Erik Blankinship is a PhD candidate at the MIT Media Laboratory. He is in the Electronic Publishing group led by Senior Research Scientist Walter Bender. Erik received an MS in media arts and sciences from MIT, an MEd from Harvard University, and a BA with a concentration in folklore and mythology from the University of Maryland at College Park.
Who's Got Game (Theory)? Articulating Strategies to Learn About Competition
Many people play strategy games against computer opponents, but don't have a clear understanding of how they function. Perhaps the best known computer opponent is IBM's Deep Blue which beat World Chess Champion Kasparov after nearly 50 years of artificial intelligence research on algorithms to beat expert human chess players. Among others, Deep Blue used the minimax algorithm to calculate and evaluate billions of potential moves and countermoves to determine an optimal strategy. These and similar game theoretical algorithms can be applied to the numerous strategy games played by children and adults. Game players may benefit from understanding these algorithms to improve their own performance. Further, strategy games can become informal contexts for learning the mathematics and computer science related to decision making.
Building your own Deep Blue is a formidable challenge, but is perhaps a good way to learn the computer science and mathematics of decision making. Since most people aren't computer scientists, they need a way to describe game strategies so they can be executed by a computer they can play against. This exhibit features a graphic toolkit for designing strategy-games and computer opponents, tailored to support learning artificial intelligence techniques for deciding which next moves are better than others. The demonstration of the toolkit is complemented by initial results of a study of how kids articulate their strategies and modify them after seeing their effect on game play.
Another feature of the exhibit is an implementation of a specialized "strategy language" and graphical editor developed to help kids articulate their strategies and see how they play out when executed by a computer. The language is a declarative framework for creating if-then heuristics, game-state evaluators, and searching state space (e.g. with minimax, maximax, minimin). The strategy language is designed to allow kids to explore artificial intelligence techniques by describing them instead of coding them.