Adaptive search space pruning in complex strategic problems

Area of Science:

• Cognitive Science
• Artificial Intelligence
• Game Theory

Background:

• Humans make complex strategic decisions despite limited cognitive resources.
• Efficiently searching vast possibility spaces is crucial in strategic games.

Purpose of the Study:

• To investigate human search strategies in k-in-a-row games.
• To identify heuristics used for pruning search spaces.
• To evaluate the effectiveness and trade-offs of these heuristics.

Main Methods:

• Studied player search strategies in k-in-a-row games.
• Employed computational simulations and analyzed behavioral data.
• Varied parameters like shutter size, complexity, noise, and computational limits.

Main Results:

• Players utilize scoring strategies and a "shutter" heuristic to prune search spaces.
• A narrow shutter strategy is dominant across most simulated conditions.
• The shutter heuristic improves deep learning network performance under computational constraints.

Conclusions:

• The "shutter" heuristic is an adaptive strategy for navigating complex strategic games.
• This heuristic offers a trade-off between search efficiency and overlooking opponent moves.
• Findings suggest implications for both human cognition and AI development in strategic decision-making.