Do My Earliest Essays Read Like AI?
On low-trust environments in education.
As a teacher, I'm highly tuned in to the multitude of ways to identify AI writing. Em-dashes, "it's not just X, it's Y", bullet point formatting, key words in bold, I know it all like the back of my hand. I once clocked a student's LLM use just by the font; when you export a PDF from ChatGPT while using a Mac, the font is SF Pro. There are obvious reasons why I need this skill as a teacher, but being able to identify AI (LLM) writing is a crucial part of media literacy for everyone. This is for a variety of reasons, not the least of which being that blatant LLM use (in most writing contexts) is a sign that the author does not care. They do not care about taking the time to write something. They do not care about their work being average (rather than great) since the statistical mean is quite literally what chatbots produce. They do not care about the environmental impact, exploitative labor in the global south, their own data security, and so on.
I should state first and foremost that I have never used an LLM to write one of my essays. I have some experience with LLMs for their more legitimate use-cases (e.g., as a coding tool) and I occasionally experiment with them for the sole means of trying to stay abreast of the technology so that I can be aware of what my students may be getting up to. Having a good understanding of the thing I'm critiquing makes me better at critiquing it, and boy do I have lots of criticisms of the tool. And don't worry, I only ever use local models so that the environmental impact is near zero and my data is secure.
Besides, the best creative practice is having weeks of writer's block before having a sudden burst of creativity and writing a whole piece in an afternoon. Subscribe today!
