I am working on a project for school, and my idea is to use genetic algorithms to make the interactions between the human user and the computer players more realistic in video games. In other words, I would like to make the responses of the non-player characters more intelligent. I was wondering if anyone who has done game development or is doing some research in game development has any ideas on this subject. Does anyone think it can work, and is it a worthwhile task??

Any help would be greatly appreciated.

I played with a couple different approaches back in my MUD days, and the biggest problems I ran into weren't in generating intelligent responses but rather in NLP (natural language process). Of course, if you're doing conversation trees or something equally deterministic you won't have to worry so much about trying to figure out what the hell the player is saying... :)

Given that, the simplest approach that I found was to give a series of flags to determine a NPC's current state, angry/sad/inquisitive/tired/etc, then use a series of tables to look up valid responses. Granted this isn't really a show of intelligence, and depending on the size of your database you'll get repetitive results fairly soon... but it is a good place to start.