Introduction to agent-based modeling and simulation in token economics (part II)

Journey to the “perfect” token model

Philipp Richter
Aug 24, 2018 · 4 min read

This is the second part of the previous article. For the first article click here.

As mentioned before ABMs can play an important role for creating blockchain environments. Certain scenarios and token systems can be simulated and with the data from ABMs Smart Contracts can become more efficient and maybe more secure. This in mind we as started researching about how agent-based modeling could be applied to token systems and token economics. Read know the second part of the basis of agent-based-modeling.

Learning agents

As already mentioned, the agents of a token economy are learning, adaptive agents with a memory. This form of agents can learn in two different ways to adapt their behavior. On the one hand, learning can take place on an individual level. Agents look back into the past to track their errors, update event probabilities and are susceptible to positive amplification. On the other hand, agents can learn as a population. This can be modelled on Darwin’s theory of evolution, imitation or ,as described in the previous section, through social influences.

The genetic algorithm is a technique inspired by the evolutionary effectiveness of mutations and differentiated reproduction. With it it is possible to simulate the development of agents with limited rationality over time. At this point, however, it remains to be clarified to what extent the use of genetic algorithms comes close to the actual behavior of the agents of a token economy. The behavior of other agents is often only copied without understanding the underlying rules. Genuine evolution and selection is therefore rather rare. Nevertheless, the use of genetic algorithms, at least in part, for the modelling of learning agents should be considered.

Simple rules and complex behavior

Even a small economy, like a token economy, with a few, well-defined agents could display complex behaviors and deliver unexpected simulation results. To gain a basic understanding of the mechanics and processes at the micro level and the results at the macro level, a step-by-step construction of the token economy is necessary. Otherwise there is a danger that the results of the simulation cannot be interpreted.

Construction of an agent-based model

With the knowledge gained, we have defined a schedule and structure for the development of an ABM. This flowchart will serve as a point of reference for the conception of the token economy in the following.

Technical implementation

Agent-based modeling and simulation is basically possible in two scales, Desktop ABMS and Large-Scale ABMS. Desktop ABMS are particularly suitable as an entry point for creating initial, simple models with a limited number of agents. They are relatively simple in design and easy to create. Furthermore, they enable the visualization of the model, statistical analyses and database routines.

Large-scale AMBS exceed the capabilities of desktop ABMS many times over and enable the modeling of millions of agents. However, this makes them more difficult to learn, they also require much more computing power and are rather unsuitable for beginners, small test models and the early development process.

SWOT analysis for agent-based modeling

ABM offers almost infinite possibilities to realistically model facts and areas of the real world and to map highly complex systems. The downside of this boundlessness, however, comes in the form of a flood of data and events. The following SWOT analysis results:


First a first simple (simply constructed, conditionally realistic) model of a token economy is created which serves as a kind of basis for the further conception. The agents maintain their original behavior/strategy for the entire duration of the simulation and a clear time horizon is defined. The simulation is now repeated to analyze outputs and obtain consistent results. Based on this, certain parameters to be defined are manipulated. The first basic behaviours should develop, which should be interpreted. If the macroscopic laws of this basic model can be clearly determined, this model is extended step by step. In this way it can be ensured that legalities can be clearly determined in each sub-step as well as in the final result.

This part and the previous are giving an introduction to ABMs. They represent the information we gained from our research for the use of ABMs for simulating token economics.

Feel free to comment on our idea and approaches on this and on upcoming articles.

We from have been looking for a token model which suits our use case — providing unique identities for goods and documents — for quite a while now but we did not find any. So we want to develop it by ourselves.

You are invited to follow us on our journey to develop a model to simulate token ecosystems. To figure out what the different motivations of the stakeholders are, how they can be incentivized to create a healthy token flow inside a system and point out where the bottlenecks and limits of the system are.

PS: you will find our journey also on

Token Simulation Model

The way to the perfect token economy