Year: 2025
Type: dissertation
Study design: Systematic Review
Abstract: Interest in hidden role games has greatly increased. The class of hidden role games consists of games where players are divided into two different sized teams. Both teams try to eliminate the other. The minority team, known as the MIN team, know the roles of all other players, while the majority team, known as the MAX team, knows only their own role. Because of their unique reliance on communication and social comprehension, hidden role games are positioned to be a useful class of games in the field of multi-agent simulation. The simulation of this class of games constitutes a nascent and understudied field. Consequently, until now, it has lacked a systematic review. To address this, we conduct a systematic literature review that examines the field from four perspectives to provide an overview of the current research, as well as to highlight unexplored areas in the field of multi-agent simulations of hidden role games. The corpus consists of 42 papers, the majority of which (28) were published between 2023 and 2025 inclusive. In addition, 27 papers are preprints. The rapid growth in 2023 to 2025 highlights the increase in interest in the field. The review shows a clear clear trends in the literature: Firstly, an increasing trend in interest due to recent developments in large language model (LLM) technology. Secondly, clear trends in terms of focus, communication modeling, and agent modeling. Thirdly, a narrowing trend in types of agent models and models of communication explored. In response, the foci of studies have widened to examining agent behavior and using SDGs to create LLM benchmarks. The state of the field shows promise, with ample opportunity to study novel ideas and phenomena. It is deeply intertwined with advances in machine learning which widen the possible ways of modeling agents and communication.
Source: Aaltodoc (Aalto University)
Author: Longhurst, Oliver
Institution: Aalto University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6996711636