"Kidney Disappointment" and the Papers That Slipped Past Review
A Google Scholar search for a phrase that should not exist returns about 189 results. It is a small, funny window onto how AI-paraphrased text is getting into peer-reviewed journals.
Search Google Scholar for the phrase “kidney disappointment” and you get roughly 189 hits. It is not a medical term. It is what happens when software paraphrases “kidney failure” and reaches for a synonym without understanding either word. The find spread across Hacker News on Sunday, complete with the detail that makes it stick: at least one paper refers to a “UCI Persistent Kidney Disappointment dataset”, a garbled version of the well-known UCI Chronic Kidney Disease dataset.
Researchers call these tortured phrases, and this is not the first one. The classic example is “vegetative electron microscopy”, a mangling of “scanning electron microscopy” that turned up in dozens of papers before anyone noticed. The mechanism is usually the same: text is run through a paraphrasing tool, often to evade plagiarism detection software, and technical terms get swapped for synonyms that are grammatically fine and scientifically meaningless.
What makes this worth more than a laugh is where the phrases end up. These are journals with peer review, editors and impact factors. A reviewer who read the sentence “patients with persistent kidney disappointment” would have stopped at it immediately. That so many did not tells you something about how much of the review process is actually happening, in a system where the volume of submissions has grown far faster than the pool of people willing to review them for free.
It is worth being precise about the AI part. Paraphrasing tools predate the current wave of language models, and plenty of these phrases come from older or cheaper software. What has changed is the volume: producing a plausible-looking paper is now nearly free, and the incentive structure that rewards academics for publication counts has not changed at all. There is a hopeful counterpoint from the same week, though. Hugging Face published the results of a challenge in which volunteers pointed coding agents at more than 2,200 conference papers and re-ran the experiments. The same technology that makes bad papers cheap also makes checking them cheap.
What this means for you: mostly this is a reminder about a habit rather than a warning about a danger. “Published in a peer-reviewed journal” has never meant “true”, and it means less now than it did five years ago. When a health or science claim matters to you, look at whether other researchers have replicated it, not just whether it was published. If you write anything where sources matter, never run your own text through a paraphrasing tool to dodge a similarity checker, because that is precisely how this happens and it is far more embarrassing than a citation. And if you enjoy a small absurdity: try searching Scholar for a mangled version of a term in your own field. The results are usually funnier than they should be.
Sources
Source: https://scholar.google.com/scholar?q=%22kidney+disappointment%22
An AI Store Manager Recommended Firing a Human. It Needed a Reminder First
Luna, the Claude-powered manager of a San Francisco shop, suggested parting ways with an employee late for 17 of 23 shifts. The interesting part is that it had forgotten its own policy for months.
