For testing out the ideas of my thesis, I am developing a game, Monopoly like, in which the users can experiment the relation between land price and the factors connected with its determination.
My objective is to constitute a software agent that can help the human agent on the resolution of the problem (aka win the game). This software agent should assume a symmetrical position than the human agent. Therefore, it wont have any hidden knowledge. For the same reason, it should solve the problem through some approximation heuristics and not through complex regression methods.
My problem can be schematized as follows:
1. A two dimensional map of the city is divided in cells;
2. For each cell there is a value of the land p, that is determined with a linear function like the one at (1):
3. The values of x, for k=1,2,…, m are independent variables. Some of them are constant during the game. Others are variable (i.e., pollution); others are defined by the game interaction (i.e., the presence of an hotel nearby).
Finally, this software agent should be able to solve the problem by approximation, participating on the game and applying some heuristics ad hoc.