Introduction: Invitation from a Game Master｜Artificial Intelligence based Werewolf
Artificial intelligence (AI), for which game play with humans has long been an important benchmark for research, can show the essence of intelligence and its requirements. AI has already defeated top-level human players in complete information games such as chess or Shogi. The complete information game Go currently has human players who can defeat the most advanced AI, but it seems only a matter of time before humans will be defeated by AI. In the field of incomplete games, Texas Hold’em Poker has a tradition of competition in the AI field. Real-time action video games are also spurring attempts at evaluating intelligence in real-time environments.
Compared to the previous game-related challenges, the communication or communicative intelligence skills that are commonly used in board games and card games have not been examined. When people play board games and card games, they converse with other players, and some games are advanced through conversation as so-called communication games. Few studies in the literature have analyzed artificial intelligence in relation to such communication games. “Are You a Werewolf?” is one of these communication games.
The Werewolf game is conducted solely through discussion and players must exert their cognitive faculties fully in order to win. In the game, players must hide information, in contrast to perfect information games such as chess or Reversi. Each player acquires secret information from other players’ conversations and behavior and acts by hiding information to accomplish their objectives. The game highlights various problems that have not been addressed adequately in the field of artificial intelligence, such as the asymmetric distribution of player information, the use of persuasion for earning confidence, and the use of speculation for detecting fabrications.
Following this, we started a project to create an “Artificial Intelligence Werewolf” (AIWolf) to play as the werewolf in place of a human player. The objective of this comprehensive project is not only to develop a game-playing algorithm, but also to develop virtual agents, real robots, and so on. To achieve this objective, there are a number of tasks that must be undertaken.
The collective intelligence approach is used for these tasks. To implement a collective intelligence approach, a common platform is indispensable. As described herein, we present the outline of a Werewolf platform for artificial intelligence (AI Wolf Platform), which has been developed as an open source project. We plan to hold an AI Wolf tournament in which researchers with various backgrounds can participate freely with the aim of realizing collective intelligence with the participants.
To achieve the above goals, we require research technologies in the following areas:
- Multi-agent simulation
- Machine learning
- Language processing
- Logic programming
- Human–agent interaction
- Cognitive science
Collaboration in these areas will be required for solving Werewolf. We anticipate that the development cycles will contribute to these fields and aid in the finding of essential features on social interaction.