In 1985 Dr Michael Palmiter, a high school teacher, first built a very innovative agent-based model called “Simulated Evolution” which he used for teaching the dynamics of evolution. In his model, students can see the visual effects of evolution as it proceeds right in front of their eyes. Using his schema, small linear changes in the agent’s genotype have an exponential effect on the agent’s phenotype. Natural selection therefore happens quickly and effectively. I have used his approach to managing the evolution of competing agents in a variety of models that I have used to study the fundamental dynamics of sustainable economic systems. For example, here is a brief list of some of my models that use “Palmiter Genes”:
- ModEco - Palmiter genes are used to encode negotiation strategies for setting prices;
- PSoup - Palmiter genes are used to control both motion and metabolic evolution;
- TpLab - Palmiter genes are used to study the evolution of belief systems;
- EffLab - Palmiter genes are used to study Jevon’s Paradox, EROI and other things.

…

EiLab - Model I - is a capital exchange model. That is a type of economic model used to study the dynamics of modern money which, strangely, is very similar to the dynamics of energetic systems. It is a variation on the BDY models first described in the paper by Dragulescu and Yakovenko, published in 2000, entitled “Statistical Mechanics of Money”. This model demonstrates the ability of capital exchange models to produce a distribution of wealth that does not have a preponderance of poor agents and a small number of exceedingly wealthy agents.

This is a re-implementation of a model first built in the C++ application called Entropic Index Laboratory, or EiLab. The first eight models in that application were labeled A through H, and are the BDY models. The BDY models all have a single constraint - a limit on how poor agents can be. That is to say that the wealth distribution is bounded on the left. This ninth model is a variation on the BDY models that has an added constraint that limits how wealthy an agent can be? It is bounded on both the left and right.

EiLab demonstrates the inevitable role of entropy in such capital exchange models, and can be used to examine the connections between changing entropy and changes in wealth distributions at a very minute level.

…

EffLab was built to support the study of the efficiency of agents in an evolving complex adaptive system. In particular:
- There is a definition of efficiency used in ecology, and an analogous definition widely used in business. In ecological studies it is called EROEI (energy returned on energy invested), or, more briefly, EROI (pronounced E-Roy). In business it is called ROI (dollars returned on dollars invested).
- In addition, there is the more well-known definition of efficiency first described by Sadi Carnot, and widely used by engineers. It is usually represented by the Greek letter ‘h’ (pronounced as ETA). These two measures of efficiency bear a peculiar relationship to each other: EROI = 1 / ( 1 - ETA )

In EffLab, blind seekers wander through a forest looking for energy-rich food. In this multi-generational world, they live and reproduce, or die, depending on whether they can find food more effectively than their contemporaries. Data is collected to measure their efficiency as they evolve more effective search patterns.

…

This model examines how financial and social top-down interventions interplay with the internal self-organizing dynamics of a fishing community. The aim is to transform from hierarchical fishbuyer-fisher relationship into fishing cooperatives.

Designed to capture the evolutionary forces of global society.

The Modern Wage Dynamics Model is a generative model of coupled economic production and allocation systems. Each simulation describes a series of interactions between a single aggregate firm and a set of households through both labour and goods markets. The firm produces a representative consumption good using labour provided by the households, who in turn purchase these goods as desired using wages earned, thus the coupling.

Each model iteration the firm decides wage, price and labour hours requested. Given price and wage, households decide hours worked based on their utility function for leisure and consumption. A labour market construct chooses the minimum of hours required and aggregate hours supplied. The firm then uses these inputs to produce goods. Given the hours actually worked, the households decide actual consumption and a market chooses the minimum of goods supplied and aggregate demand. The firm uses information gained through observing market transactions about consumption demand to refine their conceptions of the population’s demand.

The purpose of this model is to explore the general behaviour of an economy with coupled production and allocation systems, as well as to explore the effects of various policies on wage and production, such as minimum wage, tax credits, unemployment benefits, and universal income.

…

This model is an implementation of a predator-prey simulation using NetLogo programming language. It simulates the interaction between fish, lionfish, and zooplankton. Fish and lionfish are both represented as turtles, and they have their own energy level. In this simulation, lionfish eat fish, and fish eat zooplankton. Zooplankton are represented as green patches on the NetLogo world. Lionfish and fish can reproduce and gain energy by eating other turtles or zooplankton.

This model was created to help undergraduate students understand how simulation models might be helpful in addressing complex environmental problems. In this case, students were asked to use this model to make predictions about how the introduction of lionfish (considered an invasive species in some places) might alter the ecosystem.

a computer-based role-playing game simulating the interactions between farming activities, livestock herding and wildlife in a virtual landscape reproducing local socioecological dynamics at the periphery of Hwange National Park (Zimbabwe).

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

…

We model the relationship between natural resource user´s individual time preferences and their use of destructive extraction method in the context of small-scale fisheries.