Proceedings of the Annual Conference of JSAI
Online ISSN : 2758-7347
37th (2023)
Session ID : 4D2-GS-11-01
Conference information

The Effects of Following Another Player’s Attack Target in AIWolf
*Nozomi NAKAOTomohiro MURAKAMIHikaru MIYAKEHayate MURATAItsuki AMANOTaisei KOJIMAKeisuke ANDONobuhiro ITO
Author information
CONFERENCE PROCEEDINGS FREE ACCESS

Details
Abstract

Recently, the research of AI, the “AIWolf,” to play the werewolf game has attracted keen interest. The werewolf games adopt secret“whisper” conversations among only werewolves to realize cooperative behavior. This study focuses on this “whisper” because there are few studies of “whisper” in the AIWolf research, even though it is an essential factor in the werewolf games. It aims to clarify the effects of following another player’s attack target, one of the strategies using whispers. We investigated the difference between the case in which our agents follow an attack target by using“whisper” and the case in which they did not against the finalists of the AIWolf competition. As a result, it showed that werewolves improved their win rate by about 2.3 percentage points, on average, especially when following the agent with the highest win rate, and that was a significant improvement in the win rate. Therefore, we clarified that attuning to the agent with the high win rate increases the werewolf’s win rate.

Content from these authors
© 2023 The Japanese Society for Artificial Intelligence
Previous article Next article
feedback
Top