The Twenty-First Century’s Latest Social Justice Warrior is…AI?

Written by: Gisele Sanchez

Just last month, the 25-year-old nonprofit Nanowrimo decided to announce it would allow the use of Artificial Intelligence (AI) in this year’s novel writing challenge. They also claimed in their original statement, which has since been updated, that “the categorical condemnation of Artificial Intelligence has classist and ableist undertones,” and “questions around the use of AI tie to questions around privilege.”

Some AI-supporting writers might think it close-minded to not immediately agree with this statement, but it should be acknowledged that AI writing tools are too under-regulated and underdeveloped to be considered the 21st century’s latest social justice warrior. I can certainly acknowledge that AI writing tools have the capacity to help people, but Nanowrimo’s claims are severely misguided and deeply flawed. After all, AI tools are what they eat, and what they eat is made by humans. And, as we all know, humans haven’t even figured out how to talk about social inequality, let alone solve it.

There are a plethora of legal and ethical problems with AI as a whole, but perhaps the most damaging to Nanowrimo’s case is the fact that generative AI tools, which include AI writing tools like Grammarly, are susceptible to human flaws. Generative AI tools are trained to create new data by humans, meaning even though these programs should be unbiased, they are not. Not only that, but AI is owned by humans, too. In an article for the MIT Technology Review, Amba Kak, Sarah Myers West, and Meredith Whittaker emphasize a potentially harmful relationship between AI and Big Tech. They claim that AI’s dependence “on a few unaccountable corporate actors for core infrastructure is a problem for democracy, culture, and individual and collective agency.”

When it comes to generative AI writing tools specifically, all it takes is one surface-level look to realize that they aren’t going to be “leveling the playing field” anytime soon. First of all, most generative AI writing tools aren’t even free. ProWritingAid, one of Nanowrimo’s sponsors this year, has three subscription plans: Free, Premium, and Premium Pro; their premium plan prices range from $30 and $36 a month; $120 and $144 a year; or $399 and $699 a lifetime. Users have to purchase at least the Premium plan for access to “extended AI features,” but even then, neither of the paid plans offers users unlimited access to them. If writers can’t afford to pay for a human to improve their work, they certainly won’t be able to afford to pay a computer to do it.

In my experience trying out the free version of ProWritingAid, the tool did catch and correct pesky misused or misspelled words. However, instead of correcting misspelled words by predicting what I meant to type, the program categorized them as “unknown words” and provided possible replacements. In one case, instead of changing “abcense” to “absence,” the program suggested “because,” “scenes,” “backs,” “becomes,” and perhaps most curious of them all, “bones.”

Outside of words, though, the program didn’t do much else for free. Some of the most basic grammar mistakes, like run-on sentences and comma splices, were considered “Premium-only suggestions.” So, while the program did tell me I was wrong, it did not offer an explanation as to why I was wrong or how to fix it. There were also some more nuanced, or perhaps simply less obvious, grammar and punctuation mistakes, like dangling and misplaced modifiers, that the program didn’t catch at all. All I had to do to get my 17% Grammar/Spelling score to a 100% was correct only the mistakes caught by the program.

This is a fatal flaw that could lead users who are less familiar with grammar and punctuation rules into believing that their work is grammatically correct, and prevent them from seeking further correction. When AI writing tools aren’t capable of being right 100% of the time, they leave it up to the humans to catch their mistakes. The problem is, not everyone can do that. The technicalities involved in writing can be confusing and leave writers discouraged, especially when they might have additional disabilities or financial hurdles that limit their access to proper writing resources.

The mere fact that these AI writing tools are capable of detecting writing errors at all is a marvel in and of itself, and there is no denying that they will be capable of helping a lot of people someday—but, they are still works in progress. Therefore, before we go advertising AI writing tools as potential solutions to serious problems, we must make sure that they are incapable of leaving a job unfinished, or worse, locked behind a paywall.