AI as a Mirror: Turning the Chatbot Into a Thinking Partner | Miskola

In the educational landscape today, a rather simplistic view of Artificial Intelligence (AI) dominates. We are accustomed to hearing about it as a time-saving tool—something that writes an email, generates an image, or quickly summarizes a long text. In the “3 Levels of AI Usage” concept that I developed, this is merely Level 1 (Instrument). Although it improves efficiency, it often “bypasses” the thinking process, allowing students to arrive at a result effortlessly.

However, for true, deep learning to take place, we must climb higher. The focus of today’s article is Level 2: AI as a Mirror. This is the critical “turning point” where technology ceases to be merely a “doer” and becomes a facilitator of “thinkers.”

What is Level 2 (AI as a Mirror / Understanding)?

• At Level 2, AI is no longer an “executor” that does the work for the student. It becomes a cognitive opponent and scaffolding in the thinking process.

At this level, we intentionally use AI to see the gaps in our thinking rather than to obtain a ready-made result. It operates similarly to the Socratic method—the AI does not provide answers, but instead asks questions, compelling the student to clarify, substantiate, and revise their assumptions.

The student is no longer a passive recipient of information, but an active manager of the process. Here, a new digital competency emerges: text-mastery. This is the skill of directing content creation by delegating technical execution while retaining complete control over meaning and logic. If a student does not know the subject matter, they will not be able to form a quality dialogue with the “mirror” because they will not be able to distinguish hallucinations from facts.

Why is this critically important right now?

We live in an era characterized by the “Synthetic Content Loop”—the information environment is rapidly filling with AI-generated content. If students only learn Level 1 (generation), they become part of the problem. Level 2 trains “immunity”—the ability to verify, question, and improve what AI produces rather than trusting it blindly.

Practical Examples in Lessons

Here is how Level 2 can be integrated into various subjects, with clearly defined roles, task workflows, and assessment methods.

1. History: “What If…” (Counter-factual Simulation)

Traditionally, students learn history as a sequence of facts. This method turns the lesson into a strategic simulation where the student must understand cause-and-effect relationships and defend decisions in a dynamic environment.

• Task: Understand the complexity of the Cuban Missile Crisis, the tension of decision-making, and alternative historical scenarios.
• Prompt: “You are John F. Kennedy at the height of the crisis (October 1962). I am your military advisor (General Curtis LeMay). I am advising you to immediately launch air strikes on Cuban missile bases. Criticize my plan based on the geopolitical risks of that moment (especially the threat of nuclear war), and ask me for an alternative, diplomatic solution that does not look like surrender.”
• Roles:
  • Student: Strategist and arguer. They must “sell” their idea to the AI, which assumes the role of a stubborn or cautious leader. The student must know the historical context to be able to respond to “Kennedy’s” (AI) concerns.
  • Teacher: Scenario architect. The teacher sets the historical constraints and analyzes the flow of the conversation, rather than just the final outcome.
• Assessment Criteria:
  • Argumentation precision: Did the student use real historical facts and terms in the conversation (e.g., “blockade,” “quarantine,” “DEFCON”)?
  • Reaction to criticism: How did the student adapt their strategy when the AI pointed out risks?
  • Process reflection: The conversation log (chat history) is evaluated.

2. Literature: Style and Tone Analysis

Often, students use AI to “write a poem.” That is Level 1. In Level 2, we do not ask AI to write the creative work, but rather ask it to become a literary critic that helps polish what the student has created themselves.

• Task: Understand the epiphany genre and Imants Ziedonis’s specific stylistics.
• Prompt: “Here is an epiphany I wrote on the topic ‘Autumn.’ Please analyze it as a strict literary critic. Do not fix mistakes for me, but point out areas where I have lost the paradox and philosophical depth characteristic of Ziedonis. Ask me 3 questions that would help me improve the text and make it more metaphorical.”
• Roles:
  • Student: Author. They create the original text and then perform editing based on the AI’s analysis. The student practices text-mastery—the ability to evaluate whether the AI’s suggestion genuinely improves the work or makes it banal.
  • Teacher: Mentor. Helps the student understand the critique provided by the AI and evaluate whether it is justified.
• Assessment Criteria:
  • Originality: Is the “draft” created by the student themselves?
  • Iteration: How did the text change after AI feedback? (The student must submit the version before and after).
  • Metacognition: The student’s commentary—which AI suggestion they agreed with and which they did not, and why.

3. Physics / Natural Sciences: “The Feynman Technique” (Teaching the AI)

Research shows that the best way to learn something is to teach it to someone else. Here, the “someone else” is an AI that plays the role of a novice or a skeptic.

• Task: Test and solidify your understanding of Newton’s 3rd law (action and reaction).
• Prompt: “Act like a 10-year-old student who believes that if a large truck crashes into a small car, the truck hits harder, but the car does not hit back. I will try to explain the flaw in your thinking to you. Ask me skeptical questions if my explanation is not logical or is too complicated. Do not agree with me until I have proven it with a simple example.”
• Roles:
  • Student: Teacher/Expert. They must find the right words and analogies. They must overcome the “curse of knowledge” and explain the complex simply.
  • Teacher: Observer. Evaluates whether the student can simplify a complex concept without losing scientific accuracy.
• Assessment Criteria:
  • Clarity: Was the student able to explain the concept without “scientific jargon” (understanding concepts vs. rote memorization of definitions)?
  • Dispelling misconceptions: Did the student notice and correct the errors and counterarguments simulated by the AI?

4. Mathematics: The Error Hunter

This method combats “copying” from the AI. Instead, the AI helps find the error in the student’s own solution, promoting self-regulated learning.

• Task: Find the error in the solution of a quadratic equation or trigonometry problem.
• Prompt: “I have solved this problem [insert solution], but the answer does not match the back of the book. Do not tell me the correct answer and do not solve it for me! Instead, tell me only in which step (beginning, middle, or end) I made a logic or calculation error, and give a small hint about what kind of error it is.”
• Roles:
  • Student: Detective. They must review their own work based on the clues. This requires high concentration and analytical abilities.
  • Teacher: Process facilitator. Ensures that students do not ask the AI to “solve it for me.”
• Assessment Criteria:
  • Self-regulation: The ability to independently find and correct the error after receiving a hint.
  • Reflection: The ability to explain to the class or teacher why the error occurred (carelessness, lack of formula knowledge, or misunderstanding).

Implementation Challenges and Solutions

Transitioning to Level 2 is not simple because it requires changing habits.

• Student resistance: Students often desire “quick dopamine”—the ready-made answer. Level 2 requires effort. Solution: Explain that AI is a training partner, not a “cheat sheet.”
• Assessment complexity: How to assess a conversation? Solution: Introduce “Process Journals” or Thinking Traces—students submit not only the final essay, but also screenshots of the most crucial dialogue moments where the development of their thought process is visible.

Conclusion: From Result to Process

By working at Level 2, we mitigate the risks posed by blind reliance on technology. We transform the educational paradigm:

• We do not evaluate the final product (because any model can generate that).
• We evaluate the quality of the dialogue—how the student arrived at the result in collaboration with the AI.
• We evaluate the quality of the questions—at Level 2, the question asked by the student shows their level of knowledge better than their answer.

This is a step toward Level 3 (AI as a Metacognitive Partner), where technology helps us not only learn about the world but also understand our own thinking.

Are you ready in your classroom to transition from “answer generation” to “thought mirroring”? Choose one example and try it tomorrow!