003 | Refusing to use AI to write your book is classist and ableist
At least that’s what NaNoWriMo seems to think
Hello voyageur,
Welcome to HazyFables!
I wasn’t supposed to post this episode today, but when the news broke up, I couldn’t ignore what was happening and decided to write down my thoughts. I’ll link as many references as I can throughout this essay, and I’ll probably update it with another episode if necessary.
In a statement published on September 1st, 2024, updated since then, the NaNoWriMo organisation wrote:
“NaNoWriMo does not explicitly support any specific approach to writing, nor does it explicitly condemn any approach, including the use of AI.
(…) We also want to be clear in our belief that the categorical condemnation of Artificial Intelligence has classist and ableist undertones, and that questions around the use of AI tie to questions around privilege.”
— NaNoWriMo, What is NaNoWriMo's position on Artificial Intelligence (AI)?
Since then, the writers’ and artists’ communities on all social platforms have taken up arms and massively called out the organisation for such a stance in favour of the use of AI in creative fields.
But let’s roll back.
An Introduction to NaNoWriMo
The National Novel Writing Month, often called “NaNoWriMo”, is a non-profit organisation, based in the U.S., that promotes creative writing around the world. They are mainly known for their November program, during which the participants have exactly 30 days to write a minimum of 50,000 words. Every kind of genre is accepted, from non-fiction to fan fiction, video games or academic writing. No fee is required to participate, and everyone reaching 50k words is considered a winner, badges being self-awarded all through the event.
Regarding their free registration, NaNoWriMo’s finances are fed by users’ donations and merchandising1, but also thanks to sponsors. NaNoWriMo has been particularly known these past few years for its controversial sponsors (Inkitt, Manuscript Press), the latest being ProWritingAid.
What is ProWritingAid?
ProWritingAid is declared as an “AI-Powered Writing Assistant”. Past the classic grammar and plagiarism checkers, ProWritingAid also offers a full enhancement of your writing by reformulating your work, analysing and critiquing it (strengths, characters, setting, point of view, pacing), and even helping you find imagination by continuing writing for you to avoid writer’s block. Some bestselling authors are featured on their website for using their software, such as Talia Hibbert, known on social media for her romance books (Get a Life, Chloe Brown; Take a Hint, Dani Brown).
ProWritingAid is, therefore, heavily powered by AI, going beyond spelling and grammar checkers, contrary to concurrent software programs like Grammarly. But other than their features and products, it’s difficult to access some important information about ProWritingAid. As their pinky website is full of illustrations and excerpts of their services, they don’t share much about their policies. Apart from stating that they are an independent company, free of any “corporate shareholders who might pressure [them] to compromise [their] ethics for the sake of profit”2, and using a lot of jargon (that we don’t always fully understand), we don’t know much about its functioning. The indications are vague, only mentioning in a pretty diagram their “unique Natural Language Processing” and typing everywhere they can “Your text is not retain”.
ProWritingAid also supplies a blog, thanks to some authors’ participation, on which we can read articles like Why I'm Not Scared of AI Writing As a Professional Writer (with a thinly disguised ad for OpenAI and ChatGPT) or 9 Best AI Writing Tools for Essays, Business, and Blogs, in which the author states that “if you’re worried about plagiarism, you can use the built-in plagiarism checker to ensure that your writing is plagiarism free before submitting your assignments”. When you know how AI programs are trained and work, you tend to avoid such assertions. Especially when even specialised AI websites declare that ProWritingAid amends are immediately detected as AI-generated content3.
AI and plagiarism
The only intelligent aspect of Articifial Intelligence is in its name.
This is what I repeat all day long to my students. They don’t necessarily agree with me until I tell them I’ll grade them nought if they hand me an AI-generated research paper. Because there are high chance it will be signalled as plagiarised content.
That’s why, every time, I take the time to explain to them what is plagiarism and how it works. Plagiarism is defined as “the act or practice of taking someone else’s work, idea, etc., and passing it off as one’s own” (Oxford English Dictionary). When we discuss plagiarism, most people know about direct plagiarism (or deliberate word-for-word plagiarism), but they tend to forget about the other forms:
Self-plagiarism: in academics, for example, submitting ideas from your own assignments to different teachers;
Mosaic plagiarism: or paraphrasing, “A paraphrase of something [being] the same thing written or spoken using different words, often in a simpler and shorter form that makes the original meaning clearer” (Cambridge Dictionary);
Accidental plagiarism: misquote sources or unintentionally paraphrases4.
