Productive dialog during collaborative problem solving

Collaboration is an important problem-solving skill; however, novice collaboration generally benefits from some kind of support. One possibility for supporting productive conversations between collaborators is to encourage pairs of students to provide explanations for their problem-solving steps. To test this possibility, we contrasted individuals who were instructed to self-explain problem-solving steps with dyads who were instructed to jointly explain problem-solving steps in the context of an intelligent tutoring system (ITS). The results suggest that collaboratively developed explanations prompted students to remediate their errors in dialog, as opposed to relying on the ITS for assistance, which is provided in the form of on-demand hints. The paper concludes with a discussion about implications for combining proven learning interventions. Introduction As is evident to those who live and work in societies with advanced technologies, the world is not only becoming a smaller place, but the demands for collaboration are expanding across disciplinary (Schunn, Crowley, & Okada, 1998) and geographic boundaries (Friedman, 2006). Individuals are finding themselves collaborating in new ways that have been made possible by recent advances in high-speed networks and digital forms of communication. For individuals to stay competitive on a global scale, they need to develop their collaborative skills. The field of the learning sciences is uniquely positioned to provide recommendations for how to best optimize those collaborative skills. In the paper that follows, we attempt to develop the following argument. First, it is evident from the collaborative problem-solving literature that, when done “naturally,” collaboration is not much more effective for learning gains than solo problem solving (Hill, 1982). Attempts to optimize collaborative learning have included various scripting manipulations that increase learning gains (Rummel & Spada, 2005). However, novices generally do not use these effective modes of interaction to communicate their ideas; therefore, the interactions must be taught. Moreover, attempts to use computers to elicit improved collaboration via scripts have floundered because of their inability to understand natural language (Soller, 2004). The research problem addressed in the current paper is how to use computers to help increase learning during collaboration. Toward that end, the paper is organized into the following sections. First, we will highlight two effective learning situations: self-explanation and peer collaboration. Then we will introduce an intelligent tutoring system for physics, called Andes, which has also been shown to increase individual learning. After the background for the study has been presented, we will report on an experiment that contrasted self-explaining with peer explanation in the context of using the Andes physics tutor. Finally, we will conclude with a discussion about leveraging the impact of various learning interventions. Learning Alone: Self-explaining Worked-out Examples When enrolled in a course like physics, much of the learning that takes place outside the classroom is done individually. That is, students are generally responsible for learning the course material from a textbook, which often contains worked-out examples. On first inspection, worked-out examples tend to be fairly impoverished, in the sense that they typically omit information that needs to be supplied by the learner (Chi & Bassok, 1989). While examples may exclude some information, students prefer to learn from them, especially during the initial acquisition of a skill (Pirolli & Anderson, 1985). How do students learn from incomplete worked-out examples? One hypothesis is that students attempt to explain the examples, line-by-line, to themselves (Chi, Bassok, Lewis, Reimann, & Glaser, 1989). This study strategy goes by the name of the selfexplanation effect (Chi, 2000). Self-explaining is a constructive learning activity because the student is actively trying to make sense of the material from his or her own background knowledge. For instance, consider the following monolog from a student in a second-semester physics class (Hausmann & VanLehn, 2007). In this experiment, the student was asked to study an example after solving an isomorphic problem. The example was broken down into problemsolving steps, which were related to the motion of a charged particle in a region of a uniform electric field. In this episode, the student had just watched a video-based example of a step where the solver drew a force vector in the opposite direction of the electric field vector (as per the vector equation F = qE, where F is the force due to the electric field; q is the charge on the particle, which is negative in this instance; and E is the electric field). The student is attempting to make sense of this step (see Table 1). Table 1: An example of a self-explanation (SE) episode.