Selfish Search on Playing Shogi

This analysis explores the fundamental differences between human and computer decision-making processes in the game of Shogi. The central thesis posits that human players utilize a specialized, cognitively constrained decision-making method termed "selfish search," which contrasts sharply with the exhaustive, parallel-processing nature of computer-based alpha-beta search algorithms. While computers rely on deep, wide, and structured search trees, human players are limited by psychological restrictions, including finite search speeds and an inability to perform complex parallel calculations.

Methodologically, the research draws upon psychological experiments, including interviews with expert players and the analysis of verbal protocol data from participants of varying skill levels. Findings indicate that human players significantly reduce the set of potential moves to a small number of candidates, which are then evaluated intuitively. Furthermore, human prediction is characterized by a linear, sequential approach rather than the comprehensive MIN-MAX calculations employed by machines. As skill levels increase, both the speed and the number of moves considered by human players rise, yet they remain bound by a maximum processing capacity of approximately ten moves per minute and a few hundred total moves per search.

To bridge the gap between human intuition and machine logic, the study proposes the "Selfish Search" system. This algorithm integrates a knowledge-driven component—the Human Intuitive Thought (HIT) project—which uses production rules to score and rank legal moves. By mimicking the human tendency to focus on a narrow, high-priority subset of moves and performing sequential, linear evaluations, this system aims to create a more human-understandable decision-making process. The research concludes that "selfish search" represents an optimized cognitive strategy for navigating the vast search spaces of complex combinatorial games under inherent human psychological constraints.