Designer Intentionality vs. Procedural Generation - Comparing Player Experience and Behavior in Maze Level Design

Abstract

This thesis investigates how procedural content generation affects player experience in a 2D maze-based game. A playable prototype, Light the Way, was developed using the Godot engine, in which the player navigates dark hedge mazes using limited torchlight, with the ability to manipulate the level layout by burning through walls, at the cost of reducing their visibility. Maze layouts, item placements, and start and exit positions are all produced algorithmically and integrated with gameplay mechanics designed to make level structure central to the player experience. The study compares procedurally generated and handcrafted levels to investigate how level structure affects player experience. To ensure the comparison is fair, the handcrafted levels were built on similar procedurally generated layouts and used the same visual style, tile system, mechanics, and gameplay rules. This allowed the study to focus only on differences in maze structure and design intent. The findings indicate that both level types were playable and supported the core game loop. Handcrafted levels took longer to complete, required more player movement, and resulted in greater interaction with the wall destruction mechanic—outcomes that aligned with the deliberate design intent behind each handcrafted maze, where surgical modifications were made specifically to increase navigational complexity and encourage resource decision-making. Questionnaire responses showed that handcrafted levels were perceived as more structured, cohesive, and carefully constructed, while procedurally generated levels were associated with unpredictability or simpler navigation. At the same time, procedurally generated levels were still considered clear and playable. The study concludes that procedural content generation can be an effective tool in game production, particularly for providing playable level diversity and reducing manual design work. However, the findings also indicate that procedural generation requires careful design constraints and parameter tuning to produce the same sense of pacing, structure, and intentionality as handcrafted levels.