to as engagement cascade, performing the day’s most highly meaningful tasks earlier and in a concentrated manner will tend to produce higher levels of daily engagement relative to sequences in which the most highly meaningful tasks occur later or are disbursed. We use agent-based modeling to formalize the engagement cascade process and generate specific predictions regarding how daily engagement is impacted by different task meaningfulness sequences. Findings from an intensive longitudinal design study with NASA and ROSCOSMOS crew members (Study 1), an experience sampling field experiment with full-time employees (Study 2), and two controlled laboratory experiments with undergraduates (Study 3 and Supplement to Study 3) provide support for our hypotheses. Specifically, the decreasing task meaningfulness sequence engendered higher levels of daily engagement, and in turn, daily performance, relative to increasing, inverted U-shaped, and U-shaped sequences.
HOSS(-5) is an exploratory agent-based prototype of a small artificial society, written in Python. Each agent is a “ph” (punto humano, “human point”) that is born, is educated, works, consumes and dies. Its yearly decisions depend on a psychological profile (archetype × character), its economic status and its intentionality. Agents interact indirectly through families, firms in 10 sectors, a labour market with about 200 job categories, and a State that provides public services, employment and a Universal Basic Income, adjusting taxes and UBI each year to seek fiscal balance.
The model simulates a few thousand agents over decades of virtual time and records yearly aggregates (population, GDP, income, Gini index, HDI). A separate analysis script explores the results and fits Gamma distributions to individual incomes, comparing the shape parameter with Gini and HDI.
HOSS does not aim to predict real societies or claim empirical validation. It is a sandbox for comparing scenarios and thinking about the organizational design of society from an engineering perspective. Known limitations: no direct social relations between agents, prices and production follow simple rules rather than supply-demand dynamics, and organizations are black boxes.
…
Displaying 10 of 535 results from clear search