Episodic memory model for computer game characters: to make believable ai characters

Abstract

Interactive narrative is a digital entertainment which has story created or influenced by users. Emergent narrative system are Interactive narrative constructed from believable AI characters. Wan Chin Ho suggested that Episodic memory will give long-term believability to the agent, but it is not proved in his article. To test this, an episodic memory model for game character and a simple game are developed. Dating simulation is a genre of game. The goal of this game is getting close to the game characters. This type of games are selected for this research because it has simple world, and interactions between user and characters are very important in these games. An episodic memory model for dating simulation is suggested. It has two type of re-trieval algorithms: association and recollection. Association is an algorithm for retrieve a memory which is close to the search key. Recollection is an algorithm for find specific memory with a search key.Subject study is doen with 30 subjects. Each subject played a game with a character having episodic memory, and then they played a game with a character not having episodic memory. Subjects answered to a questionnaire. Effects on believability, social interaction, and game experiences are confirmed.