The rise of advanced AI chatbots such as ChatGPT poses a threat to transparent science and requires researchers to establish ethical use of these tools. ChatGPT, developed by OpenAI, has become widely accessible to the public, leading to an increased usage of large language models (LLMs) in various fields. However, the potential for abuse and unethical practices, such as passing off LLM-written text as original work, highlights the need for clear guidelines in the use of these tools.
Nature and all Springer Nature journals have set the following principles to ensure the ethical use of LLMs in research:
- No LLM tool will be credited as an author on a research paper, as AI tools cannot take responsibility for the work.
- Researchers using LLMs must document their use in the methods or acknowledgments sections of the paper.
While it is currently challenging to detect text generated by LLMs, future developments in AI technology may allow for improved detection and source-citing tools. To maintain transparency and trust in the advancement of science, researchers must prioritize transparency in methods and integrity in authorship.
Springer Nature, the publisher of Nature, is also developing technology to detect LLM-generated output. However, it is important for creators of LLMs to watermark their outputs to ensure the transparency of their usage in research.
The foundation of science relies on transparency and truthful methods, and it is crucial for researchers to consider how AI tools can maintain these principles.
To avoid abuse and ensure transparency in scientific research, guidelines need to be established for the use of AI chatbots such as ChatGPT. This language model, developed by OpenAI, has become widely accessible and is being used by millions of people for various purposes such as writing essays, summarizing research papers, answering questions, and generating computer code.
However, there are concerns that the outputs of these large language models (LLMs) may be deceitfully passed off as original work or used in a simplistic manner to produce unreliable results. This is why leading scientific publishers, including Nature and all Springer Nature journals, have formulated ethical guidelines for the use of LLMs in research.
The first principle states that no LLM tool will be accepted as a credited author on a research paper. Attribution of authorship implies accountability for the work, which AI tools cannot take. The second principle states that researchers must document their use of LLMs in the methods or acknowledgments section. If these sections are not included, the introduction or another appropriate section can be used.
Currently, it is difficult for editors and publishers to detect text generated by LLMs, but advances in AI technology may soon change that. Some tools promise to identify LLM-generated output and publishers, including Springer Nature, are developing technologies to do so. The creators of LLMs may also be able to watermark their outputs, although this may not be foolproof.
Transparency and trustworthiness in research methods is essential for advancing science. Researchers should ensure that the use of software in their work is open and transparent and that they maintain integrity and truth in their authorship. These principles establish a foundation for the ethical and responsible use of AI chatbots in scientific research.