AI is eroding critical thinking in students faster than we can
The death of the first draft
Writing is not just a way to record thoughts; it is a way to form thoughts. The process of staring at a blank page, failing, rewriting, and refining is where the actual learning happens. By using an LLM agent to handle the initial drafting, students are skipping the most critical part of the intellectual process. They aren't learning how to structure an argument or how to bridge two disparate ideas; they are simply becoming "editors" of a machine's output. This doesn't make them more efficient; it makes them cognitively passive.
The dopamine loop of the "Correct Answer"
There is a psychological toll to this efficiency. Real learning is often frustrating and slow. However, AI provides an immediate, polished answer that mimics expertise without the effort. This creates a feedback loop where the student values the result over the process. When the "answer" is always available, the curiosity that drives a student to dive deep into a subject vanishes. Why spend an hour in a library or digging through a primary source when a prompt can summarize it into five bullet points?
How to build a sustainable AI workflow for learners
To stop this decline, we need a practical tutorial for how students should actually interact with these tools. Instead of using AI as an answer machine, it needs to be used as a Socratic tutor. Here is a better approach to prompt engineering for students:
1. The Constraint Prompt: Instead of asking for the answer, tell the AI to guide you toward it.
"I am struggling with this physics problem regarding kinematics. Do not give me the final answer. Instead, ask me a series of leading questions that help me figure out the correct formula to use."2. The Critique Loop: Use the AI to challenge your existing work rather than creating it from scratch.
"Here is a paragraph I wrote about the causes of the French Revolution. Please find three logical gaps in my argument and suggest primary sources I should look into to strengthen my claim."3. The Concept Stress-Test: Use the AI to verify your understanding through teaching.
"I think I understand the concept of quantum entanglement. I will explain it to you in my own words, and I want you to tell me exactly where my explanation is imprecise or wrong."If we don't shift from "AI-generated" to "AI-assisted," we are raising a generation that knows how to operate a tool but doesn't understand the logic behind the output. The goal should be cognitive empowerment, not intellectual outsourcing.