Artificial intelligence (AI) is becoming more accessible everyday. It’s also getting better everyday, including its ability to process language, as in editing and/or writing. There is still a lot of human input required, though, but that is diminishing. So what will happen when AI becomes good enough to write or edit with minimal human input? How will anything requiring human writing, like educational assignments and writing contests, adapt to identify writing with AI assistance? And how will it adapt to judging it if it cannot?
Technology has been there for a while to improve writing, like with spellcheck and grammar check. Now you have software like Grammarly to help with more advanced language processing. With AI being capable of analysis and mimicry, or reproduction of similar content, in many fields from science to art, it is no doubt able to write with some instruction. There might not be anything commercially viable at this time, at least not widely available, but I don’t doubt the capability is there. If something as commercial as Google could already simulate conversation, what’s not publicly available is no doubt far ahead. The only questions I have about it is how far, and how soon to be publicly available.
So let’s imagine an inevitable near future where writing with AI is practical. That is, you could collaborate with AI after an acceptable level of learning to know how to communicate with it, and effort to actually communicate with it. Let’s take novel writing, for example, where there is a lot of writing work with the word total alone, to make it worth your while to rely on AI. Also, the detection of AI could arguably be less detectable given it’d be diluted among many more words. Brevity won’t be an AI writing strength for a while, I suspect.
Let’s say for your AI collaborated novel, you could do something like write an outline, with instruction for AI to fill in certain aspects with either its own capability as is, or through mimicry of how some writers’ styles based on their writings that you feed it as samples. For example, you could ask it to describe a countryside in the purple prose (prose text that is so extravagant, ornate, or flowery as to break the flow and draw excessive attention to itself) of some writer, or write a breakup conversation between two characters in the style of some playwright who are great at such things. Of course, you wouldn’t hand in this writing style quiche that would be all too obvious you hadn’t written it all, and of poor quality with all the jarring pieces and transitions. No, you would look at the results to edit and make more homogeneous before putting it back into AI, likely several times over. First, you would give it more instructions to improve from where you left the manuscript, likely several times until it’s reasonably close to the final product. Then, you would get AI to edit more technically as your editing skills wouldn’t perfect, before doing a final review and edit yourself. I don’t want to make it seem that easy, because it probably won’t be in the near future, but also because the workload required will be an important consideration to judge merit to some extent.
Now, with your product finished, you take that piece and submit it as a writing assignment, for publication, or for a writing contest. How would you be caught, even if someone had what they deem to be good evidence? Writing is a judgment call, after all, and if you even put in an effort to smooth over the pieces you and AI wrote, definitive evidence you collaborated with AI might be hard to tell. Technical fingerprints like your access to some platform may be detectable, but I’m thinking it won’t be long before this technology becomes like a word processor that would be undetectable for assistance if you wrote out your assignment by hand, for example. What then? How would merit integrity be maintained? That’s probably a different answer for each scenario, so let’s explore each on its own.
- Writing assignment. Academic integrity is important. However, like with calculators, spreadsheets, math software and websites that are widely accepted today for math courses, unless the writing course were in technical English for determination of writing proficiency rather than style, I’m betting writing collaboration with AI will be accepted without too much challenge aside from some initial fluster that comes with any broad change.
- Publication. This is a more interesting situation to consider. Publication on its own isn’t too big a deal. Lots of crap gets published. The question to the publisher is will it sell? For the reader, will they care who and/or what wrote it if it were good? I don’t personally think so. However, if an author starts to gain fame, and became under suspicion for collaborating with AI, or worse, gets busted for doing so, would readers care? Can they separate the work from the author? Would it be like in sports where fans would “turn” on cheaters? For the record, my guess is that people won’t care if the writing were good in terms of books. As for author fame, I think fans will generally turn their back on methodology, ultimately. That is, they ultimately won’t care as collaborative writing with AI will become more acceptable with time. If they did, it might be a matter of who can do that best as it will be prevalent so let’s see who can do that well to support them… maybe even aspire to be like them!
- Writing contests. This is, by far, the most interesting and challenging scenario to contemplate. Will there be efforts to try and weed out collaborative writing with AI? If so, it had better be behind the scenes because it will controversial, at best, and a complete firestorm, at worse, if people could find any patterns among authors disqualified to claim bias on other grounds. But if not, what will these contests do, ask for a simple checkmark on a self-sworn statement of authorship integrity? Then leave the rest to be the potential equivalent of an Olympics without drug tests? The writers collaborating with AI may not be able to win as often as athletes on steroids would be able to, I’m guess, but some writers collaborating with AI will be able to. Then what? Have contests with every potential winner suspected of cheating, or possibly get accused by some who didn’t win as such? Would some organizations even host writing contests any more if they couldn’t verify the writing were genuinely only written by humans? What would the roles of the judges be then, just to identify great writing, or with an ability to sniff out AI collaboration? Oh, the things to consider! As for what I think, I have no doubt there will always be writing contest. Storytelling is far too human for society to abandon it for a little technical interference. I’m guessing there will be a self-sworn statement for integrity, with prizes being stripped from those busted later. To be able to do this, there will need to be rankings of results beyond those with prize money to bump up other entries should winning ones get disqualified. Those authors will be blackmarked for integrity in the contest realm, but sliding back to publication, their works and reputations will be as that which I foresee with publication. As for the judges, their jobs will probably remain the same, no more than sport judges could be asked to detect and prove steroid use on what they see.
So that’s my take on what I see happening with writing in the near future. I don’t think it’s a question of reality or not, but only when. If you have an opinion on this, I’d love to hear what you think. Otherwise, let’s wait and see and I will revisit this post in maybe 2024?