Ethical Knowledge Creation and AI

This behavior is unethical and long term, harmful to the author who cannot write critically about subjects they are researching.  While most agree copying and pasting the text without analysis or attribution is unethical, others see AI tools as a useful technology, and given their ubiquity, continue to incorporate them into their research and writing practices.

For research purposes, LLMs can be useful in summarizing main contents of an article, highlighting key terminology, and recommending further reading.  The information generated should be vetted independently and traced back to their original sources, particularly because of LLMs’ known predilection towards hallucinations.  ChatGPT’s latest 03 and 04 mini models have shown to have hallucinated 30-50% of the time (Murray, 2025).  One should also refrain from uploading sensitive materials, classified documents, or private files that do not belong to the author into an offline LLM.

This is because this information gets ingested into the LLM, and becomes training data from which it grows its knowledge base.  In the free version of ChatGPT, the default setting allows for this, but in the paid and business versions, it does not.  The New York Times is currently in litigation with OpenAI over their use of its articles to build ChatGPT (Shamsian, 2025).

In terms of writing a manuscript, there are several ethical ways to use LLMs, including using offline large language models to rephrase one’s notes to speed up and sharpen the process, correct language and grammar, and answer simple manuscript-related questions (Zou, 2024). Publishers have adopted language on their author’s guidelines and submission pages to advise users on acceptable uses of artificial intelligence in preparing their manuscripts.

Those on the other side who are reviewing manuscripts should refrain from having LLMs generate an entire review on their behalf and passing it off as their own original work to the journal.  In addition to being highly unethical, it does a huge disservice by denying the author honest feedback.

This feedback is helpful and often essential to improve the author’s manuscript and increase their chances of it being accepted for publication.  Reviews generated in this manner are also likely to contain hallucinations and lack meaningful critique because LLMs cannot exhibit scientific reasoning.

As faculty authors, scholars, and researchers, we must ensure we maintain the highest ethical values when it comes to knowledge creation, as models to our students and the academy.

References: 

Murray, S. (2025, May). Why AI ‘Hallucinations’ are worse than ever. Forbes. https://www.forbes.com/sites/conormurray/2025/05/06/why-ai-hallucinations-are-worse-than-ever/