1 of 13

Engineering multi-agent cryptosystems

Shruti Appiah

ConsenSys

2 of 13

Methodology

The token engineering design process

Formulate problem

Design

Model & Simulate

Iterate

Validate

3 of 13

Case study

Incentive mechanisms in

Decentralized Autonomous Organizations

(DAO)

4 of 13

Objective πŸš€

Maximize f(x) - Total # of agents participating in projects in each iteration of the system

Domain 🌐

A single DAO with 10x10 participants

Constraints βš“οΈ

System constraints

Technical constraints

5 of 13

System agents/players πŸ‘«

Members of DAO

System clock πŸ•‘

Period - unit in which a full project cycle can be completed

6 of 13

Assumptions πŸ‘Œ

  • Quality of all the projects completed are similar
  • Agents aren't able to evaluate difficulty of projects
  • All token transactions are tracked

Input parameters ➑️

Altruism coefficients

Starting mechanism πŸ”°

Boltzmann Wealth Model

7 of 13

Modelling

8 of 13

Model properties

  • Spatially static, time-varying model
  • Probabilistic
  • Modelled as discrete-time Markov chain

Tech: Python’s Mesa library

9 of 13

Game

  • Organization’s members participate in company-wide projects to earn rewards.
  • Rewards can be traded, donated, and taxed.

Initial conditions

Randomly initialized, members start with an unequal distribution of coins

10 of 13

Functions

daily_transactions(self, expenditure)

project_reward(self, #rows, #columns)

donate_money(self)

pay_tax(self)

step(self)

11 of 13

Simulation

12 of 13

Limitations

  • Model cannot represent human relationships
  • Agents in the model are not globally aware, focus only on immediate neighbours
  • Informal power structures are not considered
  • Intelligence and lack thereof (bounded rationality) of agents not considered

13 of 13

Thank you

Questions?