AI Ghostwriting Is Creeping Into Science—Is That a Bad Thing?

Which words give AI away? A new study of more than 15 million biomedical abstracts on PUBMED has found that at least 13.5% of scientific works published in 2024 show signs of writing tools with the help of Ai-Asisted, especially Openai.

The studies Scientists from Northwestern University and Hertie Institute for AI in brain health at the University of Tübingen found a sharp increase in the verbal samples associated with the written AI. These were both unusual concepts – for example “immersion”, “underlining” and “demonstration” – as well as more known words such as “potential”, “finding” and “essential”.

To measure this change, scientists compared the frequency of words in 2024 with the default data from 2021 and 2022. They finally identified 454 words, which were often used by AI models, including “encapsulation”, “remarkable”, “underline”, “exploration” and “trouble -free”.

However, experts explained Unscramble This frequency of the word itself is not sufficient proof of the use of AI.

“The language is changing over time,” said Stuart Geiger, associate professor of communication in UC San Diego. “Delve” has risen sharply, and this word is now in the vocabulary of society, partly because of Chatgpt. “

Geiger stressed that AI detection in writing is not just a technical challenge; It’s also ethical.

“The only way to reasonably detect the use of LLM is, if you are there, tracing the writing process,” he said. “It comes at a high price, logistically, morally and technically.”

Stuart, however, warned against jumping to the conclusions on the basis of the surface level without knowing the whole context.

“It could be that they just saw a lot of writing generated by chatgpt and now they think it looks good writing,” he said. “This is the whole problem we are fighting with on academic soil, especially when we can’t just place students on the seats and make sure it’s just a pen and paper.”

As the text generated by AI becomes more common, educators turned to tools that claim to detect it; However, the quality of these tools varies.

In October 2024, Unscramble guidance tested AI detection Tools – including grammar, quillbot, gptzero and zerogpt. The results varied wildly: Zerogpt claimed that the US Independence Declaration was 97.93% generated by AI, while Gptzero gave it only 10%.

“A lot of snake oil is sold,” Geiger said.

According to Geiger, concerns about AI writing tools reflect past debates on spelling, wikipedia and cliffs, and reflect deeper questions about the purpose of writing, authorship and trust.

“People are afraid that when you had to write the words yourself, you had to think about them,” he said. “That’s what people respond so hard when they see something that feels suspicious.”

Professor Business University of Rice University Kathleen Perley claimed that while writing artificial intelligence often shows patterns such as repeated structures or overused words like “delve”, it depends most on whether it helps scientists without risk. This is particularly true, she said, in non -native English spokespersons or people who face other challenges.

“If AI helps scientists to overcome challenges such as language barriers or learning disabilities, and does not endanger the originality or quality of their work, then I don’t see a problem with it,” she said Unscramble. “I think it could be a general benefit, because it allows people who have had different facilities, ideas, expositions, participate in something that could be an obstacle to the lack of formal writing skills.”

AI Deans advisor in Rice Business, Pearley noted that another dilemma is the tendency of people to change the way they write for fear of being accused of using AI, and adding that they realized some words that could be described as potentially generated.

While some criticize this style as a missing personality, the pearley sees ai-asisted writing as an instrument that can democratize participation in formal research.

“Sure, we can get more” immersion “and Em dashes,” she said. “But if AI helps people from different environments to share important research, I don’t care how polished it – it’s worth it.”