The Feedback Companion: local AI as a tutor for deep learning

In today’s education, feedback is central to helping students grow. But what if we take the way we give feedback to a higher level? Recently I adapted the Feedback Companion using a locally running AI with LLaMA 3.1 8B and the “Ask for better feedback” poster by Renske de Kleijn. The Feedback Companion not only gives feedback and feedforward, but also helps students by asking the right questions that encourage them to arrive at insights themselves.

How does it work?

The renewed Feedback Companion now works as a tutor that guides students by asking questions inspired by the method of Renske de Kleijn. Her poster has also been added to this blog post. Instead of giving direct tips and improvements, students are challenged to think about their own work, justify their choices and discover possible areas for improvement. This strengthens their learning process and helps them take active responsibility for their own development.

Why is this important?

Many students take in feedback passively: they process the tips and carry them out, without really understanding why. By asking questions such as “What do you think the most important areas for improvement are?” or “How could you measure this effect yourself?”, the Feedback Companion encourages them to think more deeply about their own work. It helps them progress in their learning process without being spoon-fed the answers.

Running AI locally: control over your own data

Another important aspect of this development is that the AI runs locally on a powerful laptop, so that students and teachers keep full control over their own data. This is of great importance at a time when privacy and data management are crucial topics. With the right hardware — such as a laptop with a good graphics card and enough video memory — users can use the Feedback Companion without worries, without depending on external servers or the cloud.

An experiment: how does it work in practice?

I carried out an experiment in which I, as a fictional student, had a conversation with the Feedback Companion. The result? The AI asked targeted questions that helped me think about my work and better understand how I could improve my approach. Instead of just giving feedback, the AI led me through a process of reflection and problem-solving.

I have attached the full conversation and a poster with the feedback questions by Renske de Kleijn. See how the AI helped me think more deeply about my project and how the right questions can lead to better learning outcomes.

What can you do?

Are you curious how you can use this yourself in your educational setting? The beauty of this approach is that with a reasonably powerful laptop and the right adjustments to the Feedback Companion, you can achieve similar results. By running AI locally, you stay in control of your data and can enjoy the same benefits, without external dependencies.

Need help? I am happy to help you further! Whether you want to know more about implementing AI locally or adapting the Feedback Companion to your situation, I am ready to support you online or in person. Let me know how I can help you create this powerful learning environment!