From: AAAI Technical Report FS-02-05. Compilation copyright © 2002, AAAI (www.aaai.org). All rights reserved. Towards Incorporating Intent Inference into the Game of Go Timothy T. Huang Department of Mathematics and Computer Science Middlebury College Middlebury, VT 05753
[email protected] Abstract We begin this paper by presenting an overview of game Computer programs for the game of Go play only at the play in Go and the factors that make computer Go level of an advanced beginning player. The standard challenging. After motivating the use of intent inference in approach to constructing a program based on brute force Go, we consider the components of an intent inference game-tree search does not work well because of the game agent: the inputs it can be given, the possible outputs it can tree size and, more significantly, the difficulty in produce, and the agent architecture that might generate constructing fast, accurate heuristic evaluation functions. In those outputs. We discuss how an architecture based on this paper, we consider the use of intent inference in a Go dynamic probabilistic networks seems well-suited for program. In particular, we discuss how models of an intent inference, and we describe how inferred intentions opponent’s long-term playing style and short-term can be used to influence move selection. Finally, we intentions can direct the exploration of candidate moves and influence the evaluation of game positions. We propose a compare intent inference in Go with intent inference in probabilistic approach to user modeling and intent other fields. inference, and we note key issues relevant to the implementation of an intent inference agent.