I got a bunch of early simulations running for various "social simulation" algorithms I'm testing. My first tests were a method of creating ways for characters to interact - a glorified "topic" list, with each character more or less interested in any given topic. The end result was not staggeringly interesting, so I added a loop where the more skilled you are in a topic, the more people respect you if they like the topic. That ended up with some kind of bizarre interactions - such as the man who hated aeronautics. Damn aeronautics!
But it also created a nice "web" of personal connections which could be used for a play loop. That means it can support a game. However, I need to avoid making the people into mere tokens, and there's three elements to that: limiting the number of people, making the people have very human interests, and making the humans active rather than controllably passive.
All are within reach - the second is actually hardest. I came up with a bunch of game ideas, but the one I like best is called "To The Future!"
You inherited a hotel out in the middle of nowhere from a distant family member. So you and all your geeky college buddies go to live there. The game is played by collecting scientists of various specialties and gaining upgrades which allow the hotel to be more attractive to more capable and established scientists.
The basic idea is that the more scientists you have, the more likely an experiment is to run amok and do serious (and humorous) harm. So you really have to decide whether you want to play it safe with a slower advancement or walk the edge with a few more scientists. I'm thinking the normal number would be six, with an effective absolute max cap of a dozen. Few enough that the player can keep them all as distinct personalities in his head.
There were a bunch of other ideas - such as a superhero team - but I like the geeks best. I love the idea of every character in the game being painfully socially inept, both because it's endearing and because it hides the weaknesses of the social engine.
I also performed some basic tests on drama generation and a learning system inside COGENT. The learning system is, so far, a failure. The drama system might be cool, once I've banged out some more bugs, but it would be best as a lightweight massively multiplayer game in which players did a lot of content generation.
Interesting fact of the day: steamrollers have squeegies on their "wheels".