Using software repository authorship data in evaluation of group work

Abstract

Group work is very common in university courses that teach computer programming. This work is very hard to evaluate and grade, and usually results in very high grades being given to all students regardless of individual participation. This thesis studies the possibility of using software repository authorship data to find out how much of the project each individual student is responsible for. If such data is available, can it be used in the evaluation process to more fairly grade group work projects? It also considers challenges with groups dynamics that might be related to an individual's contribution anonymity within the group. The data used in the analysis will come from a course run at NTNU in the autumn of 2016. Other results from the course, like exam scores will be used for comparisons.

The analysis of the data shows that it is indeed possible to extract data on how much of a project each student is responsible for. It also shows that there is some correlation between the data and how well the students did in other aspects of the course, like the exam.