Researchers work on procedural fun
Developing behaviors via genetic algorithms of various sorts has been around a long time now. You come up with a basic environment and ruleset, then you let loose millions of generations of simple AIs to keep trying to surivive. You then have the AIs tweak themselves based on what survived well, attempting to evolve the best survivor.
This can be used for lots of purposes — and now it’s being applied to game design. Starting with a simple Pac-Man like environment, researchers are generating zillions of procedural games, and then testing to see which is most fun. But how to measure the fun?
It should be pretty straightforward to see how game rules can be represented to be evolved: just encode them as e.g. an array of integers, and define some sensible mutation and possibly recombination operators. (In this particular case, we use a simple generational EA without crossover.) For other rule spaces, some rules might be more like parameters, and could be represented as real numbers.
What’s the much trickier question is the fitness function. How do you evaluate the fitness of a particular set of game rules? …
Our solution is to use learnability as a predictor of fun. A good game is one that is not winnable by a novice player, but which the player can learn to play better and better over time, and eventually win; it has a smooth learning curve.
There is a PDF with more details of the sorts of games which have been completely procedurally designed thus far. Reading the descriptions, you can recognize that the things that make them not fun look like issues with starting conditions, time limits, and the like. They have not yet done very many runs of the algorithm, so there’s more to come.
Additionally, it seems like there’s still things to tune in the algorithms proper — the games that have the highest fitness are the ones easiest to win, whereas the actual fun ones would be at a point between easy and impossible. One thing I wonder is whether this pushes game grammar stuff more to the forefront as well…
Still — it’s a step on the way towards developing things like automatic tuning mechanisms, metrics for learnability, and so on.
Also, it’s another valid citation for Wikipedia. ‘Cause, you know, the theory of fun entry is really skimpy. 😉 (Oops, canvassing!)
2 Responses to “Researchers work on procedural fun”
Sorry, the comment form is closed at this time.
You’re right that there’s a lots of work left to do – this paper is really just there because I like the concept so much myself and wanted to get it out as soon as possible! I’m working on an extended version of the paper for the TCIAIG journal, which will definitely include many tweaks and fixes for the basic learning algorithms, something which will hopefully translate into better generated games. Maybe I’ll extend the rule search space by adding new axioms as well.
I don’t know much (almost anything) about game grammar, but I’ll definitely take a look at it, as it seems relevant for what I want to do. Your “Theory of fun” was one of the main inspirations for the current work, as I mention in the paper.
The “game grammar” tag on the sidebar here should point you at most of the available resources. I think it would be directly applicable to what you are doing.
https://www.raphkoster.com/tag/game-grammar/