Collaborative Learning with the Cognitive Tutor Algebra. An Experimental Classroom Study

Interest in developing improved methods for mathematics instruction has increased since TIMSS and PISA. In our project, we investigate a new way to promote learning in mathematics: we enhanced the Cognitive Tutor Algebra, a computer-based intelligent tutoring system for mathematics at the high school level, to a collaborative learning setting. Although the Algebra Tutor has shown to increase learning substantially, there are also several shortcomings. For instance, learning with the Algebra Tutor places emphasis on improving students’ problem solving skills, yet a deep understanding of underlying mathematical concepts is not necessarily achieved. To reduce these shortcomings, we extended the learning environment to a dyadic setting, thus adding new learning opportunities such as the possibility to mutually elaborate on the learning content. A script was developed to guide students’ interaction and to ensure that students profit from these new learning opportunities. Our script followed a general jigsaw-scheme, i.e. it distributed expertise for the problem between partners and allowed them to prepare individually for the following interaction, thus laying the grounds for effective collaboration. During the collaboration, the script provided additional support: it prompted fruitful interaction and adaptively supported students as they encountered difficulties. Following each problem, it guided the dyads to reflect on their interaction in order to improve subsequent collaboration. In an experimental classroom study taking place over the course of one week, we compared three conditions to evaluate the impact of collaboration in a cognitive tutoring setting: individual learning, unscripted collaborative learning, and scripted collaborative learning. After a two-day learning phase, several post tests were administered to assess learning on three levels: reproduction, transfer and future learning. In the collaborative reproduction test, we found a higher need for cognitive tutor assistance in scripted students with low prior knowledge (note that this was the first post test after script support had been removed). In the subsequent individual reproduction test, however, differences between conditions were no longer observed. In fact, the learning outcomes of students in the two collaborative conditions were comparable to those in the individual condition even though collaborative students had solved fewer problems, i.e. had had less practice, during the learning phase. Analysis of the transfer test did not reveal differences between conditions. Finally, scripted collaboration better prepared students for future collaborative learning situations as compared to the unscripted collaboration condition. In addition to presenting the study and its results, this paper discusses the advantages and methodological challenges of (collaborative) in vivo experimentation, such as the issues of betweenvs. within-classroom designs and evaluation of dyadic data. Collaborative Learning with the Cognitive Tutor Algebra 3