In the academic field, the best way to avoid plagiarism is to quote your sources correctly. As I tell my students, “No quotes = plagiarism”.
But what about AI-generated texts, then? How does it work? In her article, Is It Plagiarism To Use AI-Generated Content? The Ethics of Content Creation in an AI World, Maddy Osman explains the whole process in these words:
To generate content with an AI tool, you first have to provide instructions through a prompt. The AI tool interprets the prompt and generates content based on information in its training dataset.
Training datasets vary between tools, and we often don’t know exactly what content was used to train each tool. For example, OpenAI says its various ChatGPT models were “trained on vast amounts of data from the internet written by humans” but fails to elaborate any further.
Often, large language models (LLMs) like ChatGPT rely on information gathered by data scraping, which is the use of web crawlers to collect information from third-party websites.
While scraping data isn’t inherently wrong or illegal, in many cases, the data powering LLMs is obtained without explicit consent from the source or payment for its use.
Long story short, we can’t fact-check AI-generated texts and thus not detect plagiarism by ourselves. The rare times I accept the use of AI is if my students can find the source material used by AI generators to quote it. If ChatGPT doesn't make a mistake by quoting the thesis of an author who doesn't exist, obviously. Again: no quotes equals plagiarism.
AI and creative spaces, not a love story
Since the accession of generative artificial intelligence websites like Midjourney, OpenArt.ai, and Artbreeder, social media turned into a battlefield. Some artistic accounts display the “AI artists” mention in their bio; while “real” artists try to bring them into disrepute by creating guides on how to detect AI in images and drawings.
In 2022, artists Kelly McKernann, Sarah Andersen and Karla Ortiz filed a lawsuit against Midjourney for the illegal use of their art for image generation, the software using “LAION-5B, a nonprofit, publicly available database that indexes more than five billion images from across the Internet, including the work of many artists”5. One of their attorney, Matthew Butterick, summarized this matter with “the three ‘C’s” rule:
Copyright;
Compensations;
Credits.
The problem with AI in creative spaces is that, most of the time, AI generators and algorithms are trained with stolen databases, their use preventing fair compensation of the artists. The use of AI art is also considered disrespectful towards artists and devalues human work6. Artists also addressed the problematic use of these AI generators to create +18/NSFW (Not Safe For Work) imagery, as “AI cannot be trained to tell the difference between legal and illegal image”7.
The writing community has quickly shown its support to artists, with writers condemning the use of AI art and refusing to support fanarts of their characters created with AI. Other writers have, on the other hand, shown support or at least dismissed the harm caused by AI art, incurring their readers’ wrath as they compared the use of AI art with pirating books.
Social media bookish spheres, such as Bookstagram or BookTok, are now pretty aware of these questions and content creators are reactive in terms of denunciation and ‘cancel culture’ when this type of discussion is invited into their spaces. It's hardly surprising that the NaNoWriMo announcement, which was already under intense scrutiny after the child grooming scandal, caused such a stir, with TikToks and Instagram posts multiplying since Monday to denounce this statement, since now updated.
Why are NaNoWriMo accusations of ableism and classicism problematic?
As well as being particularly ignorant of the issues regarding the use of AI generators in creative spaces, it is, in fact, NaNoWriMo's stance that could be described as ‘ableist’ and ‘classist’.
Ableism is defined as “harmful or unfair things that people say, do, or think based on the belief that disabled people (= people who have an illness, injury, or condition that makes it difficult for them to do things that most other people can do) are not as good as people who are not disabled, or that they do not deserve special arrangements that help them to live their lives and be included in society”; while classicism is the “unfair treatment of or negative opinions about someone based on their social class (= economic and social position), especially because they are thought to be from a low social class” (Cambridge Dictionary).
Many disabled/neurodivergent and poor/underprivileged writers pointed out the very ableist and classist natures of such a statement, or even that the NaNoWriMo board contradicts itself by promoting an exorbitantly priced sponsor (ProWritingAid offering a yearly Premium plan at 120€ and a yearly ProPremium plan at 144€).
Moreover, stating that the condemnation of AI is ableist is a roundabout way of saying that disabled people cannot produce art without the help of AI. Which is deeply ableist on its own.
The author C.G. Drews (@paperfury on Instagram) goes even further, explaining in a story posted on September 2nd, 2024 that the very concept of NaNoWriMo is ableist:
Also NaNo as a concept is not ableist?? “Write 50k in a month” is not achievable for everyone but neither is running a marathon. It’s just a for-fun challenge and no one has to do it. Being a fast or slow writer is not better, it’s just you do you.
