Authors:
Aloïs Rautureau, Eric Piette

Venue:
arXiv preprint, 2025

Topics:
human-like AI, general game playing, cognitive psychology, Monte Carlo Tree Search

Links: PDF · arXiv

While artificial intelligence systems have reached or surpassed human performance in many games, their behaviour still differs strongly from the intuitive and pattern-based decision-making processes observed in humans.

This paper presents an overview of findings from cognitive psychology and previous work on human-like artificial agents, and discusses how these ideas can be applied to General Game Playing.

It introduces CogniPlay, a work-in-progress model designed to move toward more human-like game-playing agents, with a particular connection to Monte Carlo Tree Search and cognitively inspired decision processes.

This paper is closely aligned with the objective of developing human-like AI agents rather than systems that merely maximise performance.

In the context of general game playing, this means designing agents that better reflect human cognitive processes, including intuitive pattern recognition and bounded reasoning, instead of relying only on brute-force optimisation.

As a work in progress, CogniPlay helps formalise a research direction that connects artificial intelligence, game-playing, and cognitive science, and fits naturally within a broader agenda on general and human-like agents.

Rautureau, A., Piette, E. (2025). CogniPlay: a work-in-progress Human-like model for General Game Playing. arXiv preprint, arXiv:2507.05868.

@article{rautureau2025cogniplay,
  author  = {Rautureau, Alo{\"i}s and Piette, Eric},
  title   = {CogniPlay: a work-in-progress Human-like model for General Game Playing},
  journal = {arXiv preprint arXiv:2507.05868},
  year    = {2025},
  url     = {https://arxiv.org/abs/2507.05868}
}