AI Tools in the Classroom: Curse or Blessing?

The chalk is still dusty on the blackboard, but a revolution has long since taken place on the screens of students and pupils. When we talk about education today, there is one topic we can no longer ignore: Generative Artificial Intelligence. Tools like ChatGPT and GitHub Copilot are no longer futuristic concepts; they are the new seatmates in our lecture halls and classrooms.

For us lecturers and those interested in education, an existential question arises: Do we damn this technology as the ultimate tool for plagiarism and cognitive laziness, or do we celebrate it as the greatest lever for individual learning since the invention of the printing press?

The Elephant in the Room: The Fear of Losing Control

In the staffroom break, one often hears the same concerns: "Are students even writing for themselves anymore?", "Does homework XY still have any value?" or "Are we currently unlearning how to think?". These fears are legitimate. When an AI writes a flawless analysis of Goethe's Faust or provides a complex mathematical derivation within seconds, our traditional assessment system begins to falter.

The impact is particularly noticeable in the field of computer science and programming. Copilot completes functions before the learner has even finished their first thought. But before we lapse into pessimism, we must ask ourselves: What is actually the goal of education? Is it the memorization of syntax, or is it the ability to solve complex problems methodically?

Learning to Program 2.0: From Coder to Architect

The use of AI is fundamentally changing the way we learn to program. Previously, beginners spent hours searching for frustrating syntax errors or laboriously leafing through documentation to find out how to read a file in Python. Today, the AI provides the solution immediately.

This has two sides:

• The motivation curve: The entry barrier drops drastically. Success experiences occur more quickly, which is crucial for young learners in particular to stay on the ball. The frustration of the "semicolon error" gives way to pride in the functioning project.
• The level of abstraction: We are moving away from "writing lines" towards "designing systems". The learner becomes an architect earlier, needing to understand how components interact instead of getting lost in the syntax jungle.

The Dilemma of "Boilerplate Code"

There is a heated debate about whether one still needs to master the basics when AI writes the so-called boilerplate code (standardized code segments). Why should a student learn how to write a sorting algorithm by hand if list.sort() or an AI-generated function does it faster?

The answer is simple: Because you cannot build a stable house without a foundation.

When AI writes the code, the training of the "logical muscle" is omitted. Anyone who has never learned how data flows in memory or what the time complexity of an algorithm means cannot judge the quality of the AI output. We run the risk of raising a generation of "StackOverflow copy-pasters on steroids" who can indeed click programs together but are completely lost during the first serious troubleshooting (debugging).

A pilot must be able to land the plane even if the autopilot fails. In software development, AI is our autopilot, not our replacement.

My Method: First Abstinence, Then Symbiosis

In my teaching, I follow a clear strategy that I recommend to both colleagues and learners. It can be divided into two phases:

Phase 1: The AI-Free Zone (The Basics)

At the beginning, I consistently recommend refraining from using AI tools. Why? Because the brain needs resistance to grow. In the first few weeks, the fingers must feel the code. One must experience the frustration when a program doesn't compile, and the euphoria when one has found the error themselves after searching for 30 minutes. These "painful" learning experiences anchor knowledge deep in the long-term memory. Anyone who has never debugged a loop "by hand" will never develop a feeling for runtime errors.

Phase 2: The "Understanding Contract" (Productive Use)

As soon as the basics are solid, I encourage the use of AI. But with an iron rule: Never have a line of code generated that you do not understand.

When Copilot makes a suggestion, the learner has two options:

1. Researching: One reads up on what the specific commands or parameters mean.
2. Simplifying: One asks the AI to solve the problem in a way that corresponds to one's current level of knowledge (e.g., "Write this without lambda functions").

AI as a Tutor, Not a Replacement

Let's not imagine ChatGPT as a homework machine, but as a personal tutor who is available 24/7. The potential for personalized learning is gigantic:

• The Explain-Mode: Instead of "Write me the code," learners should ask: "Explain the concept of recursion to me using an example with pancakes."
• Code Review: "Here is my code. It works, but is it efficient? Which best practices have I overlooked?"
• Troubleshooting: "I'm getting an IndexOutOfBoundsError. Can you help me understand why my loop counter is running too far?"

In this scenario, the cognitive load remains with the learner. The AI serves as a mirror and mentor that accelerates the learning process instead of replacing it.

The Psychological Trap: The Illusion of Competence

A major risk of AI use is the so-called "illusion of competence." When we read and understand an AI solution, we often falsely believe we could have generated it ourselves. In psychology, it is known that recognition is not the same as recall.

In the classroom, we must therefore create formats that expose this phenomenon. This means: less assessment of submitted files, more focus on live coding, oral explanations, and transfer tasks. We must check whether the student is the "master of the code" or just the "curator of the AI output."

New Requirements for Teachers: From Knowledge Provider to Curator

For us lecturers, the role is changing radically. We are no longer the only source of knowledge in the room. Our task is now:

• Prompt Engineering as a cultural technique: We must teach how to ask the AI the right questions in order to receive high-quality answers.
• Critical Evaluation: We must show that the AI hallucinates, suggests outdated libraries, or incorporates security vulnerabilities.
• Ethics and Copyright: Where does the AI's data come from? Who owns the code? These questions must become part of the curriculum.

Conclusion: The Role of Teachers is Changing

Is AI in the classroom therefore a curse or a blessing? The answer lies in our hands. If we continue as before and try to ban AI, it will become a curse, a tool for cheating and superficiality. An arms race between AI texts and AI detectors would emerge that no one can win.

If we actively integrate AI, however, and shift the focus from creation to evaluation and understanding, it is a gigantic blessing. We no longer just teach how to program. We teach how to collaborate with intelligent systems without losing our own power of judgment.

Mastering the basics remains essential, not despite, but precisely because of AI. Because at the end of the day, the AI is only as smart as the person operating it. A bad cook will not conjure up a Michelin-star menu with a professional Thermomix; they will only produce mediocre food faster.

To my colleagues: Do not be afraid of the tool. Be the guardrails that ensure the shortcut does not become a wrong turn. Change your forms of assessment, away from the result, towards the process.

To the learners: Use the power of AI, but remain the boss in the ring. Your brain is the operating system; ChatGPT is just a plugin. Anyone who takes the shortcut may arrive faster, but they miss the view and the training for the truly steep mountains that are yet to come.