An experimental study on collaborative problem solving with Huamn/Agent via the Internet

350303150 Yoshie BABA


In this study, we investigate empirically how people are affected by a partner when they work together. In the collaborative work, a partner is a very important factor that affects both the performance and process of problem solving. Especially in a situation in which people communicate via computer networks, there may be a case where an actually partner is not identical to an imaginary one; i.e. a person thinks to collaborate with a human (an imaginary partner), but the fact is that he or she collaborates with a computational agent program (an actually partner). And the reverse situation may exist. In this experiment, we set up an environment where people solve a problem together via the Internet. We investigate whether or not (1) by the instruction, the image of a collaborative partner affects the process of collaborative problem solving, (2) the difference of strategies of partner's solving problem affects the subject's problem solving. We used Wasonfs 2-4-6 task as experimental task. The reason for using this task is that it has been used as a standard experimental task in studies on human discovery. Especially focusing on hypothesis formation and testing in scientific activities, we attempt to understand (1) through hypothesis verification how people's hypothesis-testing strategy is influenced by a partner's strategies, and (2) via hypothesis formation how people change to refer to another hypothesis given by a partner. The experimental factors are as follows: Imaginary partner (Human/ Agent) x Actual partner (Human/ Positive test Agent/ Negative test Agent). Positive test Agent and Negative test Agent are used for controlling the partner's hypothesis testing, where Positive test Agent uses only positive tests for hypothesis testing and Negative test Agent uses only negative tests. The subjects are ninety-six undergraduates and are assigned to each condition randomly. The results of the experiment are: (1) regardless of the difference of a partner, a bias in human behavior, such as the positive test bias in hypothesis testing, is not influenced, (2)the degree of using information given by a partner such as reference to a partner's hypothesis changes, and (3) the factor of the difference of the imaginary partner doesn't affect on the problem solving.