Here are a few AI articles which could be applied to Magic. The Hoyle program has a set of advisors which help the program learn and act more like a human. Temporal difference is a type of re-enforced learning. The TD-Backgammon program learned to play backgammon just by playing itself over and over again. The program was good enough to compete against world class players.
These articles are very high level which means that are very dense and hard to read. There is some good info in them but you have to take your time.