Having it all: auto-graders reduce workload yet increase the quantity and quality of feedback

Abstract

Due to COVID-19, teaching has moved online at an accelerated pace, and this movement will partially be permanent. Online teaching implies an automatic assessment of exercises. Using automated grading, the studied web development course (N=257) managed to serve students promptly and increase the amount of feedback received by students even if the number of submissions increased remarkably.<br/><br/>Automatic graders guaranteed the uniformity of feedback, equal treatment, and most importantly, reduced the routine work of the personnel. Being less burdened, the course personnel could concentrate on assisting students in online discussion channels, where discussions were targeted for the students needing more help and support. Compared with previous manually assisted course implementations, the workload moved from "in situ" to prior to the course, where the most laborious part was the design of the exercises and the implementation of automatic graders. The amount of work for grading the exercises and assignment was decreased by about 70 per cent.<br/><br/>In the graders, the feedback given by them is of paramount importance and should suggest necessary improvements. The graders enforced good coding conventions and other targets set for the code (e.g., maintainability and accessibility). In some cases, this feedback was modified during the course based on the difficulties experienced to give more targeted advice. Automatic grading provided a way for students to iteratively improve their code based on the feedback. The software and methods used in this course could be applied to such other courses and domains, where automatic grading is considered helpful.