Interlude: AI Detective, Co-Occurring Retractions, and Finished Cake
I’ve been reading a new non-fiction book that was published last year, and something about the writing isn’t sitting right with me. The alliteration is a little too clever and consistent. Although the author might have a gift for coining phrases, I have a strong suspicion that this book had a lot of help from AI.
I’m not going to knock the author for using AI. Most of us are using AI tools in some capacity, and they can be helpful for your writing. And I know that a few weeks ago, I even said not to worry too much about “AI tells” in your writing. I still believe that.
But as I’ve been reading this book and noticing myself scanning for AI fingerprints, I’ve realized that what bothers me isn’t about the writing quality. It’s about trust—whether we can trust the person behind the words.
The more I question whether the text was AI-generated, the less I trust the author. And I wonder if this trust is why we keep wanting to identify “AI tells” in the first place.
I’ll also add that trust is not about using or not using tools. Writers use all sorts of tools—spellcheckers, grammar checkers, translators, readability calculators, and more. But AI has a power far beyond the tools that have been available in the past. And that power comes with a real risk to trust.
The author of the book did not include a disclosure or acknowledgment that they used AI (like we need to include in published manuscripts), so there is no way of really knowing. And I plan to finish the book because it still offers insight on the topic that I’ve found useful.
Maybe that’s the lesson here: that disclosure is what helps readers trust without playing “AI detective.” This disclosure is required for our manuscripts, which supports trust in the authors and in science. This disclosure is not (yet) required for grant proposals, but I encourage you to disclose AI use in those documents as well to build trust with reviewers. And until this disclosure becomes the norm everywhere, we need to read carefully, trust selectively, and keep disclosing our own use, so that trust (not suspicion) is at the heart of our relationships with readers.
Now onto this week's round-up...
Round-up
Reading
Retraction and correction of scientific articles due to conflicts of interest: a cross‑sectional analysis of the retraction watch database
“Editorial actions explicitly linked to CoI [conflicts of interest] remain relatively uncommon compared with the broader volume of retractions and corrections, and CoI are rarely cited in isolation. The frequent co-occurrence of CoI with other forms of misconduct or procedural problems suggests that undisclosed or poorly managed CoI often signal deeper weaknesses in disclosure practices, editorial oversight, and research governance rather than isolated administrative lapses.”
Just over 400 U.S. patents contain citations to retracted scientific papers
”Of the more than 13 million patents in the USPTO database, 136,162 contained citations to at least one scientific paper. Of these, 401 patents cited retracted papers: 195 cited the retracted paper before it was filed, 92 cited papers retracted while being examined, and 114 cited a paper that was retracted after the patent was granted, the study revealed.”
Watching
This Is Not the Cake
“When you’re writing something, it is incredibly tempting to look at other people’s books. And you’re comparing your flour and eggs with finished cake. Don’t do it. This is not the cake. You have not finished. You’ve barely started. Keep going.”
Tools
Remove AI overviews from Google
When doing a Google search, do you wish you could just skip the AI overview and get right to the search results? You can trick Google into removing this overview by add "-AI" after your search query.
Quote
“Reading, like writing, is a creative act. If readers only bring a narrow range of themselves to the book, then they’ll only see their narrow range reflected in it.” – Ben Okri
Thank you so much for reading.
Warmly,
Crystal
