HTA-Driven Structuring of Gameplay Loops: Development of an LLMAssisted Analysis Tool for Early-Stage Game Design

Abstract

In the early design stage of game development, it is often challenging to conduct task analysis for new game designs that address diverse player needs, making it difficult to visualize potential design flaws or imbalances in the experience structure. To address these challenges, this study proposes an LLM-driven analysis tool capable of analyzing new game designs that respond to diverse player needs. The tool uses GPT-4o to parse natural language descriptions provided by designers, automatically extracting the game flow and associated mechanics. Based on this, it constructs a task tree following the principles of Hierarchical Task Analysis (HTA) and derives a Gameplay Loop from its branch structure. Furthermore, it applies multiple experience tags—defined by the MDA framework and PXI—to each task node, and visualizes tag distribution and potential structural issues. This enables designers to grasp the overall structure of their game and identify latent problems without requiring a complete prototype.