Abstract:

In higher education, the assessment of a subject often involves weighting grades obtained from evaluations of various components such as theory, laboratory work, and presentations. Determining the appropriate weighting is a challenging task. While each teacher understands the relative importance of these components, expressing them as numeric weights can be difficult. Moreover, if multiple teachers are involved in assessing the subject, they may define different weightings despite having similar perceptions of the relative importance of each component.

Using a weighting scheme that does not align with a teacher's perception of the component's importance can result in inconsistencies. For instance, a teacher may believe that student A should receive a higher grade than student B, but the chosen weighting scheme calculates the grades in the opposite order.

In this work, we propose a technique for defining weightings based on the global preferences of teachers regarding grades for each component. The teacher only needs to determine which students should receive higher grades, using the grades of the individual components as input. The technique can incorporate preferences from different teachers, and it generates a weighting scheme that aligns with the majority of expressed preferences.

This process can be carried out after completing a subject, once all grades are known, and before defining the weightings for the next academic year. The work not only presents the theoretical definition of the process but also describes a case study involving the definition of component weightings in the subject of Database for the computer science program at Oviedo University.

Keywords:

Assessment, weighting, preference learning.