Anyway, if you think LaMBs are cool and would like to fund more research related to them, you know where to find me 😎

As a final note, I cheekily picked examples based on some my favorite games in this paper. I could not finish my PhD without citing WoW at least once 😜 (DID SOMEONE SAID [Thunderfury Blessed Blade of the Windseeker] ????????)

LaMB is the results of countless reading, discussions and meetings with plenty of fascinating people, and it would take forever for me to thank all of them. From the field of AI, but also game studies, and even people with a background in archeology, architecture and urbanism.

There are several approach for LaMBs: comparing their properties or their distribution in an artefact, the playing experience they create altogether, etc. As you might guess, there is an avenue for plenty of new research based.

Read the paper for all details!

Then, we also argue that LaMB can be useful for characterising an evaluated artefact. Rather than crunching it down to a series of computed values, it focuses on the core elements of the player's experience, the one that are noticed.

So what is LaMB useful for? Our intent with LaMB is to create a bridge between the technical field of PCG and other fields such as gamedesign, level design, and others even outside of games.

There are plenty of theories from other fields that could actually be used in LaMB identification. For instance, you can look at my previous work on isovist in games. With LaMB, we provide a framework in which such theory can be applied in the field of PCG for games.

One approach could be working with gamedesigners to create tailorred implementations. But in the paper we also mention fields that could lead to future implementations: agent-based evaluation, player modelling, player's profiling, etc.

Or simply by tuning the identification algorithm based on the game's design, maybe even by flagging some assets as potential landmarks, like taking into account that the rarest items in a given game are orange.

For monuments and beacons, things are more subjectives, and therefore harder to capture.

The automated identification of landmarks could be based on analysing the salience of a given feature of the artefact, or by looking at some of its properties such as its color. I recommend looking at "Towards a procedural evaluation technique: Metrics for level design" for some examples.

It is a huge sword with tons of FXs, making it highly noticiable. But a seasoned player can recognise it, and deduct its wielder is a powerful character. And in a PvP context, players might be inclined to runaway from someone holding this blade.

All these categories are nonetheless fluid. Monuments and Beacons are subject to interpretation, which means a player might not perceived them as such. One example we look at in the paper is [Thunderfury Blessed Blade of the Windseeker], one of the best weapon in original World of Warcraft.

They are tall building, standing out from the landscape. They usually are historic buildings, such as dungeons, minarets, cathedrales. And they also unlock parts of the map when they are climbed. Therefore, a player spotting a tall building in such games might be inclined to climb it.

Beacons are monuments with a "Call to Action", meaning that not only they remind the player of something, but they also suggest them to pursue a certain action. The best example for them are the vantage points in Assasin's Creed games.

Monuments are landmarks which evoke something. It could be related to the game story, real life tropes or cultures, etc. PCG settlements in NMS or Starfield, can have visibles variation depending on past event, the faction holding them, etc. The wreck in Subnautica gives all needed context.

Landmarks are the features that the player notices. Either because of their salience (e.g. how they stand out from the rest of the generated artefact), or properties such as a flashy color, loud SFX, etc. Think of a tall elevation on a Minecraft map, or the bright red paths in Mirror's Edge level.

Based on our previous research, we wanted a framework aiming at capturing 3 types of features: noticiability, evocativeness, and offered affordances.
LaMB is based on these three features and draw inspirations from various field such as theme park design.

We decided to create 3 nested categories, each of them being a specific type of landmark. What started as a study for level design led to a canvas that could be applied to any kind of game artefacts.
It is built on top of concepts such as indicators, perceived affordances, and indexical storytelling

It introduces a new theoric framework for automated evaluation of PCG artefacts, based on 3 categories: Landmarks, Monuments, and Beacons (LaMB for short). We wanted to have a tool that captured the uniqueness of a level. The concept of "landmark" kept coming back when we looked at other fields.