Inferring Lexicographically-Ordered Rewards from Preferences

Code author: Alihan Hüyük ([email protected])

This repository contains the source code necessary to replicate the main experimental results in the AAAI 2022 paper “Inferring Lexicographically-Ordered Reward from Preferences.” Our proposed method, LORI, is implemented in files src/ and src/ for the problem settings considered in the paper: cancer treatment and organ transplantation respectively.


First, install the required python packages by running:

    python -m pip install -r requirements.txt

Then, the experiments in the paper can be replicated by running:

    ./src/        # generates the results in Tables 2 and 3
    ./src/  # generates the reward functions in (10) and (11)

Note that, in order to run the experiments for the transplantation setting, you need to get access to the dataset for liver transplantations as of December 4, 2020.


If you use this software please cite as follows:

  author={Alihan H\"uy\"uk and William R. Zame and Mihaela van der Schaar},
  title={Inferring lexicographically-ordered rewards from preferences},
  booktitle={Proceedings of the 36th AAAI Conference on Artificial Intelligence},


View Github