TLDR: This research paper introduces a new framework of nested concepts—Landmarks, Monuments, and Beacons—to improve the algorithmic evaluation of procedurally generated game content. It addresses the “Oatmeal Problem” where generated content often feels too similar, by focusing on player-centric perception, evocativeness, and intrinsic “Calls to Action.” The framework aims to bridge technical game AI and humanities research, enabling more accurate and human-aligned assessment of game content’s impact on player experience, with discussions on practical implementation and future research directions.
In the dynamic world of video game development, especially with the rise of procedurally generated content (PCG), creating experiences that truly resonate with players remains a significant challenge. While algorithms can churn out vast amounts of game content, evaluating whether this content aligns with human perception and enjoyment has been a persistent hurdle. Many existing metrics fall short, failing to capture the nuanced “human experience” of a game.
A recent research paper, “Landmarks, Monuments, and Beacons: Understanding Generative Calls to Action”, introduces a novel framework designed to bridge this gap. Authored by Victoire Hervé, Henrik Warpefelt, and Christoph Salge, the paper proposes a set of nested concepts – Landmarks, Monuments, and Beacons – to help developers and researchers better understand and evaluate complex game environments from a player-centric perspective.
The “Oatmeal Problem” and the Need for New Metrics
The core issue in PCG evaluation is often dubbed the “Oatmeal Problem”: procedurally generated content, despite its complexity, can often feel too similar or generic to players. Traditional metrics, while useful for comparing broad characteristics, frequently fail to capture what makes a piece of content unique or perceptually different. This means that even if an algorithm generates technically distinct levels, players might still perceive them as bland or repetitive, leading to a less engaging experience.
The authors argue that this problem stems from the complexity of “composite artifacts” – game content made up of many smaller, often similar components. Players don’t experience every single part of a large map or settlement equally; instead, their experience is shaped by a meaningful subset of these components. Current evaluation methods often treat all parts equally, missing the elements that truly stand out and influence player perception.
Introducing Landmarks, Monuments, and Beacons
To address this, the paper introduces a hierarchical framework:
Landmarks: These are noticeable features within the game that act as foundational elements for the player’s experience. For something to be a Landmark, it must be perceivable (visible, audible, readable, or experienced through game systems) and stand out from its surroundings. This “salience” can be achieved through differences in size, color, pitch, or design. For example, a towering mountain in Minecraft, a distinct sound effect, or the brightly highlighted climbable objects in Mirror’s Edge all serve as Landmarks. They help players orient themselves and set initial expectations, but aren’t necessarily memorable on their own.
Monuments: Building upon Landmarks, Monuments are features that also possess an evocative nature. They strongly relate to a player’s past experiences, conjuring memories, feelings, or a specific atmosphere. Unlike simple Landmarks, Monuments carry a meaning beyond their immediate sensory perception. Think of the crashed spaceship in Subnautica, which immediately tells a story of survival, or the varied architectural styles of bases in No Man’s Sky that hint at different narratives. Musical themes associated with characters or areas, or the visual cues of item rarity in World of Warcraft, also function as Monuments. They are more memorable and guide the player’s interpretation of the game world.
Beacons: At the highest level of this hierarchy, Beacons are Monuments that intrinsically create a “Call to Action.” This means they encourage players to perform specific actions, not because of an explicit game reward, but due to an intrinsic desire to explore, investigate, or interact. This concept is closely tied to “affordances” – what actions a player perceives as possible. An iconic example is the tall buildings in Assassin’s Creed; they are Monuments (historical structures) but also Beacons, calling players to climb them for vantage points or to unlock map sections. Other examples include the distinct sound of Nirnroot in Skyrim prompting players to search for it, or the cooing of BB in Death Stranding warning of enemies. Beacons influence a player’s future actions and motivations, shaping their gameplay journey.
Also Read:
- Aligning AI Agents with Human Behavior in High-Stakes Simulations
- Navigating Conflict: How Language Models Are Reshaping Open-Ended Wargames
Practical Implications and Future Directions
The paper discusses how these concepts can be practically implemented. Identifying Landmarks can involve computational methods like visibility tests and salience checks, potentially adapting existing PCG metrics. Detecting Monuments and Beacons, however, is more complex due to their subjective nature. This might require empirical testing with players (surveys, recall tests) or leveraging advanced AI like Large Language Models to interpret associations and memories. Designers could also pre-define expected interpretations and calls to action.
Ultimately, this framework aims to provide a common language between technical game AI research and game studies in the humanities. By focusing on the elements that truly impact player experience, developers can design more engaging PCG systems, create more embodied artifact comparisons, and better approximate the uniqueness of generated content. This approach promises to accelerate the design-and-test loop in PCG, leading to richer and more player-centric game worlds.
