Ribe Gymnasium

Ribe Gymnasium is exploring how generative AI can become part of teaching without replacing students’ subject knowledge, critical thinking, and responsibility. FERC – Frame, Explore, Refine, Commit – provides a practical structure for keeping human judgment active throughout the use of AI.

The central idea is to shift attention from the final product toward the learning process: if AI can produce parts of an assignment quickly, teaching needs to make visible where students need to make choices, evaluate alternatives, and take responsibility for the result.

Teachers at Ribe are discussing not only how to bring AI into education, but the more fundamental question of how to educate in an AI age. This became concrete when an English teacher questioned what written production should look like if AI can already generate strong academic texts. The discussion therefore moves beyond protecting existing assignments toward asking what the core purpose and learning of a subject is when parts of the traditional product can be generated by AI.

At the same time, teachers remain constrained by existing learning objectives and examination formats. The challenge is therefore to identify, subject by subject, which parts of an activity are increasingly prediction-like and can be supported by AI, and where students still need to contribute context, subject understanding, and judgment. This reflects a central Hybrid Intelligence question: if AI can take over parts of the production, where should human effort be moved so that it still improves the process and the outcome?

The Hybrid Intelligence element lies in deliberately designing this distribution of work rather than simply adding AI to existing assignments. AI can generate possibilities and accelerate predictable parts of a task, while the student remains responsible for defining what matters, evaluating the alternatives, bringing in relevant context, and deciding which direction to pursue.

For example, a generic physics problem may be relatively easy for AI to solve. One example discussed at Ribe was to add contextual constraints, such as finding a solution using only materials available in a particular room. The task then requires knowledge that is not contained in the written exercise itself. AI can still support the problem-solving process, but the student has to connect its general suggestions to the actual situation. This was explicitly linked to the distinction between prediction and judgment.

FERC provides a practical structure for keeping this human contribution active throughout the process:

• Frame: Make the purpose, context, preferences, and constraints explicit. AI can support the process, but the human defines what matters.
• Explore: Ask AI for several different possibilities rather than accepting one fluent answer, creating something the student can compare.
• Refine: Criticize, combine, reject, and revise the alternatives. Through iteration, the direction should increasingly reflect the student’s own judgment.
• Commit: The final choice remains human. The learner commits only when they can stand behind the result.

The important HI principle is that human judgment is not added only at the end as approval of an AI-generated answer. It shapes the direction throughout the interaction. In the FERC process described at Ribe, the goal is for the first AI-generated suggestions to become progressively less visible as the learner compares, changes, and develops them into something they can take responsibility for.

This process can also become visible in assessment. Students could, for example, submit an AI conversation and identify the points where they changed the direction of the work, shifting assessment from only evaluating the final product toward also examining the learning process.

Another exercise format involves deliberately creating a chatbot that is wrong about a specific subject concept. Students then have to identify the error and explain why it is wrong. Rather than simply warning students that AI can make mistakes, the activity systematically trains them to challenge convincing AI output using their own subject knowledge.

AI-generated efficiency is therefore not treated only as time saving. If AI reduces the time spent on production, that time can instead be reinvested in activities such as choosing an angle, clarifying purpose, comparing alternatives, and improving originality and quality. This is where the combination of human and AI capabilities is intended to create more value than AI alone.

The Ribe case is still at an early stage, so the outcomes are envisioned rather than demonstrated.

The approach aims to develop students who are better able to formulate what they want, compare alternatives, challenge convincing AI output, and explain why they accept or reject suggestions. Rather than learning only how to produce an answer, students are trained to remain active participants in the reasoning and problem-solving process.

For teachers, the approach offers a way to embed AI literacy within individual subjects rather than treating it as a separate generic competence. It also opens a different approach to assessment, where students’ framing, revisions, choices, and influence on the process can become educationally relevant alongside the final product.

FERC should, however, be understood as a scaffold rather than a guarantee of meaningful human involvement. Its value depends on tasks that genuinely require students to contribute knowledge and judgment, rather than mechanically completing four steps. The discussion at Ribe also makes clear that FERC is particularly relevant for creative and complex problem-solving activities and should not necessarily be treated as the framework for every form of AI use.

More broadly, Ribe points toward a shift from asking how existing assignments can survive AI toward asking what students should learn when AI can perform parts of the work for them. The goal is not to preserve human activity for its own sake, but to design learning so that AI acceleration strengthens, rather than removes, the places where students need to understand context, exercise judgment, and take responsibility.