Prof. Kwame Appiah, the latest (and arguably the most ethical) in a long line of proprietors of the Sunday Times “The Ethicist” column has long provided me with fodder for ethics posts, often critical ones. Appiah might finally have jumped the shark however: I don’t know that I can continue to regard him highly after his collection of rationalizations employed to answer a TV screenwriter’s query about whether it is ethical for him to use generative artificial intelligence bots to write screenplays he is paid for and puts his name on. “So what ethical line would I be crossing? Would it be plagiarism? Theft? Misrepresentation?” the inquirer asks.
My answer is simple: using AI as inspiration or even a model isn’t any of those things, just like a screenwriter reading other writers and watching movies with deft screenplays is legitimate source material inspiration. Most artists “steal” from other sources, altering their models sufficiently to pass as original, and rightly so. There’s a line where imitation and inspiration becomes theft and plagiarism—like when the Beach Boys lifted Chuck Berry’s “Sweet Little Sixteen” almost note for note—but short of that line is just art as usual. At least, however, the artist is the one doing the adapting and ethical tight-rope walking, not a machine. I feel the same way about authors using AI to write their products exactly the way I feel about AI judging: the human being, his or her experience, quirks, patterns, world view and more is why a screenwriter has the job. Using a bot, and I don’t care how it has been programmed, to produce full scenes and dialogues is lazy and dishonest. Individuality is a writer’s, indeed any artist’s, most valuable commodity.
In short: what the screenwriter is proposing is unethical. Now here’s “The Ethicist’s” take. I’m going to post it all, and leave it to you to name the rationalizations, which you can find here.
“We’re done here.” Some years ago, sleepless in a hotel room, I flicked through TV channels and landed on three or four shows in which someone was making that declaration, maybe thunderously, maybe in an ominous hush. “We have nothing more to discuss.” “This conversation is over!” Do people really talk like that? Possibly, if they’ve watched enough television.
“My point is that a good deal of scripted TV has long felt pretty algorithmic, an ecosystem of heavily recycled tropes. In a sitcom, the person others are discussing pipes up with “I’m right here!” After a meeting goes off the rails, someone must deadpan, “That went well.” In a drama, a furious character must sweep everything off the desk. And so on. For some, A.I. is another soulless contraption we should toss aside, like a politician in the movies who stops reading, crumples the pages and starts speaking from the heart. (How many times have we seen that one?) But human beings have been churning out prefab dialogue and scene structures for generations without artificial assistance. Few seem to mind.
“When screenwriters I know talk about generative A.I., they’re not dismissive, though they’re clear about its limits. One writer says he brainstorms with a chatbot when he’s “breaking story,” sketching major plot points and turns. The bot doesn’t solve the problem, but in effect, it prompts him to go past the obvious. Another, an illustrious writer-director, used it to turn a finished screenplay into the “treatment” the studio wanted first, saving himself days of busywork. A third, hired to write a period feature, has found it helpful in coming up with cadences that felt true to a certain historical figure. These writers loathe cliché. But for those charged with creating “lean back” entertainment — second-screen viewing — the aim isn’t achieving originality so much as landing beats cleanly for a mass audience.
“So why don’t the writers feel threatened? A big reason is that suspense, in some form, is what keeps people watching anything longer than a TikTok clip, and it’s where A.I. flounders. A writer, uniquely, can juggle the big picture and the small one, shift between the 30,000-foot view and the three-foot view, build an emotional arc across multiple acts, plant premonitory details that pay off only much later and track what the audience knows against what the characters know. A recent study found that large language models simply couldn’t tell how suspenseful readers would find a piece of writing.
“That’s why I hear screenwriters talk about A.I. as a tool, not an understudy with ambitions. I realize you’ve got another perspective right now: “We’re not so different, you and I,” as the villain tells the hero in a zillion movies. But don’t sell yourself short. You fed the machine your writing before you asked it to draft a scene. You made it clear what dramatic work was to be done. And so long as you and the studio or production company are consenting parties on this score, you’ll be on the right side of the Writers Guild of America rules. Your employers wanted a script; you’ll be accountable for each page they read. And though generative A.I. was trained on the work of human creators, so were you: Every show you’ve watched, every script you’ve read, surely left its mark. You have no cause to apologize.
“Does the entertainment industry? It was hooked on formula, as I’ve stressed, long before the L.L.M.s arrived. Some contrivances endure simply because they’re legible, efficient and easy to execute. Take the one where one character has news to share with another, but is interrupted by the other’s news, which gives the first character reason not to share her own news. Then comes the inevitable: “So what was it you wanted to tell me?” Ulp! Writers have flogged that one for decades; why wouldn’t a bot cough it up? The truth is that many viewers cherish familiarity and prefer shows, especially soaps and franchise fare, to deliver surprises in unsurprising ways. Still, there will always be an audience for work that spurns the template — for writers who, shall we say, think outside the bot.
“That’s the bigger story. In the day-to-day life of a working writer, the question is less abstract. If people press you about your A.I. policy, point to the guild’s rules. Tell them that every page you submit reads the way you want it to. Then announce: We’re done here.
