How can you play a computer game and simultaneously communicate freely with game characters without limitations? Current traditional story-based games often restrict player communication with characters to predefined sets of actions and dialogue options, limiting broader interaction. Ing. Jiří Macháček, a student at FIT CTU, proposed an innovative solution in his thesis with a game called HaLLMark. He focused on utilizing large language models that serve to tell game stories through non-player characters (NPCs). The goal was to create a playable prototype that would allow free interaction with NPCs and enable players to tell their own unique game stories. This work earned him the Dean's Award for the summer semester of the 2023/2024 academic year.
After analyzing typical computer and board story-based games, Ing. Macháček implemented the best language models in the HaLLMark prototype. Large language models are advanced AI systems capable of processing, understanding, and generating human-like text. They are built on deep learning techniques and trained on vast amounts of text data, such as books, articles, and websites. The HaLLMark game prototype uses an appropriate story setting and algorithms for story creation and character development. These models facilitate free interactions with NPCs, allowing them to generate immersive and coherent stories in real-time.
HaLLMark is aimed at players who enjoy creating their own stories and strategy enthusiasts. It allows players to create their own character— a king who rules a fantasy kingdom and addresses the problems of his subjects. Players can communicate freely with in-game characters through text, and their decisions impact the rules of the game world and kingdom metrics (wealth and strength of the kingdom, king's popularity, and the court's patience). However, the player must adhere to their royal role. If the player makes nonsensical decisions or the conversation strays off-topic, the patience metric of the king's court significantly drops. If any metric reaches zero, the player loses the game. Therefore, it is in the player's interest to write meaningfully and reasonably. Additionally, the game characters are instructed not to stray from the topic and remain in their roles as subjects in the fantasy world. If the player digresses, the characters will try to steer the story back to the problems the player needs to address.
“Players were satisfied with the prototype and the generated story and often praised the concept of the game. Based on research into story creation, plot twists were added to the game, leading to many funny moments during testing and significantly improving the game. The testing results show previously unseen levels of consistency and interest in the generated story, achieved through the use of vector databases and prompt engineering techniques,” says Jiří.
The results demonstrate that language models can effectively create consistent stories, opening new possibilities in game development. In the future, new models and advanced features could be added to the prototype, enhancing the player experience. The game could also be translated into other languages and new game mechanics introduced, allowing players to actively engage, not just respond to matters presented by characters.
Jiří has already presented his work and the HaLLMark game at two conferences— at the Meaningful Play 2024 conference in Pittsburgh in October and the AI Hotspot conference in November, where he won a young talent award for it.
