Long before the printing press or ChatGPT, Plato was already worried about what new technologies might do to us. In the Phaedrus, Socrates explains why he never wrote anything down. He tells us that when writing was invented, its inventor boasted that it would make people wiser and help them remember. But in his view, it would do the opposite. It would make people forgetful, and give only “the appearance of wisdom, not its reality.” Written words are prone to misinterpretation; they are indifferent to their audience. They are a kind of “dead speech.”

As a philosopher of science and AI who also teaches ancient Greek philosophy, I’m struck by how closely our current situation mirrors the concerns of the ancients. LLMs like ChatGPT are new technologies that promise to make us wiser, to read, write, and think on our behalf. But like writing in Plato’s time, they threaten to erode the skills they imitate. In a recent Templeton Ideas essay, Richard Lopez articulated three perceived harms: AI gives us quick answers but robs us of understanding, it displaces real human interaction, and it stunts personal growth. Many others are sounding the alarm and calling for tech-free spaces.

But would Plato agree? The irony is that we know Socrates’ warning about writing only because Plato wrote it down. Did he forget his own criticism, or decide the trade-off was worth it?

When I teach the Phaedrus, I follow my former philosophy professor at Boston University, David Roochnik, who argues that Plato was well aware of the force of his own critique of writing, but believed his work could overcome its limits. Plato wrote dialogues because they could come alive in the reader’s mind. They invite reflection, questioning, and response. Such “living speech” enhances the reader’s faculties and even discriminates among readers, refusing some and rewarding others. Socrates, for example, often gives arguments that are mediocre. This is deliberate—some will be satisfied and move on, while the more philosophically minded will object and dig deeper. In that sense, every Platonic dialogue is two conversations: the one Plato wrote, and the one that unfolds between Plato and the reader.

Crisis in the Classroom

How might we do that? Reflecting on Plato’s model of education in The Republic as a painful and difficult “turning of the soul,” the arrival of AI that can write and think for us is oddly invigorating.  It invites us to reconsider writing, not just as a tool of communication, but as an art and practice of thought itself. 

LLM technology and its wide accessibility push us to ask what understanding, creativity, and authorship truly are: Can I understand something if I can’t say it in my own words, or if AI can say it better? If the idea is mine but the text was created or polished with AI, am I still the author? There is no doubt that unreflective LLM use might lull us into complacency. But the questions raised by these systems are vital. For Plato, such intellectual provocation is the beginning of wisdom. Like the prisoners in the allegory of the cave, turning from shadows toward the light, we are being forced to turn from taking writing for granted to questioning its value in our lives.

That turning can begin anywhere. We might build AI systems that optimize for truthfulness, are explicit about their reasoning, and reveal their biases. In everyday life, we could foster forms of literacy that teach us how to understand the new algorithms shaping our world. And in the classroom, AI can become a kind of living speech, a teacher, editor, and interlocutor.

As a teacher, for instance, I can encourage my students to treat LLMs like a Socratic partner: to challenge and test their views, to help them discover what they don’t know and expose blind spots, and to use AI in what Agnes Callard calls an “exploratory battle” to discover “inquisitive opportunities.” And since even the best teachers have limited time with students, we can show them how to use an LLM as a personal line editor—not to replace their writing, but to offer sentence-by-sentence suggestions that help them become better writers. Of course, such techniques are no substitute for human interaction. But they do address Lopez’s earlier concerns about substituting quick answers for real understanding and stunting intellectual growth.

Still, beneath these practical strategies lies a deeper issue. Across academia, from lecture halls to the editorial boards and university committees on which I serve, the same questions resonate: what should we still teach our students, and what standards should guide our own writing? What, exactly, is writing for?

What Writing is For

Familiar answers come readily to mind. Writing matters because it’s how we learn and nurture reflection, distinguish between strong and weak prose, bring human individuality, culture, and taste into play, practice the aesthetic craft, and so on. Personally, I often don’t know what I think about a topic before I sit down to write about it. But perhaps that’s a far cry from claiming writing is indispensable for thought.

There are different forms of thinking and knowing. My own view, echoed by friends and colleagues, is that writing plays a special role in cultivating philosophical thinking, a kind of “dialectical literacy” that we all need.

To see what I mean by dialectical literacy, let’s look at what good reasoning is all about. In some cases, all that matters is logical form. Take the syllogism: “All Greeks are mortal. Socrates was a Greek. So Socrates was mortal.” We could substitute any content, as long as the structure holds. 

But most real arguments live in messier territory. These are what philosophers call “inductive” or “ampliative” arguments. They hinge on word choice, background facts, and specific content—what is actually said and whether it’s true, not just the logical form. That’s reading-and-writing work all the way down.

Reinventing Writing

Consider, then, how LLMs increasingly present chains of reasoning, not just answers. When you are dealing with real-world reasoning, evaluating those chains can’t be cleanly separated from reading and writing. And yet, this is precisely where LLMs can shine if used in the right way. You can literally interrogate them, ask for reasons, demand sources, and evaluate their arguments. It’s much more like a Platonic dialogue than static writing that cannot answer back. Recall, for instance, Socrates’ worry about how writing is a one-way communication that doesn’t adapt to its readership. Nigel Warburton argues that with digital technology such as online chats, written communication is now interactive, thereby overcoming these limitations. Even more so, I’d add, in the era of LLMs.

We might even say that, like Socrates in the Phaedo, LLMs let us “take refuge in discussions and investigate the truth of things by the means of words.” In other words, today’s AI systems make the dialectical practice cherished by philosophers readily available. This is no substitute for Socratic conversation in the classroom, but it does suggest teaching, writing, and doing research in a way that can complement this practice. LLMs can function as real empowerment. The challenge is to use them in ways that deepen, rather than diminish, our thinking.

For example, just as LLMs can write our sentences, and AI apps like Suno can compose songs, before long, a robot may even play all our instruments. But a musician doesn’t stop playing just because a machine can make the notes. The point is the playing. So too with writing: its worth isn’t just in the product, but in the process, in what it draws out of us.

So, once more, why does writing matter? The suggestion here is that it’s essential for the kind of philosophical thinking and dialectical literacy that I described—in short, good reasoning—that it’s an art with its own intrinsic value. And how can LLMs function as empowerments, rather than substitutes for human thinking? We can use them to broaden our research reach, refine our writing through personal editing, and engage in Socratic dialogue—a sustained, thoughtful exchange of prompts and responses that becomes a new genre of writing, an “AI dialogue” akin to Platonic dialogues.

Nonetheless, if we look to this essay for the answer or to anything specific that Socrates says, we miss the point. Plato wrote dialogues to draw readers into thinking for themselves. Likewise, the task is ours, alone and in community, to decide why writing matters to us and where it doesn’t. That’s harder. And it’s work we can’t outsource.

Elay Shech is a Professor at Auburn University who writes about science and philosophy. His books include Idealizations in Physics and The Metaphysics of Color, and his articles appear in outlets such as The New York Times and The Washington Post.