A BDI Game Master Agent for Computer Role-Playing Games. (Extended Abstract) Bao Vo Luong, John Thangarajah, Fabio Zambetta and Mahmud Hasan School of Computer Science and IT, RMIT University, Melbourne, Australia {bao.luong,john.thangarajah,fabio.zambetta,mahmud.hasan}@rmit.edu.au ABSTRACT In this paper we describe an approach for developing an in- telligent game master (GM) for computer role-playing games. The role of the GM is to set up the game environment, man- age the narrative flow and enforce the game rules whilst keeping the players engaged. Our approach is to use the popular Belief-Desire-Intention (BDI) model of agents to de- veloping a GM. Categories and Subject Descriptors D.2.10 [Software Engineering]: Design—methodologies Keywords BDI Agent Systems, Computer Role-Playing Games, Game Mastering. 1. INTRODUCTION AND MOTIVATION In recent years, computer games and interactive story- telling research have been emphasising the need for a ra- tional ‘coordination agent’ that orchestrates the behaviour of several autonomous characters according to some high- level, often plot-oriented, policy without taking direct con- trol of all their actions. Examples of such a trend have been narrative controllers or director agents in interactive story- telling [2], AI directors in computer games [1] and, directly inspired from real-life, game masters in a table-top or live- action Role-Playing Game (RPG) [3]. A Game Master (GM) in table-top RPGs is a person in charge of organising engaging game sessions for cooperative multi-player experiences, usually in the range of 4-8 partic- ipants. The GM describes the events taking place in the game’s fictional world, she gives life to the characters pop- ulating it, and then finally communicates the outcomes of players’ decisions, enforcing any game rules as needed. A table-top RPG game session is often regarded as an emergent theatrical production where players take the role of lead actors and the Game Master serves as the director that provides stage, scenery and a flexible plot that adapts as a result of the interaction between players and the fictional game world. Examples of interaction between Game Mas- ters and players abound in specialised press, game books, Appears in: Proceedings of the 12th International Confer- ence on Autonomous Agents and Multiagent Systems (AA- MAS 2013), Ito, Jonker, Gini, and Shehory (eds.), May, 6–10, 2013, Saint Paul, Minnesota, USA. Copyright c  2013, International Foundation for Autonomous Agents and Multiagent Systems (www.ifaamas.org). All rights reserved. academic publications as well as dedicated web sites 1 and on-line gaming communities. Our work contributes a solution to the design and imple- mentation of an automated GM agent for computer role- playing games. The GM agent will set up the game en- vironment, manage the narrative flow in a flexible fashion and it will enforce the game rules as needed while keeping the players engaged. Such a GM agent will help address is- sues in current computer role-playing games, which resort to scripting every possible event in the game’s narrative. Not only does this lead to a substantial increase in design and programming time for the game [5], but it also reduces the narrative flexibility, which will have a negative impact on players’ engagement. Our design uses the Belief-Desire-Intention (BDI) model of agents to flexibly pursue the goals of the plot. BDI agents are designed and implemented using mental attitudes such as beliefs, goals, plans and intentions [4]. The use of such mental attitudes in software design allows developers to break complex decisions into smaller, more intuitive compo- nents. This technology is most useful when building systems that operate in highly dynamic environments as it allows for real-time online reasoning. The design of such systems al- low incorporating alternatives strategies for achieving the one task, thus enabling flexible and robust systems. These traits, make BDI agent technology ideal for developing an automated software GM. 2. DESIGNING A BDI GM AGENT There are 3 main steps in designing a BDI based GM agent: (i) Constructing a ‘game-flow’ diagram out of a sam- ple game scenario; (ii) Constructing a ‘goal tree’ based on the game-flow diagram; and (iii) Constructing the agent’s detailed ‘goal-plan tree’ by adding plans, to achieve the iden- tified goals, and additional sub-goals as appropriate. (i)Constructing the game-flow diagram: The first step, is to construct a game-flow diagram for the scenario. We be- gin with an abstract diagram as shown in Figure 1, where in order to Get inside the Hideout the player has to go through two main stages: Gain Access To Hideout Door and Player Opens Hideout Door. The game-flow diagrams for each of them were subsequently developed and they are shown, re- spectively, in Figure 2 and Figure 3. (ii)Constructing the goal tree: In BDI agent design method- ologies an initial task in the requirement analysis phase is to 1 For instance, see http://www.darkshire.net/~jhkim/ rpg/whatis/tabletop.html. 1187