Daniel MacDonald occasionally mentions video games. It turns out some researchers use strategy video games for exploring decision theory models. (Partially observable Markov decision processes and partially observable stochastic games are examples of such models for decision theory.) Frans A. Oliehoek and others at the Intelligent Systems Lab, at the University of Amsterdam, have developed the Multi-Agent Decision Process (MADP) Toolbox, "an open source C++ library for decision-theoretic planning under uncertainty in multiagent systems."
Some researchers, I guess in this lab, have integrated MADP into StarCraft, a real-time strategy game. StarCraft has a science fiction setting, and MADP routines are used to calculate policies for one of the three races in the game.
I consider Edward Castronova one the most interesting researchers exploring the intersection of computer games and economics.
Don't tell Brad DeLong (who at one time found he had to choose between playing Civilization or doing economics) about this post.
0 comments:
Post a Comment