An agent-based model of saving and dissaving behaviour under quasi-hyperbolic (β–δ) discounting. Building on the individual decision problem of Cao and Werning (2018), the model embeds present-biased agents in a Watts–Strogatz small-world network and adds three configurable mechanisms of social influence — information diffusion, peer comparison, and social-norm conformity — across five heterogeneous behavioural profiles (Planners, Moderates, Procrastinators, Inverse Procrastinators, and Impulsive agents).
Each profile’s saving policy is approximated by value-function iteration over a discretised wealth grid; the solved policies are cached and applied as agents interact over their network neighbourhoods. The model tests whether each social mechanism can alter the saving and wealth trajectories that present-biased agents would otherwise follow in isolation, and characterises the direction and size of each effect on median wealth, wealth inequality (Gini), and the incidence of severely depleted agents.
The deposit includes the core model (Model.py), an analysis and visualisation pipeline (analyze_results.py), a standalone ODD description (ODD.md), and pinned dependencies.

Overview

The Weather model is a procedural generation model designed to create realistic daily weather data for socioecological simulations. It generates synthetic weather time series for solar radiation, temperature, and precipitation using algorithms based on sinusoidal and double logistic functions. The model incorporates stochastic variation to mimic unpredictable weather patterns and aims to provide realistic yet flexible weather inputs for exploring diverse climate scenarios.

The Weather model can be used independently or integrated into larger models, providing realistic weather patterns without extensive coding or data collection. It can be customized to meet specific requirements, enabling users to gain a better understanding of the underlying mechanisms and have greater confidence in their applications.

…

An agent-based microsimulation of insecticide-treated net (ITN) distribution and adoption in Kenya (2003–2024), integrating the Theory of Planned Behaviour, Rogers diffusion, Weibull net decay, and a GPS-based two-layer social network. 8,561 household agents calibrated via Approximate Bayesian Computation to six DHS/MIS survey waves, achieving 2.42 pp mean absolute error on Kenya-level ownership. The analysis chain supports mechanism counterfactuals and policy experiments on equity outcomes of ITN distribution strategies.

This repository contains the Python implementation of an agent-based model investigating how localized boundary-crossing dynamics generate large-scale connectivity in structured multi-attractor landscapes.

Agents evolve in a continuous two-dimensional environment composed of attractor basins. A fraction of agents exhibits exploratory higher-mobility dynamics, while the remaining agents remain locally constrained. The model analyzes how localized configurational transitions accumulate into transition networks that progressively integrate the explored state space.

The repository includes:

…

ABMIND, the Agent-Based Model of Individual Psychological Distance, is a modeling framework developed to examine how psychological distance influences environmental protection behavior in coastal farming communities in southern China. Using household survey data and empirically estimated behavioral pathways, the model represents how uncertainty shapes four dimensions of psychological distance, namely temporal, spatial, social and hypothetical distance, and how these dimensions guide protection and degradation decisions. Agents include households, government actors and mangrove ecosystem patches, connected through social networks and ecological feedbacks that affect learning, expectations and perceived benefits. Policy interventions such as rewards, penalties and publicity guidance efforts work by modifying uncertainty and psychological distance rather than directly controlling behavior. ABMIND is implemented as a spatially explicit model following the ODD protocol, and a concise user guide is provided. In developing ABMIND we introduce a structured validation workflow that links statistical mediation analysis with simulation-based diagnostics, allowing empirical cognitive mechanisms to be systematically embedded and tested within the ABM. This integrated approach strengthens the credibility of psychological-mechanism models and supports their use in policy evaluation. The framework offers a methodological platform for integrating cognitive mechanisms into agent-based environmental behavior modeling and for evaluating policy strategies that support ecosystem protection.
Model paper:
ABMIND: An empirically informed agent-based model of psychological distance and environmental protection behaviour
Ecological Modelling
https://doi.org/10.1016/j.ecolmodel.2026.111700

The FishCensus model simulates underwater visual census methods, where a diver estimates the abundance of fish. A separate model is used to shape species behaviours and save them to a file that can be shared and used by the counting model.

This model simulates a simple aquatic ecosystem containing fish and food. It explores how individual interactions such as movement, feeding, and reproduction shape the population dynamics of fish over time.

This model visualizes gradient descent optimization - the fundamental algorithm used to train neural networks and other machine learning models. Agents represent different optimization algorithms searching for the minimum of a loss landscape (the “error surface” that ML models try to minimize during training).

The model demonstrates how different optimizer types (SGD, Momentum with different parameters) behave on various loss landscapes, from simple bowls to the notoriously difficult Rosenbrock “banana valley” function. This helps build intuition about why certain optimization algorithms work better than others for different problem geometries.

HOW IT WORKS

…

MASTOC-LLM extends the classic Multi-Agent System Tragedy of the Commons (MASTOC) model by replacing hard-coded behavioral rules with autonomous decision-making powered by large language models (LLMs). Three heterogeneous agents manage herds of cows on a shared grassland commons. Each tick, an agent receives a structured prompt describing current resource levels, its own herd size, peer behavior, and — optionally — a rolling memory of recent rounds and messages from neighboring agents. The LLM returns a stocking decision (add, remove, or hold cows) together with a natural-language rationale and, when communication is enabled, a short message to broadcast to peers.

The model is designed to test whether LLM agents spontaneously develop Ostrom-style common-pool resource governance (mutual monitoring, graduated sanctions, graduated rule revision) or instead fall into identifiable failure modes. Preliminary experiments with Claude Haiku 4.5, GPT-5.4-mini, and DeepSeek R1:32b have revealed four recurring collapse patterns — Cooperative Paralysis, Defection Cascade, Overshoot-Panic, and Hybrid Architecture Failure — whose onset timing is sensitive to memory length, inter-agent communication, and the post-training alignment approach of the underlying model.

MASTOC-LLM is intended as a laboratory for generative agent-based modelling (GABM) methodology: it provides a clean, well-understood commons baseline against which LLM behavioral hypotheses can be systematically tested and compared across models, parameter sweeps, and alignment regimes.

Exploring learned cooperation, coevolution and free-riding. Learning is achieved through Multi-Agent Deep Reinforcement Learning (MADRL) in an ecological environment. The environment emits no other than sparse reproduction rewards. No reward shaping, no explicit cooperation signal.