On relating local and global factors: a case study from the game of Go

Traditional artificial intelligence (AI) approaches to programming the game of Go are based on the translation of local information (modelled by pattern recognition processes) to global symbolic form (modelled by rule-based systems) for access by symbolic reasoning processes. We explore the converse process, seeking to relate local and global factors by integrating global factors into a local representation which can then be accessed by symbolic reasoning processes. We demonstrate one method for such an integration, the interaction of bottom-up and top-down processing. The algorithm we use in our simulation integrates global factors by directly modifying the numeric information contained in the local representation of the board. We use Go as an example of a domain with the characteristic that local and global factors cannot be identified independently of each other. Thus, to form a representation of a Go board requires an interaction between bottom-up processing (to identify local factors) and top-down processing (to identify global factors). In the final section we briefly relate these constraints to other domains