Your article brings up so many valid points - thank you for sharing it. I've already been guilty of editing for the tool, and I 100% agree that it dilutes the power of my words in the process of trying to reduce the score. For those who write self-improvement articles, lists of potential solutions tend to raise the AI flags, which is the reason for the article itself. I've given up trying to please the machine. If someone disengages because of a Pangram score, so be it.
It’s really hard to ignore as there seems to be a lot of judgment associated with a high score, but I have to keep reminding myself that my genre invites a lot of flowery language - the stuff memes are made of - so that’s going to impact my work.
Genre awareness would make the detector less unfair, but I think it still leaves the deeper category error intact. Even a perfectly calibrated detector can only estimate how the text was produced, not where the argument originated or who understood and owned it. Fact-checking fixes a different problem: whether claims are true. If the goal is authorship trust, the evidence probably has to move from the text to the process.
You may find AI detectors to be even more problematic than that in my latest article:
https://midhunkmadhu.substack.com/p/so-you-think-you-can-tell-whether?utm_source=share&utm_medium=android&r=1q7mkj
Follow or subscribe, or simply like and restack it if you find it interesting!
Thanks for sharing! This is great I'll take a look :) 🙏
Your article brings up so many valid points - thank you for sharing it. I've already been guilty of editing for the tool, and I 100% agree that it dilutes the power of my words in the process of trying to reduce the score. For those who write self-improvement articles, lists of potential solutions tend to raise the AI flags, which is the reason for the article itself. I've given up trying to please the machine. If someone disengages because of a Pangram score, so be it.
I love your comment, Beth! Giving up on editing for the machine is the only way to keep your writing alive. 🚀
At least to me, it is completely backwards that writers are now bending their actual craft to appease a flawed detector.
So glad the piece resonated with you, Beth—keep writing from the page, not the score. 🤗
It’s really hard to ignore as there seems to be a lot of judgment associated with a high score, but I have to keep reminding myself that my genre invites a lot of flowery language - the stuff memes are made of - so that’s going to impact my work.
Such a good point!
The exact language that makes self-improvement or personal writing resonate is the same stuff a blunt classifier flags as synthetic.
I can't with this irony 😅. We're living in wild times definitely!
Genre awareness would make the detector less unfair, but I think it still leaves the deeper category error intact. Even a perfectly calibrated detector can only estimate how the text was produced, not where the argument originated or who understood and owned it. Fact-checking fixes a different problem: whether claims are true. If the goal is authorship trust, the evidence probably has to move from the text to the process.