This model is a part of an ongoing research project on Multiagent Reinforcement Learning (MARL). The ODD protocol is included in the model. In this version of the model, Proximal Policy Optimization (PPO) is designed in the agent behaviors. It also includes a designed experiment in its Behavior Space which is used in the Response Surface Methodology and training of an Artificial Neural Network (ANN) based Recommender System.

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 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

…

A dynamic identity model for misinformation in social networks, an agent-based model of social identity and misinformation dynamics.

I developed this model as a part of my master’s thesis, “Does social identity drive belief and persistence in online misinformation? An agent-based modelling approach” at University College Dublin, Ireland (2024-2025).

The purpose of this model is to further understand the dynamics of misinformation sharing as an expression of social identity. I introduce a framework to understand the influence of self-categorisation on misinformation persistence in social network. It integrates a social learning model with the Dynamic Identity Model for Agents (DIMA) using simple logic to simulate the social trade-offs driving misinformation and observe the effects on misinformation spread.

This model is a minimal agent-based model (ABM) of green consumption and market tipping dynamics in a stylised two-firm economy. It is designed as an existence proof to illustrate how weak individual preferences, when combined with habit formation, social influence, and firm price adaptation, can generate non-linear transitions (tipping points) in market outcomes.

The economy consists of:
1) Two firms, each supplying a differentiated consumption bundle that differs in its fixed green share (one relatively greener, one less green).
2) Many households, each consuming a unit mass per period and allocating consumption between the two firms.

…

This agent-based model, developed for the study “Online Protest and Repression in Authoritarian Settings,” examines how online protest and repression evolve in authoritarian contexts and how these dynamics affect ordinary users’ attitudes and behavior on social media. The model integrates key theoretical and empirical insights into social media use and core political factors that shape digital contention in authoritarian settings. The following questions are addressed: (1) how online protest–repression dynamics unfold across different levels of authoritarianism and varying compositions of committed accounts, and (2) how ordinary users’ internal propensity to protest and their perceived probability of successful repression change during online protest-repression contestation. The model is evaluated against two empirically grounded macro patterns observed in the real world. The first is enduring protest: online protest becomes dominant as vocal protesters grow to outnumber vocal repressors, shrinking the pool of silent users and stabilizing a pro-protest majority. The second is suppressed protest: online dissent is contained as vocal repression and silence expand in response to protest, yielding a sustained majority of repressive and silent accounts. Together, these dynamics demonstrate how dissenting voices are empowered and suppressed online in authoritarian settings.

ARMM is a theoretical agent-based model that formalizes Murra’s Theory of Verticality (Murra, 1972) to explore how multi-zonal resource management systems emerge in mountain landscapes. The model identifies the social, political, and economic mechanisms that enable vertical complementarity across ecological gradients.
Built in NetLogo, ARMM employs an abstract 111×111 grid divided into four Andean ecological zones (Altiplano, Highland, Lowland, Coast), each containing up to 18 resource types distributed according to ecological suitability. To test general theoretical principles rather than replicate specific geography, resource locations are randomized at each model initialization.
Settlement agents pursue one of two economic strategies: diversification (seeking resource variety, maximum 2 units per type) or accumulation (maximising total quantity, maximum 30 units). Agents move between adjacent zones through hierarchical decision-making, first attempting peaceful interactions—coexistence (governed by tolerance) and trading (governed by cooperation)—before resorting to conflict (theft or takeover, governed by belligerence).
The model demonstrates that vertical complementarity can emerge through fundamentally different mechanisms: either through autonomous mobility under political decentralization or through state-coordinated redistribution under centralization. Sensitivity analysis reveals that belligerence and economic strategy explain approximately 25% of outcome variance, confirming that structural inequalities between zones result from political-economic organization rather than environmental constraints alone.
As a preliminary theoretical model, ARMM intentionally maintains simplicity to isolate core mechanisms and generate testable hypotheses. This foundational framework will guide future empirically-calibrated versions that incorporate specific archaeological settlement data and geographic features from the Carangas region (Bolivia-Chile border), enabling direct comparison between theoretical predictions and observed historical patterns.

Boyds (Boids that Fight) is an agent-based model in NetLogo that extends the classic Flocking model with multi-faction competition, a local fight–flight heuristic, and a target locking/“taking” mechanism. The model separates perception (vision) from engagement range (lock distance) and uses per-faction steering bounds to explore how local numerical superiority, sensing, and bounded turning affect victory, losses, and emergent formations.

The model examines cattle herd dynamics on a patchy grassland subject to two exogenous pressures: periodic raiding events that remove animals and scheduled management culling that can target males and/or females. It is intended for comparative experiments on how raiding frequency, culling schedules, vegetation dynamics, and life-history parameters interact to shape herd persistence. The model was specifically designed to test the scenario of cattle herding in the arid grasslands of southern Arizona and northern Sonora during the mission period (late 17th through late 18th centuries, CE). In this period, herds were locally managed by Spanish mission personnel and local O’odham groups. Herds were culled mostly for local consumption of meat, hides, and tallow, but the mission herds were often targets for raiding by neighboring groups. The main purpose of the model is to examine herd dynamics in a seasonally variable, arid environment where herds are subject to both intentional internal harvest (culling) and external harvest (raiding).

HUMLAND Fire-in-the-Hole is a conceptual agent-based model (ABM) designed to explore the ecological and behavioral consequences of fire-driven hunting strategies employed by hunter-gatherers, specifically Neanderthals, during the Last Interglacial period around the Neumark-Nord (Germany) archaeological site.

This model builds on and specializes the HUMLAND 1.0.0 model (Nikulina et al. 2024), integrating anthropogenic fires, elephant group behavior, and landscape response to simulate interactions between humans, megafauna, and vegetation over time.