Recently I became interested in the idea of computer games as living creatures. The primary function of any species is to ensure its own continuing survival — otherwise the species simply dies off, and its genes will not propagate. What if we apply this same principle to the study of computer games that evolve to optimize for their own continued existence?
I looked around a bit and found that there has been very limited research in this area, but not really very much. And it doesn’t seem to have been studied as a formal problem. By “formal” I mean breaking the problem down into genotype and phenotype. A “genotype” is the genetic description of something — the list of instructions, as it were. For example, our DNA contains our genotype, and a cooking recipe is the genotype of that delicious dish your mom cooked last weekend. A “phenotype” is a description of the kind of thing you get when those instructions are carried out, like a human being or a yummy ratatouille.
Rather than design specific games, I’m thinking it would be interesting to create a computer program that can create lots of different games. When you tweak various parameters in this game generator, different games come out. I’m not saying that there is some magic way to create this program — you’d still need to carefully design the game generator, using all the skills that a good game designer must have. But when you were done you’d have not one game, but a universe of possible games.
Most of these games would be terrible, but certain combinations of parameters would produce magical results — truly marvelous games that are fun and exciting to play. But how do you find those particular games within the mass of possibilities?
This is where crowd-sourcing comes in. You put these games up on-line, and let anybody play them. Some of the games will be boring, and people won’t be drawn to them. But others will find an audience. The genotype of those more fun and playable games will gradually spread, as more people play them.
Meanwhile, you continually tweak the parameters behind the scenes, so that the game is slightly different for each player. Certain tweaks will make the game more fun and popular — people will continue to play it longer — and others will have the opposite effect.
Eventually, games might emerge that are fun and exciting for many people. Such games will not have been built by any individual, but rather evolved organically, through the collective mind-share of the community of players.
The species will have ensured its own survival.
I think the games would need to be really easy to get into, so that players wouldn’t mind trying a few different ones to find a good one.
I’ve thought for a while about writing a multiplayer game involving Karl Sims’ creatures. As players protected and crossbred their favorite beasts, the game would evolve towards maximum fun. But that’s different than evolving the rules of the game.
Perhaps if you started with a population of viable games, and kept the mutation rate low, you could attract a large enough base of players to make it work?
Good point! Hmm, maybe we could work on this together.
That sounds fun. But I need to finish my book first.
In fact, this is what Zynga is doing … they publish many variations of their social games and do a bunch of data crunching to see which features to propogate globally for each title.
Cheers!