I'm interested in social intelligence. By this I mean the problem of acting in an environment which includes other agents like youself which have goals which may or may not be compatible with your goals. To put it another way, I'm interested in games with the following properties:

1. A large number of players (more than 2, and potentially thousands)
2. Non-zero sum (This is somewhat optional, as social activity can still exiss in zero-sum games with more than 2 players)
3. Communication between players (not necessarily information about the environment, but definitely some channel for communication of social agendas)
   1. An interesting side point is the study of how to allocate your communications resources when your communication channel is limited. I'll probably start out experimenting with games in which players can communicate as much as they want to between moves, but it would be interesting to limit their communications, and see how that affects their social strategies).
4. Iterative play (Players need to interact repeatedly in order to build relationships)
5. Trade of materials/Currency (This is also somewhat optional - I'm including it because I think that it opens up the set of possible strategies enormously. I might leave it out initially though, just for simplicities sake)
6. Cooperative and hostile interactions (It's easy to cooperate when you have no other choice. Contrariwise, it's impossible to cooperate if there are no mechanics to interact peacefully with other players. There should be a choice. Strictly speaking, this is another optional one, as there is still room for negotiation when you can only hurt others (mainly in the realm of choice of targets) but I think it's simpler if you have both options available)

Here are the subgoals I'm interested in pursuing:

1. I'd like to develop a social language or set of concepts that computers could use to reason about social information. It should be as limited as possible, while still conveying sufficient information to be useful.
2. I'd like to develop an online game in which human and computer players could interact, using the previously mentioned language, which could be used to test algorithms. I would like to have human players, because they are very efficient at locating exploitable flaws in computer AI. The Neverwinter Nights platform looks like it might be useful in this regard.
3. Recognizing that developing an online game which is also entertaining to humans would be very difficult, I'd like to develop a computer only environment which will support socially interesting interactions, like trade, alliance, war, coordination of activity, justice, information transfer etc.