Breaking Bias: Measurements, Potentials, and Limitations for Modelling Study Success by Performance and Diversity Factors

Maria Krakovsky, René Krempkow, Larissa Bartok, Karl Ledermüller, Julia Spörk

Publikation: Beitrag in Buch/KonferenzbandBeitrag in Konferenzband

Abstract

This study investigates the influence of performance and diversity factors on student success using machine learning models. Two case studies from Austrian universities are presented, comparing the predictive power of models with and without diversity related factors. While performance indicators seem to have larger impact on student success, diversity factors can slightly improve model accuracy and help identify at-risk students. However, the importance of the use of diversity indicators in predictive models varies depending on the study program, the student population and on the aim with which the analysis is carried out. The study highlights the potential and limitations of using machine learning models to predict student success and emphasizes the need for context-specific analysis to avoid generalization and ensure fair and effective interventions.
OriginalspracheEnglisch
Titel des Sammelwerks28th International Conference on Science, Technology and Innovation Indicators (STI2024)
Untertitel des SammelwerksBerlin, Germany, 18-20 September 2024
VerlagFraunhofer Institute for Systems and Innovation Research
Seitenumfang15
DOIs
PublikationsstatusVeröffentlicht - 2024

Zitat