Plagiarism and Artificial Intelligence: How should researchers react ?

© Element5 Digital/Unsplash

Since With the advent of ChatGPT, generative AI is profoundly changing the problem of plagiarism in the academic world. Although the phenomenon is not new and we are sometimes praised for its merits, #8217;anti-plagiarism weapons, these technologies raise new questions as to the very definition of illicit intellectual property.

Jonathan Bailey, a New Orleans-based copyright expert, aptly points out the complexity of the situation: “There's a spectrum of AI usage, from fully human-generated writing to fully AI-generated writing — and in between there's this vast expanse of confusion.” In this rapidly changing context, researchers are faced with a major challenge: how to reconcile the beneficial use of these tools with maintaining academic integrity ?

When AI blurs the lines of plagiarism

Large language models (LLMs) certainly offer significant advantages in terms of efficiency, clarity and linguistic accessibility. However, they also raise completely new ethical dilemmas.

Fed by vast pre-existing textual corpora, they can easily lead to subtle forms of plagiarism, whether through the misattribution of machine-generated work or through the production of content that is too similar to uncited sources.

Pete Cotton, an ecologist at the University of Plymouth, highlights this reality: ” Defining precisely what we mean by academic dishonesty or plagiarism, and figuring out where the boundaries are is going to be very, very difficult ».

A 2023 study of 1,600 researchers found a widely shared concern: 68% of respondents believe that AI will make plagiarism easier while complicating its detection. Debora Weber-Wulff, a plagiarism expert at the University of Applied Sciences in Berlin, also explains the existence of a paradox: “Everyone is worried about other people using these systems, and they are worried about not using them when they should “.

Plagiarism or unauthorized generation: the blurred line

The central issue that emerges is to determine whether the unattributed use of content entirely generated by artificial intelligence constitutes an act of plagiarism. Some bodies, such as the European Network for Academic Integrity, prefer to call this phenomenon “unauthorized generation of content” » rather than plagiarism in the strict sense.

Weber-Wulff clarifies that, in her view, plagiarism necessarily involves the appropriation of elements attributable to an identifiable author. However, while AI can produce texts that resemble existing human content, they are generally not similar enough to be considered plagiarism in the strict sense.

Some actors go further, arguing that these tools infringe copyright. In December 2023, The New York Times filed a lawsuit against Microsoft and OpenAI, alleging unauthorized use of millions of articlesfor training their language models, which would constitute an intellectual property infringement.

200% Deposit Bonus up to €3,000 180% First Deposit Bonus up to $20,000

The complaint stated that ChatGPT-4 sometimes reproduced nearly identical paragraphs from the journal. In response, OpenAI moved to partially dismiss the complaint, arguing that ChatGPT is not a substitute for a journal subscription; a rather cynical stance. Microsoft, for its part, advocates for responsible advancement of legally developed AI tools.

The impact of AI on scientific publishing

Since the release of ChatGPT in November 2022, the use of AI in academic writing has grown dramatically. A study by Dmitry Kobak of the University of Tübingen estimates that at least 10% of biomedical abstracts published in the first half of 2024 were written using language models, demonstrating an unprecedented impact on the scientific literature.

Graph showing the marked increase in the use of stylized words in academic articles following to the introduction of AI language models. © Kobak, D., González-Márquez, R., Horvát, E.-Á. & Lause, J. Preprint at arXiv

Kobak notes an increased prevalence in non-English speaking countries, while highlighting the possible difficulty of detection in regions where English is predominant. Although the undisclosed use of assistive technology software While writing is not a new phenomenon in academia, AI tools are generating an unprecedented level of concern.

Other scientists, such as Hend Al-Khalifa, a researcher at King Saud University, believe that AI still has some potential: it would allow non-English speaking researchers to focus more on their work, by breaking free from language barriers.

However, there is still considerable uncertainty about the line between legitimate use of AI and plagiarism or unethical behavior. For Soheil Feizi, a computer scientist at the University of Maryland, using AI to paraphrase content from existing articles is clearly plagiarism.

However, he believes that using AI to clarify the expression of ideas should not be penalized, as long as it is transparent. ” We should enable researchers to exploit language models to express their ideas more fluently ,” he says.

Currently, guidelines on the use of AI in academic writing are still very vague. They vary considerably and the very rapid emergence of these tools is outpacing the ability of institutions to adapt properly. The problem is complex and will therefore not find a simple solution. Higher education and research are therefore facing a real labyrinth, which will require collective reflection to be resolved. A dialogue will have to be established between researchers, teachers, students and political decision-makers, au risks missing the boat and limiting the potential of AI in the service of research.

• The arrival of LLMs and AI models on the market facilitates plagiarism in academic production.
• A disruption that also blurs the lines between plagiarism and the generation of unauthorized content.
• Since the release of ChatGPT, the field of scientific publishing has been radically disrupted.

📍 To not miss any news from Presse-citron, follow us on Google News and WhatsApp.

[ ]