Abstract

The demand for high-quality video games is ever increasing. This entails every aspect of video game creation, such as graphic fidelity, sound design, gameplay innovation, or novel game design approaches. However, the ambition to tell truly interactive and dynamic stories is also a significant factor contributing to this trend. This thesis aims to create a novel game mechanic that supports narrative gameplay. The presented game mechanic is based on exchanging information between the player and NPCs and among NPCs themselves. By modeling the NPCs as intelligent agents in a multi-agent system, techniques from this field are used to give the agents an autonomous behavior as they move through a 3D environment. A classification for different types of gameplay focusing on dynamic gameplay is created, and the developed prototype is compared to other current games in this classification. The resulting system creates a playground for varied narrative scenarios that are not previously defined by a game designer but can be explored by the player and are supported by the game systems.

Project Description

In this thesis, I developed a game mechanic based on modeling information into objects that exist in the game world. They are created by interacting with the world and are perceivable by agents in the system next to the player. By allowing the non-player agents to react to newly learned information, the model creates an interplay among all agents in the system. The resulting game mechanic then is based on retrieving relevant information about characters or other objects in the game world and introducing new information as the player. This mechanic could even be used to create factually wrong information and thus allow the concept of lying to agents. For the prototype game that was also developed as part of this thesis, I chose the following setup to showcase the game mechanic. The player controls a knight in a freely traversable 3D environment that represents a medieval kingdom. There are several key locations like a castle, a village, or a farm that are inhabited by other non-player characters (NPC) that represent the agents of the multi-agent system. The NPCs move around the kingdom and execute different tasks, which bring them into contact with each other. When they interact, they will also exchange information and thus learn new things about the game state. The player can also interact with characters and items and learn pieces of information. They can also engage in a conversation with NPCs and ask for or give out specific information. By adding quests to the game, a narrative framing exists that the player can use to motivate their actions. Elevating the exchange of information in multi-agent systems to a core game mechanic and allowing players to exploit the information flow is a novel system that not many games support today. It is also a step towards creating an actual interactive story because it allows for genuinely influencing the game world but still enabling developers to define rules for how agents and the environment reacts and thus enforcing a narrative structure. Implementation The game simulates a small medieval kingdom and its inhabitants. The player controls a knight who is sent on a quest by the queen. The king is missing, and the player shall return the king to the castle. The player can move around and interact with non-player characters (NPCs) and some key items. First, I introduce a Information Model that I created to define information in the game. Additionally, I provide concrete implementations of both the information heuristic and the consensus protocol.

Results

I have analyzed several games through the lens of classification criteria. The selected games all feature mechanics that fall under the umbrella term of dynamic storytelling but apply the creation of narrative content differently. The final result of the classification shows that the games all behave in different ways when replayed multiple times. This is not yet true for the prototype game since the game does not yet properly react to the player. The generated narrative is purely emergent. All the considered games except for Road 96 and the prototype game actually react to player actions. Games like Road 96 follow a more "choose your own adventure" approach where all options are known upfront and are then mixed and matched to create new variations. Three of the analyzed titles, Road 96, Crusader Kings III, and the Game Master Mode in Divinity: Original Sin 2, have capabilities to generate narrative content that follows a dramatic structure. They do so differently, though by reordering encounters in a more dramatic fashion, using the game AI to let characters behave in specific ways, or by introducing a human narrator to steer the storyline. It is worth noting that all games except for Crusader Kings III contain some form of existing storyline or narrative frame in which the game takes place. Only Crusader Kings III lets its gameplay systems alone in combination with the player actions dictate the unfolding, fully emergent story.

Conclusion

This thesis presents a novel game mechanic that supports dynamic storytelling by allowing NPCs to exchange game state information and allowing the player to retrieve and introduce pieces of information to the system. The developed prototype game applies techniques from multiple disciplines of computer science in order to implement the designed game mechanic. Concepts from Emergence Games, Communication Theory, Multi-Agent Systems, State Machines, Consensus Protocols, MapReduce, and Inference Engines are brought together to create systems that interact in a deep and complex way to allow the NPCs to make decisions and act on them. This approach shows potential that encourages the further development of this game mechanic. The developed prototype provides exciting and varied narrative scenarios depending on the actions the player executes in the game. There are, however, several shortcomings that hinder the prototype from being more immersive in its narrative implications and more varied in its gameplay options. I suggest areas where further work is most promising and where the basic groundwork is already present in the prototype to be expanded.

[PDF]

[ Slides Final ]