Good news for AI feedback bots such as the Feedback Companion – using AI in education without restrictions!

Over the past few months we have worked hard on the further development of the Feedback Companion, and now we can announce the good news that we can use AI in education without major restrictions. Why? Thanks to an important update from the AI Regulation (AI Act), there are clear guidelines indicating where AI systems in education do and do not fall into the high-risk category. This is very good news for tools such as the Feedback Companion, which are intended to improve and support education!

What does the AI Regulation say?

The AI Regulation (AI Act) states that systems with a direct influence on educational decisions, such as admission, evaluation of learning outcomes and monitoring pupils during exams, fall into the category of “high-risk applications”. This means these systems must comply with strict rules.

But… there is good news! If the AI system does not take direct decisions or materially influence a decision, it does not fall into the high-risk category. Think of systems that:

• Perform a narrow procedural task.
• Improve the result of a human activity, such as giving feedback on work.
• Support the process by detecting deviations, without replacing decisions.
• Perform a preparatory task for an assessment.

Why is this good news for the Feedback Companion?

The Feedback Companion, as we are developing it now, fits perfectly into these latter categories. The tool is not intended to take final decisions, but to support teachers and students by offering feedback and feedforward based on previously submitted assignments. It therefore does not count as a high-risk application, which means we can keep using AI without restrictions to improve the feedback process and make it more efficient.

A practical example: When a student submits an assignment, the Feedback Companion analyses the work and suggests improvements. This supports the learning process without replacing or influencing the teacher's assessment. It is always the teacher who takes the final decision. It can also improve the dialogue between student and teacher, for example because you can discuss the performance in the submitted work more concretely.

What does this mean for the GDPR and data in the cloud?

Although the AI Regulation gives us a lot of freedom in applying AI, we still have to take the GDPR (General Data Protection Regulation) into account. This means users' privacy must be safeguarded at all times. Fortunately there is good news on this front too: as long as personal data is processed and stored within the EU, we can use cloud solutions that meet the GDPR requirements.

Some important points to bear in mind:

1. Consent and transparency: Users must know how their data is processed. This must be communicated clearly and plainly.
2. Data minimisation: Only collect the data needed to improve the learning process.
3. Data storage and security: Whether you store data locally or in the cloud, the security of the data must always be guaranteed. So choose a cloud provider that complies with the GDPR guidelines.
4. Users' rights: Users must be able to view, correct and delete their data whenever they wish.

The next step: further embedding AI in education

With the current developments, we can safely continue working on AI-supported solutions such as the Feedback Companion, without falling into the high-risk category. This opens the door to more innovation and use of AI in education, as long as we strictly observe the GDPR requirements.

For anyone who wants to read more about the AI Act and the GDPR, there are important sources that can help you further:

• AI Act Regulation (2024): AI Act | Shaping Europe’s digital future (europa.eu).
• Dutch government guide to the AI Regulation, October 2024: Gids AI-verordening | Brochure | Rijksoverheid.nl.

With the Feedback Companion we can use AI with confidence and flexibility to give feedback and feedforward on student work, without taking over teachers' core decisions. That is good news for the future of education!

Running AI locally: why it is a game changer for education

Running AI locally, as with the Feedback Companion, offers many advantages for education. Instead of depending on cloud solutions, you can install and use AI models locally on your own computer or school server. This has a number of important advantages:

1. Full control over data: Because everything runs locally, you are in control of the data. This means you comply with strict privacy rules such as the GDPR, because the data stays within your own system and is not sent to external servers.
2. No internet needed: Because you run the AI locally, you do not depend on a constant internet connection. This makes it ideal for schools with limited or unstable internet access.
3. Cost-efficient: Local AI solutions save costs in the long term, because you do not pay for expensive cloud infrastructure or services. You make optimal use of the hardware you already have.
4. Flexibility and adaptability: You can adapt and train the AI models yourself based on your specific educational needs, which leads to personalised learning experiences for your students.

By running AI locally, for example with smaller LLMs such as LLaMA on an HP Fury laptop with at least a T1000 GPU, teachers and students can easily use AI without privacy or internet issues. This is also a nice addition to a possible future of personalised and scalable education, without worries about external dependencies!

Stay up to date with all developments by following the knowledge clips on YouTube, and visit our website for further explanation about AI and education.