Reducing drop-out rates with the help of AI: Unicorn University’s experience

We can only do that if we have data — and enough of it. That is absolutely essential for data-driven study management, which is ultimately what we are aiming for.

At Unicorn University, we have implemented a study management system built around a large number of activities that students complete throughout their courses. These activities are logged in our information system, which allows us to monitor in considerable detail how each student is progressing and, more recently, to predict potential academic failure in advance.

At the moment, based on continuous assessment results, we inform students whether they are at risk of failing. I believe this is valuable in itself, and it has already helped reduce the failure rate. But we are going further. We have now developed a trained model that combines classical statistics with AI.

The student then receives a comprehensive, context-based report containing study recommendations, including the option to use an AI study assistant. This is a step we would like to implement as soon as possible.

The key is to have data in the first place — and to have enough of it. This is directly linked to the need for both study materials, including self-assessment applications, and graded activities such as tests, assignments, and other tasks within the study system.

Based on this, we are able to build a database and use it to model the risk of academic failure. Simply put, the more we know about students and the way they study, the better we can identify their weaknesses and help them.

AI helps in two ways. First, we have a language model that acts as a study advisor. It can not only guide students through administrative matters, such as how to interrupt their studies or apply for an ISIC card, but also support them directly with their coursework — for example, by testing them on Algorithms or helping explain a specific topic they do not understand.

At the same time, AI is part of the predictive model and, most importantly, generates the content of the report provided to each student. The key difference is that the message is no longer simply, “You are falling behind, you should work harder.” Instead, students receive specific information about the areas they need to focus on, and AI also generates a recommended study plan tailored to those needs.

AI will change society as a whole — and that is already happening. It is difficult to predict exactly how, because the development is extremely fast. In general, I hope AI will become an excellent tool in human hands and under human control. We should not allow it to be the other way around.

In education, I hope students will use AI to study more effectively — as a tool that helps them process and understand information more easily. As a result, this could lead to higher study completion rates and, more broadly, to a better-educated society.

At the same time, I do have some concerns that students may start to see AI as something that allows them to complete their studies without actually understanding the subject matter. If that happened, the effect on study success — and, more broadly, on society’s level of education — could be exactly the opposite.