An agent-based model of decentralized dispute resolution oracle.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
court
.gitignore
LICENSE
README.md
requirements.txt
run.py

README.md

Court Simulation 🔬


This is a agent based simulation supporting Aragon's research towards a decentralized oracle for dispute resolution.

The model will instantiate a number of Jurors with some amount of initial tokens.

For each dispute the true result will be determined by the model, and each agent will produce an individual belief. Jurors activate each of their whole tokens for each dispute, and tokens are drawn by the scheduler randomly.

Each Juror will vote honestly with each of their tokens which have been drawn in the current step based on their belief. If the result does not align with global truth then the dispute will be appealed, and the model will select additional tokens until all tokens have participated in the dispute, or the verdict aligns with the truth.

In this model we assume that all agents act honestly and we assume that the agents estimation is normally distributed around the truth. By itself this is not a particularly useful model, but it is intended to provide a baseline representation of the court mechanism, which can be extended to study the dynamics of agents based on various more complex and realistic assumptions.

Getting Started

Python


Make sure you have Python 3

Install Mesa


Install Mesa on Python 3:

$ pip3 install mesa

Dependencies


Install all dependencies either manually or by using

$ pip3 install -r requirements.txt

Run


Download this repository. cd into the main directory for this repository. And run

$ python3 run.py

View


The server should host it on http://127.0.0.1:8521/

In the main screen you will see:

  • A top bar menu, with flow controls to the right.
  • Some parameter sliders:
    • Number of jurors included in the model.
    • Belief threshold: Agents must sample within this many standard deviations from the true value to be coherent.
    • Number of tokens in supply, tokens are split evenly among jurors at initialization.
    • Dispensation percentage: Percentage of activated tokens which are redistributed from incoherent jurors to coherent. See here for a reference.
  • A main chart showing the amount of disputes, split in susccessful and failed.
  • Another chart showing the resulting Gini coefficient for the inequality distribution of tokens after each dispute.