Is Good Algorithm for Computer Players also Good for Human Players?

Naoki Konno, Kyoichi Kijima

Abstract


This paper aims to examine effectiveness of rational strategies for rough reasoning human players. Nowadays, computer players beat human champion players in many games (ex. Chess, Reversi, etc.) Actually, since computational power of computers transcends the human players, accuracy and volume of the search ability of computer players are superior to the human champion players in the end game phase. Then, the problem is that these computer algorithms are also effective for human players? The algorithms are basically composed by backward induction that is equilibrium concept for rational players. However, human players sometimes make wrong reasoning unlike computer players. In order to investigate the problem, we first propose a rough reasoning model that describes human imperfect reasoning abilities. This model is characterized by following two assumptions. The first is that as the payoff difference decrease, reasoning accuracy tends to decrease. The second is that as length of the tree increase, reasoning accuracy tends to decrease. We then make some examples of games and play them by some kinds of rough reasoning players with various algorithms. In the real game situations, accepted theories sometimes contradict to the rational strategies. We try to reveal the validity and effectiveness of the theories.

Keywords


bounded rationality; algorithm;game theory; rough reasoning

Full Text:

PDF