GAME INTERACTIONS WITH “HUMANS” AND “MACHINES” AND THEIR RELATIONS TO TACTICAL BEHAVIOR AND SUCCESS1
Albert E. Myers, Cecil A. Gibb, Carolyn B. McConville · ETS Research Bulletin Series · 1963
ABSTRACT An experiment was performed in which Ss engaged in a “Path and Obstacles” game where the S had the task of tracing a path on a 5 times 5 plugboard in such a way as to avoid hitting any of the five obstacles that had been placed by the other player. Half the Ss were told that the other player was a machine which had been programmed to play the game and the other half were led to believe that they were playing against another human. In addition, half the players were told to compete against the other player while half were told to cooperate. In every instance the setting of the obstacles by the other player was determined by a prearranged plan. In the Human condition the other player was an accomplice who played according to a prearranged plan; the same plans, which determined the placement of every obstacle on every trial, were used in the Machine condition. Two strategies of obstacle placement were used. One, called the Fixed Path strategy, moved the obstacles in such a way as to never block one certain path. The trial to trial movement of the obstacles was basically random. The other strategy, called the Variable Path strategy, did not move the obstacles randomly from trial to trial, but rather in a highly systematic way. Whereas the Fixed Path strategy encouraged the S to use the same path repeatedly, the Variable Path strategy could not be solved by repetition. The latter encouraged the S to find a principle which was responsible for the movement of the obstacles. Striking differences in success were found. The Ss performed much better when their competitive opponent was a “machine” than they did when he was a “human.” On the other hand, cooperation with “humans” was significantly more successful than cooperation with “machines.” It was concluded that an S's expectations concerning the other player are a critical element in determining whether his performance on the task will be affected by a “human” or “machine” other. It was also tentatively concluded that “human” others induced more irrationality than “machine” others.