Summary

This thesis studies human-like game-playing agents beyond the usual focus on behavioural alignment, where agents are mainly evaluated by how closely their moves resemble those of human players.

The manuscript argues that this view narrows the definition of play by overlooking social, interactive, and cognitive dimensions. It reframes human-like agents through a broader, multi-dimensional classification, covering behavioural alignment, cognitive plausibility, modelling range, and generalizability.

The work reviews methods for measuring behavioural similarity, explores statistical approaches that account for individual player policies, and investigates cognitively plausible search algorithms inspired by constraints such as memory, selectivity, and bounded rationality.

Reference

Rautureau, A. (2026). Human-Like Game-Playing Agents Beyond Behavioural Alignment. Master internship report, UCLouvain.

BibTeX

@mastersthesis{rautureau2026humanLikeAgents,
  author = {Rautureau, Alois},
  title  = {Human-Like Game-Playing Agents Beyond Behavioural Alignment},
  school = {UCLouvain},
  year   = {2026},
  type   = {Master internship report},
  url    = {https://piette.info/eric/master/2026%20-%20Human-Like%20Game-Playing%20Agents%20Beyond%20Behavioural%20Alignment.pdf}
}