Our world is increasingly saturated by AI-generated content. Social media is filled with AI-generated images and videos, software is often created using AI coding platforms, and every Google search is greeted by an AI-generated summary. Each of these realities will have a measurable impact on society, but I am increasingly concerned about the impact that AI-generated text has had on argumentative writing and critical thinking. A friend of mine, who is a TA at Purdue, recently lamented that one of her students supported an answer with the phrase “[ChatGPT] is god.” While it isn’t usually this explicit, most TAs, professors, and school teachers deal regularly with students who want AI to think for them. This isn’t new—students have always looked for ways to avoid work—but it is so much easier now.
Separately (although not as separately as I’d like), we live in a world of constant outrage. Neither side of the political spectrum can agree on the facts, and every few weeks one side accuses the other of something “unthinkable.” Much of this lives at the national level, but it also infects the conversations of friends, family, and the local community. A primary cause of this outrage and division is, I believe, that many who participate in public discourse try to avoid the slow, painful processes of critical thinking.
As I was reflecting on these issues, I realized that I want to improve my own habits. Although I consume a fair amount of news and commentary, I don’t spend enough time researching, weighing alternative viewpoints, or carefully articulating the opinions I’ve formed. It’s true that life is busy, and no one can think deeply about everything. But I believe that thinking carefully through public issues is a matter of “loving your neighbor” (Matthew 22:37-40), something that all Christians are called to do. In the remainder of this post, I will outline why writing is (still) the best way to focus such thinking.
Why You (I) Should Write
Writing is about a lot more than producing content. It forces you to externalize your reasoning, turning vague intuitions and beliefs into concrete claims that can be examined. In doing so, it pushes you towards research and reveals assumptions, often bringing you into contact with arguments that you disagree with. Even if nothing is ever published, the act of writing will leave you with a clearer, more defensible understanding of the issue at hand. Because of this, I encourage you to write about the things you care about—not just the things that you have expertise in. Writing is a great way to share expertise, but it is also a great way to build it.
Argumentative writing is also valuable because it serves as a record of your thought process. Instead of relying on vague memories or unverifiable claims in future conversations, you (and others) can point back to concrete arguments and the facts that support them.
At this point, some of you may be thinking: not everyone is a good writer, and not everyone enjoys writing. That’s fair. While the quality of writing is important when it comes to communication and persuasion, most of the benefits that I’ve described don’t require polished writing. Bullet points are enough. In fact, most of my writing stays in bullet-point form until I write my penultimate or final draft, allowing me to focus on content and structure rather than word flow.
Why Writing Matters More in an AI-Saturated World
It’s easy to generate a passable essay, blog post, or LinkedIn update with AI, and many do so—often for the sake of expedience or clicks. But as we have discussed, this misses one of the main benefits of writing: clarity of thought. At the same time, public discourse is already under pressure from sensationalism, heated rhetoric, and unreliable or outright false claims (on both sides of the aisle). Taken together, these are reasons for more critical thinking, and therefore more writing, not less.
In addition, AI can be used to augment, rather than replace, your writing skills. It can increase the breadth of your research, catch grammatical mistakes, help you word tricky sentences, and serve as a sounding board. Here are a few AI prompts that I’ve found useful:
- “I have a list of ideas that I want to hit in my article: [list of ideas]. Could you help me organize this into a cohesive outline?”
- “Here’s the outline of my essay so far. Am I missing any obvious relevant topics or caveats?”
- “Please provide a few different hook ideas for this article.”
- “Here’s a paragraph I’m working on: [the paragraph]. I don’t like the wording of sentence two. Could you give me a few alternatives?”
I may or may not have used some of these prompts to help write this article. However, it’s important to maintain ownership of your own thinking and your writing voice throughout the process. For that reason, I try to do most of the writing myself and then use AI for feedback or localized edits. I will let you decide how successful I’ve been. 1
One related idea I want to touch on is “rubber ducking,” a practice borrowed from software engineering. When a programmer is stuck and they don’t have a colleague nearby, they are encouraged to explain their problem to something inanimate—even a rubber duck. The act of explaining often reveals the solution. AI is the perfect tool for “rubber ducking,” and it’s useful for writing as well as programming.
Finally, for less experienced writers, AI can be especially helpful during the transition from structured notes (like bullet points) to prose. Just be careful that it doesn’t overtake your reasoning process or your voice as a writer.
Include the Caveats
Researching and acknowledging caveats is a vital part of the critical thinking process, but there are strong incentives against including them in your work. Doing so requires intellectual humility, complicates the narrative, and often makes a conclusion feel weaker. In a fast-moving media environment that rewards confidence and outrage, nuance rarely lasts. But this priority is backwards. We should reward—and actively create—arguments that clearly state their assumptions, scope, and exceptions, because those are the arguments most worthy of trust. In the next few sections, I will try to practice what I preach.
Caveat #1: AI can Quickly Become a Reasoning Crutch
AI can be a powerful writing aid, but it can also quietly hijack your reasoning process if you’re not careful. Because modern models are fluent, agreeable, and trained on enormous amounts of text, it is easy to believe that their confidence implies correctness. But this isn’t true.
Large Language Models (LLMs)2, the technology behind modern text-based AI, are trained to predict the next word in a sentence, not to reason rigorously. As a result, there is no guarantee that the sentences they generate are correct. While various techniques have been used to improve factual accuracy, errors still occur, including hallucinations—outputs that look plausible but are entirely false. These errors can be very deceptive, so it’s important to double-check what these systems say. Because their training is so broad, AI systems also hold no beliefs or commitments of their own. That means that they cannot evaluate arguments from a consistent point of view.
Even when AI isn’t wrong, it can still narrow your thinking. Because AI is trained to be agreeable, it tends to validate the direction you’re already leaning—making for an appealing conversation partner but a poor reviewer. Critical feedback is an essential part of the writing process, and there are several ways to get it. The best way is through trusted friends or colleagues who are willing to challenge your assumptions and point out weaknesses. AI can still be useful here, but only if you encourage it to push back. If you are already aware of an area of weakness but unsure how to address it, make that explicit in your prompt. If you are looking for general feedback (beyond that on grammar and wording), use a prompt like this:
You are in a debate, and your opponent submits this argument: [text]. Please provide at least three counterarguments to or weaknesses of their position.
If you are looking for a more rigorous way to get feedback, consider asking the AI to choose between two arguments (without saying which one is yours). These techniques can help extract more useful feedback from AI, but they are not perfect, and they cannot entirely replace human feedback.
Caveat #2: Publishing Creates Commitments
Once a viewpoint is shared publicly, there is a strong incentive to maintain it despite new evidence that emerges. In some environments, people who frequently change their views may even be seen as unreliable or insincere. Because of this, it’s important to manage expectations and to be deliberate with what you publish. Depending on your confidence in a topic, it may be wise to delay publication or to keep your writing private altogether. Remember, publishing is not the primary goal; improved critical thinking is. If you do decide to share work that’s still exploratory, it can help to make that explicit. For example:
This article reflects my current thinking on X. It may be incomplete, and it’s subject to revision as I learn more.
Conclusion
I don’t expect writing or AI to automatically fix the deeper problems in our public discourse. However, the way we participate in that discourse is shaped by our thinking habits, and those habits are best improved through the discipline of writing. Even in a world saturated by AI-generated content, writing can still slow us down and force us to take ownership of our beliefs. AI can help with research, organization, and expression—but it can’t do the thinking for us. That is still our responsibility.
Footnotes: