RAPP – Responsible Academic Performance Prediction
From March 2021 to February 2024, the RAPP project was carried out at HHU, which was funded by the Federal Ministry of Education and Research (BMBF). The focus was on so-called Academic Performance Prediction Systems, or APP systems for short, which are used for the early detection of dropouts, the prediction of study times and the prediction of student performance.
Such systems are often faced with the problem that students consider the analysis of their study progress to be problematic for data protection reasons; these concerns are also shared by many students at Heinrich Heine University. The aim of the RAPP project was to develop a non-discriminatory, anonymized and fair approach in order to make ethical statements about study progress and to take away students' fear of APP systems. An interdisciplinary team of computer scientists, sociologists, communication and media scientists and various university institutions, such as the data protection officer or the student advisory service, investigated the question of how such systems can be designed responsibly. The researchers focused on the question of which data can lead to discrimination and how good the predictions made are without this discriminatory data. In a second step, the focus was also on the perception of the system by those affected and the development of action guidelines for student advisory services.
On a technical level, the RAPP project was based on a combination of machine learning and explainable AI processes, which should make it possible to identify the factors that lead to students dropping out, for example. In the summer semester of 2021, a total of 34 expert interviews were conducted with academics and lecturers as well as students from the Social Sciences and Computer Science degree programs. One of the things that became clear from the interviews was that success in studies needs to be viewed in a differentiated way. Based on these findings, two concepts of success in studies were developed.
Established concept of success: this includes completing the course within the standard period of study, a low drop-out rate and a good average grade.
Student-centered definition of success: students define success primarily in terms of personal development, satisfaction with their studies and a good work-life balance
In the further procedure, panel surveys were carried out in which the students' performance data was analyzed in addition to socio-demographic characteristics and the intention to drop out or transfer. In summary, the researchers' aim was to expand established APP systems to include the concept of responsibility in order to make non-discriminatory predictions and provide recommendations to HHU that can be used to support students in successfully completing their studies.
The project was funded by the BMBF over three years (01.03.2021 - 29.02.2024) with around 1.3 million euros. More information can be found here or on the official project website. The DIID members Prof. Dr. Frank Marcinkowksi (KMW), Prof. Dr. Ulrich Rosar (Sociology), Dr. Christopher Starke (KMW) and Dr. Johannes Krause (Sociology) and Prof. Dr. Stefan Conrad (Computer Science), as well as other HHU researchers, are involved in the project.
A good and brief overview of the project work is provided in the accompanying précis. Detailed results can be found in the final report.
Contact
Prof. Dr. Stefan Conrad (stellvertretender Sprecher)