The Ethics of Using AI in Writing

Commentary

The etiquette and ethics—they are related—of any new technology take some time to settle. That’s because new techniques need time to integrate into the human experience and be tested against moral intuition, practicality, and trade-offs of costs and benefits. So it is with artificial intelligence and the literary arts.

I’ve noticed two early intuitions and concluded that both are wrong. Some people say that writing with AI is perfectly fine in every area of life. Put in your prompts and your thesis and let the engines write a book for you. There is no issue of copyright, so you can legally and plausibly claim that it is yours.

To my mind, this works for routine correspondence or summaries or purely functional prose. It does not work for literature as such. That people who do this are inclined to hide the fake tells you all you need to know. If there is nothing disreputable about this, why not announce it on the cover page? Just say that you didn’t have time for research or enough knowledge not to make mistakes so you let the machine do your thinking for you.

People do not admit this because it is unseemly and readers would revolt. We sign things we write for a reason. It is because we write them that we sign them. To be sure, there has always been a gray area called ghostwriting. I’ve done plenty of it in my lifetime. The idea is that if a person is too busy and can pay for it, it perhaps makes sense that he hire someone to do his writing.

I’m strongly of the belief that the primary authors who are not authors should tell readers of the existence of a ghostwriter. It can be stated as “by Bob and David.” Or Bob can write in the introduction of his gratitude to David who provided extensive research and drafting help. All we need is something to identify and credit the person. Speaking for myself, I should have insisted on this far more than I have in the past.

So isn’t writing a book or term paper with AI a bit like using a ghostwriter? There are certain commonalities, but the rule should still pertain. You have to say that AI wrote this, or specify what parts in particular or how you used AI in its preparation. You can say that AI did the first draft and you adapted the text to your purposes. That’s the respectable thing to do.

On the other hand, this is useless if the point of the writing project is to show what you know. Students should absolutely be banned from relying entirely on AI to write. If they are caught, they should be failed from the course. I’m profoundly aware that these days, that might mean that three-quarters of students or more would fail. Given that, they should be given advance warnings that one in 10 students will be randomly chosen for an oral exam on the contents of papers. That should help discourage this practice.

Publishers are currently in chaos over this problem. They are getting scads of submissions that are written by AI. I can testify to that as a publisher. The great problem is that writers never admit it and feign offense when you ask. It’s hard to prove, but publishers are correct to insist on this unless they are publishing purely technical manuals or summaries of known information.

A particular problem has hit publishers who farm self-publishing platforms for popular books and offer money to take them on for a fresh publication. Many have later discovered that the original author did not write the book, so now they are stuck. This is a huge issue right now and everyone is dealing with it.

All these problems could give rise to a kind of prohibitionist puritanism about the matter. Never use AI under any circumstances! I don’t find this position viable either. AI is enormously useful for researching, fact-checking, getting up to speed on a topic, or seeing what others have said about something. I use it constantly simply because it is far more efficient than searching the web or digging through stacks.

I was shopping at a flea market the other day and found multiple copies of “Bartlett’s Familiar Quotations.” This is the longest-lived and most widely distributed American reference collection of quotations. First published in 1855, it remains in print and is currently in its 19th edition—facts I learned from AI, which in turn got them from Wikipedia.

Why was it so successful? Because it allowed readers, speakers, toastmasters, or writers access to the best of the best writing without having to read the whole book by the original author. In other words, “Bartlett’s” was nothing but an early version of AI. It gave you what you needed without having to dig it up.

It’s no more cheating than using an encyclopedia is cheating. We have been searching for methods to gather information and present it in an accessible fashion since St. Isidore of Seville first put his monks to work writing an encyclopedia of all known knowledge. This was in the 6th century! It’s just part of literacy that we do this.

In this way, the rise of AI is nothing unusual, just the latest and greatest tool for doing what we have always done. Of course, it does a bit more. “Bartlett’s” did not actually write your books or reports for you. And that’s the problem. AI can confer valor on you that is entirely unmerited. That’s a serious problem.

From the perspective of the author’s reputation, you are risking everything by using AI for anything other than how you might otherwise use “Bartlett’s” or an encyclopedia or dictionary. If you are discovered using it to write your material—to ascribe to you knowledge and work that is not yours—you can lose your credibility entirely.

It’s like carrying a fake Louis Vuitton bag: It works until it does not. If everyone you know is only standing at a distance, you can get away with it. But if one person happens on the elevator with you who knows what’s what, you will sink in his eyes. Maybe it is not immoral, but it is deeply dangerous.

Given all these considerations, I propose the following first draft of some rules of thumb. If you are using it to write routine correspondence or purely functional matters for work, there is nothing wrong with it and much right about it. Your authorship here does not really matter and AI can do a spectacular job. Same with many forms of legal documents and so on.

If you are using it as you might use a book of familiar quotations—a shortcut to information that would otherwise take you a day to get—it’s great and useful. Nothing wrong here.

If you are using it as a writer to get ideas or sharpen language, or point you to other research about which you might not know, this also seems perfectly fine and defensible. It is improving the work that you are doing.

The same is true for people who draft an article or book and use AI to sharpen the final product. This can be helpful, although I would strongly warn you to be judicious here. AI has no real judgment, tends to make sharp points dull, and forces all content into a kind of boring mediocrity. Be very careful here. It can help but cannot be wholly trusted.

The no-go territory is for people who are expected to exercise judgment and creativity or share their expertise to rely on AI to draft most of the content. Something about this strikes me as wrong, and not necessarily for the reason that it feels like plagiarism. It is because readers rightly expect something else and you have given them a pastiche instead. This is a kind of exercise in fraud. It is not illegal, but it skirts very close to being unethical.

So there we go, my first draft of an etiquette and ethics of using AI to write. Expect this to evolve over time as we discover more about this strange and unprecedented visitor in our midst. Nothing about writing will ever be the same.

Views expressed in this article are the opinions of the author and do not necessarily reflect the views of The Epoch Times.