Methods of Comparing Teacher and Student Programming Behavior During Live Coding

• In Event: A.I. and the Future of STEM Instruction: Designing New Models to Automate Feedback to Teachers

Abstract

As pushes are made to introduce computational thinking and programming to students in K-12 education, teachers are already beginning to integrate computing activities within their core classrooms (Cateté et al., 2018). Though professional development programs exist for preparing educators for this task (Jocius et al., 2020), many teachers are not formally trained in how to lead these activities. To mitigate this, we have worked with teachers through research practice partnerships to co-develop lessons and offer small-scale professional development and in-class support (Cateté et al., 2018). We taught teachers to instruct classroom lessons through the popular programming teaching strategy, Live Coding (Selvaraj et al., 2021), where teachers display the coding process to their students and students are asked to follow along and mirror actions.
This work has captured logged interaction data within the programming environment, similar to other systems that utilizes this data for adaptive support such as on-demand hints or feedback (Price et al., 2017). However, unlike most other contexts where students generally work independently, this work includes interaction data from both the students and their instructor. Comparatively less research has gone into adaptive systems that utilize both teacher and student interaction data. Investigating the relationship between student and teacher programming behavior during classrooms through data-driven methodologies might allow us to build support systems for teachers leading classrooms. This could take the form of systems that can inform educators about how the classroom experience is going either after the fact or in real-time. We believe these data-driven approaches could lead to both scalable offline and online analysis of programming instruction, informing the development of intelligent dashboards designed to aid both novice and expert programming teachers during live coding instruction.
This work uses data from nearly 300 students completing a computational modeling activity within a middle grade science course. Students spent three days going through the Use Modify Create progression (Lytle et al., 2019) in which students first learned to use and modify simulation code before creating an extension to the simulation. All three days of activities were led by an instructor, most being the classroom teacher. Our data set contains classroom interaction data from these 9 instructors leading a total of 19 classroom sessions on the final day of their activity (Lytle et al., 2020). We present a variety of methodologies for comparing the interaction patterns between instructor and students, focusing on ways to visualize how teacher actions lead to student actions and how the paths that teachers take to complete the assignment influence the paths students take. We propose methods for this data to be meaningfully given back to teachers as they are leading classroom sessions. We then propose potential interfaces for how these information sources can be added to real-time systems to better support novice teachers introducing computing activities in core classrooms.

Authors

• Nicholas Lytle

• Veronica Catete, North Carolina State University

• Yihuan Dong

• Mehak Maniktala

• Danielle Cadieux Boulden, Code.org

• Eric N. Wiebe, North Carolina State University

• Tiffany Barnes