(…) Not everything works for everyone but it doesn’t make it automatically ableist (Ableism would be if you could ONLY get published if you participated in NaNo or something unhinged like that. Which has never been the case lol.) people need to STOP insulting disabled people by saying we need AI to write ffs.
— C.G. Drews (they/them) on Instagram Story
Conclusion
So, where are we heading?
The NaNoWriMo board updated their statement since then, clarifying:
We recognize that harm has been done to the writing and creative communities at the hands of bad actors in the generative AI space, and that the ethical questions and risks posed by some aspects of this technology are real. The fact that AI is a large, complex technology category (which encompasses both non-generative and generative AI, applied in a range of ways to a range of uses) contributes to our belief that AI is simply too big and too varied to categorically support or condemn.
(…)
We made mistakes in our initial expression of this position. We speak to those mistakes in this letter to our community, and we've simplified the language on this page to reflect our core position.
— NaNoWriMo, What is NaNoWriMo's position on Artificial Intelligence (AI)?
This new statement hasn't caused as much of a stir as the first one, and for good reason: NaNoWriMo has been definitively buried by many. But this case, and all the positions that have emerged, raise questions about an even wider problem: AI is taking more and more space in creative fields and we have to understand it to better regulate its use.
Because whether we like it or not, AI is a tool, just as Wikipedia is. And people will use it, whatever we say about it. Remember when people considered that photography wasn’t an art? The most important in all of this is the use we have of AI.
Because yes, I’m not fully against AI. I think in certain situations, AI-powered tools can help, as an assistive technology, not a substitute one. I’m a true believer in education, and that’s why I think it is important to train AI ethically but also to teach people how to use it correctly (i.e. critical thinking, and fact-checking).
Nevertheless, I also endorse the need for control and laws against the authors’ rights infringements that come with AI. The massive reaction to the NaNoWriMo statement should be similar to the latest news that shook the literary world, when “Meta ‘discussed buying publisher Simon & Schuster to train AI’” or “An academic publisher has struck an AI data deal with Microsoft – without their authors’ knowledge”.
Because, there is today a massive problem with how AI generators are trained, as companies like OpenAI or even ProWritingAid do not disclose all the details of their functioning. On this very matter, in their Private Policy, ProWritingAid states:
When using ProWritingAid, your texts will not be used to improve the quality of our services. Beware: other providers do not take this approach, and you may granting them rights to use your content as they wish.
We use texts hand corrected by professional copyeditors to train our algorithms to ensure that our data is of the highest quality.
If knowing that our works aren’t retained can be reassuring (even if it’s to each their own to decide if they believe them or not), it is alarming to have no other information on which data exactly is used to train their algorithms. Published books? From the free domain? And if not, do authors consent to the use of their works to train AI algorithms? Probably not.
And, unfortunately, we can't do much until copyright is strengthened to protect authors in all situations, particularly in the face of the AI giants.
Thank you for reading this episode. I hope you’ll enjoy this newsletter, don’t hesitate to subscribe to be notified every time I post. I’ll keep it free, so no worries about that! :)
Until then, à la prochaine!
Sasha
Bibliography
Enchanted_Bookworm on Instagram, “The Bookish Girlies Guide to navigating AI art”, on Instagram, April 18, 2024.
Kyle Chaika, “Is A.I. Art Stealing from Artists?”, The New Yorker, February 10, 2023
Office of the Dean of Students, “The Common Types of Plagiarism”, Academic Honesty and Plagiarism - Bodwin
Todd Spangler, “George R.R. Martin Among 17 Top Authors Suing OpenAI, Alleging ChatGPT Steals Their Works: ‘We Are Here to Fight’", Variety, September 21, 2023
NaNoWriMo, “How do I support NaNoWriMo”, last update in 2022
Originality.AI, “ProWritingAid Review”, August 8, 2024
Office of the Dean of Students, “The Common Types of Plagiarism”, Academic Honesty and Plagiarism - Bodwin
Kyle Chaika, “Is A.I. Art Stealing from Artists?”, The New Yorker, February 10, 2023
Enchanted_Bookworm on Instagram, “The Bookish Girlies Guide to navigating AI art”, on Instagram, April 18, 2024.
Enchanted_Bookworm on Instagram, “The Bookish Girlies Guide to navigating AI art”, on Instagram, April 18, 2024.