Background and motivation
Choosing an academic or career path is not a straightforward decision. Students often have to balance their interests, abilities, expectations, and external constraints, all within an increasingly complex educational landscape. While guidance systems typically rely on measurable indicators such as grades or employment prospects, one important aspect is often overlooked: student satisfaction. Yet, satisfaction plays a crucial role. It influences engagement, persistence, and overall success. The challenge, however, is that satisfaction is subjective and shaped by many interacting factors, making it difficult to capture and even harder to predict.
This led us to ask a simple but important question: can student satisfaction be anticipated before decisions are made?
A hybrid methodological approach
To explore this question, we developed a hybrid framework that brings together psychometrics and machine learning.
On the psychometric side, we designed and validated a measurement instrument to assess student satisfaction in a reliable and structured way. This allowed us to capture key subjective dimensions such as perception, expectations, and motivation. On the machine learning side, we used Support Vector Machines (SVM) to model the relationship between satisfaction and a wide range of variables, including academic performance, socio-economic background, and contextual factors. This approach makes it possible to capture complex, non-linear interactions that traditional methods often miss.
By combining these two perspectives, we aimed to connect human-centered evaluation with predictive modeling.
Key findings
The model achieved an accuracy of 89% in predicting student satisfaction. More importantly, the results show that satisfaction is not driven by a single factor. Instead, it emerges from a combination of influences, including family support, prior information about the institution, environmental expectations, and learning preferences such as group work.
This confirms the importance of considering both individual and contextual dimensions together.
Implications for guidance
These findings open the way for more informed and personalized guidance systems. By integrating psychometric insights into predictive models, it becomes possible to identify potential dissatisfaction early and support better decision-making. At the same time, such tools should be seen as a complement to human expertise, not a replacement.
Looking ahead
This work highlights the value of combining rigorous measurement with advanced analytical techniques. Future research could further explore explainable AI methods and extend this approach to larger and more diverse datasets.
Access to the article :
https://doi.org/10.1007/s44217-026-01337-9