Seeing the big picture: Influence of global factors on local decisions

This paper extends recent work studying the developmentof human expertise in the game of Go. Although itappears like a simple game on the surface, Go is actually themost difficult of all established games for artificial intelligence,with no computer program yet reaching the top internationallevel on a full 19×19 board. On smaller boards with sizes like9×9, computers are competitive, implying that the understandingof complex global interactions is the key to human superiority.The temporal analysis of game positions yields some interestinginsights in into local/global analysis. By mining thousands ofpositions from online games, we show that at some player levels,the sequence of plays leading up to a local position is a strongerdeterminant of the next move than the position alone. Thissuggests that the sequence of plays is an indicator of globalstrategic factors and thus provides a context for the next movein addition to the local position. Using perceptual templatesintroduced in other work, we demonstrate that this global contextappears at the very earliest stages of cognition.