Evaluating the Effect of Social Conversation on Learning, Interaction, and Perceived Interdependence in a Collaborative Math Problem Solving Environment

In this paper, we explore the effect of social prompts offered by a computer supported collaborative learning (CSCL) environment on student attitudes and behavior towards each other. We do this by experimentally contrasting collaboration in two configurations, one that presents students with social questions in between math problems and then uses the answers to those problems to tailor the cover story and associated hints for the next problem, and a control condition that uses the same problems, but does not use social questions for tailoring. Our finding is that the social prompts changed the attitude students displayed towards one another. Whereas in the control condition students took on a more competitive attitude, in the experimental condition students were more playful. On the questionnaire students reported exchanging more help in the experimental condition. An analysis of the corpus reveals no significant differences in amount of help exchanged, but there was a significant difference in terms of non-help-related conversation such that the proportion of episodes where help was provided was higher in the Experimental condition. In terms of learning gains, the trend was consistently in favor of the experimental condition in terms of learning between the pretest and the quizzes after each lab session as well as the posttest, however these trends were not statistically significant overall. The strongest learning result is that students in the experimental condition learned marginally more on one unit of the material on the second lab day than students in the control condition (p=.06, effect size .55 standard deviations).