Using self-paced, `flipped' teaching to promote deep learning in an Earth Sciences programming course

Abstract:

Over the last year we implemented a complete restructuring of a second year Matlab-based course on numerical modelling of Earth processes, with changes aimed at 1) strengthening students' independence as programmers, 2) addressing student concerns about support in developing coding skills, and 3) improving key modelling skills such as choosing boundary conditions. To address this, we designed a mastery-based approach where students progress through a series of small programming projects at their own pace. As part of this, all lectures are `flipped' into short videos, allowing all contact hours to be spent on programming. The projects themselves are structured based on a `bottlenecks to learning' approach, explicitly separating out the steps of learning new commands and code structures, creating a conceptual and mathematical model of the problem, and development of more generic programmings skills such as debugging before asking the students to combine all of the above to build a numerical model of an Earth Sciences problem.

Compared with the previous, traditionally taught cohort, student questionnaires show a strong improvement in overall satisfaction. Free text responses show a focus on learning for understanding, and that students particularly valued the encouragement to slow down and work towards understanding when they encountered a difficult topic, rather than being pressured by a set timetable to move on. Quantitatively, exam performance improved on key conceptual questions, such as boundary conditions and discretisation, and overall achievement also rose, with 25\% of students achieving an `A+' standard of work. Many of the final projects also demonstrated programming and modelling skills that had not been directly taught, ranging from use of new commands to extension of techniques taught in 1D to the 2D case: strong confirmation of the independent skills we aimed to foster with this new approach.