Introduction to agent-based modeling and simulation in token economics

Journey to the “perfect” token model

Agent-based models represent certain systems and environments, e.g. biological organisms, social groups, institutions or physical units. These models specify rules that impose a certain behavior on the parties (agents) inside, based on numerous inputs. All agents individually assess their situation in the modeled environment and make decisions based on the assigned rules of conduct. Agents can act freely in the created world, they pack their knowledge about their environment into rules and act according to them. Depending on the requirements of the model, agents can take a variety of forms.

Functionality of agent-based models

A system (environment) is defined in which the agents are located, as well as the properties of the agents. During a simulation, the agents interact with the environment, influence it and other agents and are influenced themselves. All this happens in discrete time steps. This results in a dynamic micro behavior which ultimately leads to a macro behavior of the system. ABM builds a bridge between the two levels. Each level has new, evolving laws, so macro behavior is no longer just an aggregate. The aim is to develop an understanding of the fundamental processes that occur in a system and derive laws.

Special features of ABMs of token systems

With regard to the modelling of a token economy, further circumstances arise which must be taken into account. Tokens and their economies are an elementary component of blockchain technology, which in turn means that smart contracts and incentive mechanisms of the economy or blockchain must also be included in the model. Smart-Contracts can be simply considered as a set of instructions (program code). Their execution is the sole responsibility of the system and none of the agents has the ability to make changes to them.

These smart contracts impose certain rules of conduct on the model and sometimes also coordinate the interactions of the agents. In addition, there are incentive systems designed to trigger the desired behavior on the part of users. The design of such systems must also be included in the modelling of a token economy, since they potentially influence the behavior of the agents. This influence does not necessarily have to bring about the desired behavior. Incorrectly or incompletely designed mechanisms can work against the system. Finally, the token in question must also be included in the simulation itself. As mentioned above, there is a wide range of tokens with different properties and functions. Stakeholders, their motivations and behavior vary greatly depending on the nature of the token.

What are agents?

In the case of the token economy to be simulated, the agents are stakeholders, i.e. persons who have a legitimate interest in the course and performance of a token economy. These can be investors or speculators, users or the company itself for example.

Properties of an agent

The characteristics that an agent must fulfill for the model of a token economy to be constructed and the subsequent simulation are described below. The determination of these properties is crucial for the success of the simulation, since the agents and their behavior form the basis of the model. The definition of a uniform framework for all agents lays the foundation for adequate modelling. For the further process an agent is characterized as follows:

- Identifiability: All agents or classes of agents must be uniquely identifiable. An individual set of rules and characteristics exists, through which the considered unit of other units in the model is adjacent.

- Environment: The agent is part of the model/simulation and interacts with it. Protocols are needed to control the behavior with its environment and enable the recognition and differentiation of properties of other agents.

- Targeted: Agents pursue a defined goal. This goal must be clearly definable, in the form of a goal function, and measurable.

- Autonomous: The agent is self-directed, it interacts with its environment without exogenous control exclusively on the basis of its set of rules.

- Flexible: Agents have memories, they learn through experience and can adapt the behavior of other agents. To this end, they need rules according to which they can adapt existing rules of conduct.

- Dynamics: All agents are dynamic, which means they constantly interact with their environment, other agents and process information.

Interaction and information flow of agents

The collective behavior and the flow of information between stakeholders inside an token economic environment can be easily illustrated by the standing ovation problem. The scene in this scenario is a lecture hall of a university at the end of an extraordinary lecture. The listeners begin to clap, this builds up and some rise. Now the question arises: will there be standing ovations or will enthusiasm collapse? Whether a listener rises or not depends on many factors. First of all, this decision depends on the subjective perception of the agent. On the other hand, the behavior of his environment can also influence the agent’s decision. How agents are influenced by their environment depends, among other things, on how they receive and process information about them. Even these relatively simple question can have enormous effects on the agents subsequent behavior.

These steps show that agents inside an ABM should be able to make decisions dependable from different circumstances.

This topic is described in more detail in the following articles.