Walker - Features of Adaptive Assistance that Improve Peer Tutoring in Algebra

In our project, we analyze how conceptual and elaborated content in student dialogue during a peer tutoring activity influences the learning of both the help-giver and the help-receiver. We further investigate how we can support poor collaborators in producing those beneficial interactions by automatically assessing student dialogue and providing assistance when appropriate. We have recently conducted a study comparing adaptive support to peer tutoring to fixed support, and found that the adaptive support indeed improved student peer tutoring interactions. We intend to take two further steps: 1) Continue analyzing the data to examine whether the improved interaction related to increased collaboration skills and learning outcomes (to be completed by October 15), and 2) Conduct a user study examining how public and private features of different types of assistance relate to student feelings of accountability and subsequent likelihood to incorporate the assistance into their interactions (to be completed by January 15). By the completion of the project, which we plan for at the end of May, we will have an increased understanding of how specific features of adaptive assistance can support students in interacting productively and thus in learning from collaboration. By that time, we will have analyzed the data from the January study, and we will have produced an adaptive system capable of supporting peer tutoring in Algebra, and demonstrated how features of the adaptive system influence student peer tutoring behavior.

This project relates the more general thrust goals as follows. It is examining how features of assistance affect the three aspects of accountable talk: accountability to knowledge, accountability to rigorous thinking, and accountability to the learning community. Steps are being made toward the ambitious goal to operationalize and assess these aspects of accountable in real time as students interact and receive assistance in this computer-mediated environment. There is also a potential to code the three way dialog (student tutee, student tutor, and computer tutor) for transactivity. In particular, the student tutee's dialog moves have not yet been coded, but appear to have interesting elements, like asking for specific help or self-explaining, that may well connect to transactivity codes. Finally, there is a potential to analyze the computer tutor's reflective prompts for similarity with accountable talk moves and associated effectiveness. Some of the prompts were indeed inspired by accountable talk moves.