This tool helps to create argumentation graphs that represent reward functions for learning moral values in a reinforcement learning algorithm.
It is meant to work in conjunction with the ethical-smartgrid simulator: the Python code this tool produces can be (almost) directly used to create reward functions that work with the Ethical Smartgrid simulator.
The idea of using argumentation graphs that represent reward functions is described in our paper:
Benoît Alcaraz, Olivier Boissier, Rémy Chaput, and Christopher Leturc. 2023. AJAR: An Argumentation-based Judging Agents Framework for Ethical Reinforcement Learning. In Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems (AAMAS '23). International Foundation for Autonomous Agents and Multiagent Systems, Richland, SC, 2427–2429. DOI: 10.5555/3545946.3598956
@inproceedings{10.5555/3545946.3598956,
author = {Alcaraz, Beno\^{\i}t and Boissier, Olivier and Chaput, R\'{e}my and Leturc, Christopher},
title = {AJAR: An Argumentation-Based Judging Agents Framework for Ethical Reinforcement Learning},
year = {2023},
isbn = {9781450394321},
publisher = {International Foundation for Autonomous Agents and Multiagent Systems},
address = {Richland, SC},
booktitle = {Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems},
pages = {2427–2429},
numpages = {3},
location = {London, United Kingdom},
series = {AAMAS '23}
}