AI in Class: Why We Need to Teach Students to Use—Not Ban—AI Tools
The debate around artificial intelligence in education isn’t about stopping its employ, but about learning to use it thoughtfully. That’s the perspective of Maximilian Milovidov, a freshman at Columbia University, who shared his experience in an “AI-first” writing course – one where the technology isn’t banned, but actively explored – with NPR this week. His insights come as universities grapple with how to address the rapidly evolving landscape of AI tools and their impact on student learning and academic integrity.
Milovidov’s course, described as a “living thought experiment,” directly confronts the anxieties surrounding AI’s potential to undermine critical thinking. Instead of treating AI as a shortcut to be avoided, the class encourages students to experiment with large language models (LLMs), critique their outputs, and understand their limitations. This approach, he argues, is more realistic – and ultimately more valuable – than simply prohibiting AI use, a strategy he likens to attempting to halt the spread of the printing press.
The Cognitive Offloading Concern
A central concern surrounding generative AI is the idea of “cognitive offloading,” where reliance on machines weakens a person’s own understanding. Milovidov acknowledges the validity of this worry, but suggests it only materializes when AI is treated as an infallible source of truth. When students actively engage with AI, questioning its suggestions and analyzing its reasoning, it transforms into a collaborative learning tool. He describes it as a readily available study partner, offering benefits and drawbacks, and even serving as a substitute for a teaching assistant or tutor.
The course structure itself is designed to foster this critical engagement. Students don’t simply submit assignments generated by AI; they bring their own ideas and drafts, then use chatbots to identify gaps in their reasoning or explore alternative perspectives. As Milovidov explains, it’s akin to seeking feedback from a friend – a valuable input, but not a replacement for original thought. This mirrors findings from research, including a study suggesting that moderate AI assistance can actually improve human cognitive performance.
Leveling the Academic Playing Field
Beyond fostering critical thinking, AI tools also have the potential to democratize access to educational resources. For students who lack the financial means for private tutoring, chatbots can provide personalized practice questions, mock exams, and feedback on writing. These tools can act as virtual teaching assistants, offering support and guidance without the associated cost. A 2025 Harvard University study found that students using an AI tutor experienced learning gains more than double those in traditional classrooms, and reported increased engagement.
However, Milovidov emphasizes the importance of recognizing the inherent biases within these systems. LLMs are trained on vast datasets, often dominated by Western-centric perspectives. He found that feeding drafts into a chatbot didn’t automatically produce polished prose; instead, it often amplified existing weaknesses in his writing, highlighting the demand for careful evaluation and revision. This experience underscored the value of his own imperfect work, teaching him more about his writing process than any traditional assignment.
The “Friend Test” and Beyond
The professor of Milovidov’s “Writing AI” course introduced a concept called the “friend test” – a way to gauge appropriate AI use. The idea is that you would ask a friend for feedback on your work, but you wouldn’t ask them to write it for you. This analogy highlights the importance of maintaining ownership of one’s ideas and using AI as a tool for enhancement, not replacement.
This approach is particularly relevant as AI becomes increasingly integrated into the workforce. The World Economic Forum suggests that AI may automate many entry-level jobs, making it crucial for students to develop the skills to collaborate effectively with these technologies. Education, needs to evolve to prepare students not just for the jobs of today, but for the jobs of tomorrow – jobs that will inevitably involve working alongside AI.
Addressing “AI Shame” and Moving Forward
Milovidov points to a concerning trend on college campuses: “AI shame,” where students fear punishment for using AI tools, driving them to use the technology in secret. His course provides a safe space to openly discuss these issues, fostering a more nuanced understanding of AI’s potential and limitations. This open dialogue is essential for developing responsible AI practices and ensuring that students are equipped to navigate this evolving landscape.
The conversation extends beyond the classroom. New York City schools are already exploring AI literacy programs, recognizing the need to prepare students for a future shaped by artificial intelligence. The key, as Milovidov’s experience demonstrates, is not to fear AI, but to embrace it as a tool for learning, growth, and innovation – and to teach students how to wield it responsibly and critically.
Looking ahead, the ongoing development and refinement of AI tools will necessitate continuous evaluation and adaptation of educational practices. Universities and educators will need to stay informed about the latest advancements, assess their impact on student learning, and adjust their approaches accordingly. This is not a one-time fix, but an ongoing process of learning and adaptation.