Supporting Effective Collaborative Learning in a Computer Science Intelligent Tutoring System

The standard paradigm of an Intelligent Tutoring System (ITS) exclusively affords one-on- one learning between a student and a computer tutor. However, the benefits of collaborative learning are extensive and include learning for transfer and development of higher order thinking skills such as planning. With this research as motivation, my work shifts the ITS paradigm to accommodate multiple-student, computer-mediated learning. The primary research objective is to investigate how design choice within a Collaborative Intelligent Tutoring System (CIT) impacts student learning and interaction. To begin, we synthesized research from Computer Supported Collaborative Learning and ITSs to develop a framework for design and evaluation of CITs. From there, we grounded our development and assessment of three versions of our CIT based on this framework. Over 200 students used the systems in the classroom and on average experienced significant learning gains. We compared our collaborative ITS to its one-on-one counterpart and found that students learn equivalently while being more efficient in timing, example use, and code correctness. Additionally, the evaluations showed that design choice influences student interaction such as symmetry, planning, and on-topic discussion. Finally, we discovered that dialogue-based analysis can provide significant correlations to measures of group cognition. Overall, my current research is motivated by my personal experience as an underrepresented student in the field with a strong desire to be connected to others as I learn CS.