Introducing atomic, re-usable feedback for the semi-automated assessment of mathematics tasks

Abstract

In mathematics education, students often exhibit recurring mistakes in their solutions, referred to as ‘structural error patterns’. These patterns offer an opportunity to streamline the feedback process by re-using feedback across similar errors, particularly when fully automated assessment is not an option (e.g. paper-and-pencil tests). However, to maximise re-usability, feedback must be formulated with care. To this end, we introduce the concept of ‘atomic feedback’, which involves breaking down feedback into small, independent units that address specific errors. While atomic feedback enhances re-usability and results in a greater quantity of feedback, it does not lead to time savings for teachers. Moreover, more feedback does not necessarily equate to better feedback: atomic feedback tends to be more general and descriptive, making it less to the point and less concrete than traditional feedback approaches. In this chapter, we explore the concept of atomic feedback in depth and present research results. The study was conducted using a custom-developed tool within Moodle, involving 45 mathematics teachers. Finally, we discuss applications and the potential for integrating AI and recommender systems to further enhance the intelligence of feedback suggestion systems in the future.