Authors:
Eric Piette, Matthew Stephenson, Dennis J. N. J. Soemers, Cameron Browne

Venue:
IEEE Conference on Games (CoG), 2021

DOI:
10.1109/CoG52621.2021.9618990

Topics:
general game playing, game concepts, Ludii, transfer learning, explainable AI, game recommendation

Links: PDF · ACM · IEEE · arXiv

Abstract

This paper proposes to formalise the notion of game concept for general board game playing, inspired by terms commonly used by players and designers.

Through the Ludii general game system, concepts are described at several levels of abstraction, including the game itself, the moves played, and the states reached.

The introduction of game concepts opens many new research directions, including hyper-agent selection, transfer learning between games, explainable AI, game reconstruction, and board game recommendation.

Context

This paper is a foundational contribution toward higher-level representations for general game playing.

While ludemes provide structured symbolic descriptions of games, game concepts operate at a more human-understandable level, making them especially relevant for transfer learning and AI explainability.

The work also broadens the scope of Ludii beyond game playing itself, by connecting it to recommendation, reconstruction of historical games, and human-facing interpretations of AI behaviour.

Full reference

Piette, E., Stephenson, M., Soemers, D. J. N. J., Browne, C. (2021). General Board Game Concepts. In IEEE Conference on Games (CoG). DOI: 10.1109/CoG52621.2021.9618990

BibTeX

@inproceedings{piette2021concepts,
  author    = {Piette, Eric and Stephenson, Matthew and Soemers, Dennis J. N. J. and Browne, Cameron},
  title     = {General Board Game Concepts},
  booktitle = {IEEE Conference on Games (CoG)},
  year      = {2021},
  doi       = {10.1109/CoG52621.2021.9618990},
  url       = {https://arxiv.org/pdf/2107.01078}
}