ChatGPT is everywhere these days. The AI chatbot has taken the world by storm, astounding people with its human-like conversational abilities. ChatGPT for scientific writing is among the many controversies amid this rapid evolution taking place right before our eyes.
Many experts warn against using generative AI in any form of scientific writing, citing the potential for inaccuracies and ethical issues. Nonetheless, ChatGPT can aid scientific writing in various ways if applied judiciously. It’s not an all-or-nothing proposition, and whether we like it or not, our future content, in all forms, if not our future lives, will involve AI. Learn it, even if you don’t love it.
This article provides tips on harnessing ChatGPT effectively and ethically to enhance scientific writing. If you understand your responsibility as a science communicator (researcher, journalist, or other), you can leverage this amazing tool for good.
- Understanding ChatGPT’s capabilities and limitations
- You hold full responsibility for the final content
- Use ChatGPT for presentation refinements, not core content
- Generate multiple versions for comparison
- Apply an active, critical mindset
- Ethical considerations
- Specific uses
- Finding a place for generative AI in your scientific writing
Understanding ChatGPT’s capabilities and limitations
At its core, ChatGPT is an advanced predictive text generator. It produces human-sounding text by recognizing patterns from its training data, not through comprehension or reasoning. This process has major limitations:
- ChatGPT has no inherent knowledge or facts. It cannot validate responses. If unsure, it will confidently invent information, including fake statistics and nonexistent citations (Anthropic, 2022).
- Its knowledge cuts off in 2021, so it lacks awareness of current events, the latest research, and new discoveries. But that will change sooner than later. Already, other GPTs, like Claude, are much more up-to-date. Real-time updating is just a matter of time.
- ChatGPT aims to sound convincingly human-written. But sounding credible does not guarantee accuracy or truthfulness.
So, ChatGPT cannot replace human intelligence and critical thinking. Its output comprises “intelligent” guesses, not verified facts. You must diligently validate its suggestions rather than accepting them as true.
You hold full responsibility for the final content
As the author, you retain complete accountability for your published content, regardless of what tools you utilized during the writing process. This remains true even if ChatGPT introduced errors, inaccuracies, or ethically problematic material. Your name is on it. You own it.
You’re expected to thoroughly oversee ChatGPT’s contributions to ensure the final product meets scientific standards and publication ethics policies.
Specifically, authorities like the Committee on Publication Ethics (COPE) outline stringent ethical obligations for scientific writers:
- Authors must ensure their work is original and not plagiarized from other sources. Any text suggested by ChatGPT must be thoroughly reworked or clearly attributed.
- References must be accurately cited. Fake citations introduced by ChatGPT violate COPE principles on citation manipulation.
- Facts, statistics, and interpretations should adhere to scientific validity and integrity. Invented “facts” from ChatGPT contradict ethics guidelines.
- Studies involving human or animal subjects demand strict ethical compliance. ChatGPT lacks the reasoning to ensure this.
- Standards of inclusivity and non-discrimination must be upheld—biased language from ChatGPT conflicts with mandates from COPE and scientific journals.
And if ChatGPT’s involvement is detected, merely disclosing its use does not absolve authors of responsibility, as the COPE principles emphasize author accountability.
Authors retain full obligation to meet publication ethics policies and scientific validity standards regardless of drafting aids used. This requires diligent oversight of ChatGPT’s contributions before finalizing any writing for publication.
Use ChatGPT for presentation refinements, not core content
Given its limitations, ChatGPT is best utilized for polishing presentation, not generating substance. Supply key facts, ideas, and reasoning yourself. Then use ChatGPT to refine and enhance how that content gets communicated. For example:
- Have ChatGPT suggest grammar, syntax, or continuity improvements to existing drafts.
- Request it rephrase sections in different styles (more formal, more conversational, etc).
- Ask it to express your points more concisely and punchily.
- Use it to paraphrase sentences and avoid self-plagiarism.
- Give it bullet points to turn into complete sentences under your guidance.
Carefully review all details, statistics, citations, and facts from ChatGPT before accepting. Delete any false or unverified information. The core content must come from you.
Generate multiple versions for comparison
ChatGPT has a stochastic element, providing varying responses to the same prompt (Mast, 2023). Capitalize on this variation:
- Try your draft several times, closing ChatGPT between sessions.
- Independently judge the merits of each revision.
- Combine the best aspects of different versions into superior hybrids.
- Recognize that many valid ways exist to communicate ideas.
By generating multiple options, you expand your choice set. But you remain the decisive gatekeeper for selecting what best suits your work.
Apply an active, critical mindset
Using ChatGPT well requires engaging with it critically and staying actively in charge. View its output as input into your mental process, not the final say. Ask yourself:
- Is this better than my original? Worse?
- Does it spark promising new directions or concepts?
- What specifically works or doesn’t about each suggestion?
- How can I integrate its strengths with my own?
Maintain a discerning, participatory stance. ChatGPT’s role is to aid your communication, not control or replace it.
Like any powerful technology, ChatGPT holds the potential for misuse. Keep your application ethical:
- Cite any sources ChatGPT provides to avoid plagiarism.
- Do not present its text as your own without extensive reworking.
- Reject any unverified claims, facts, or fake references from ChatGPT.
- Use inclusive, non-biased language.
- Do not outsource core authorial duties like generating content.
- Disclose ChatGPT usage if the publisher requires it. Even if it’s not, you should probably take the initiative.
Uphold your obligations as an ethical author first. Let ChatGPT support that role, not define it.
Within an ethical framework, ChatGPT can aid scientific writing in many ways:
- Improving draft clarity and flow: Give ChatGPT a paragraph or sentence and have it suggest improvements to readability, brevity, continuity, or style. Selectively use its top recommendations.
- Restructuring/reorganizing: Ask ChatGPT to propose new arrangements or paragraph orders to strengthen logical flow. Critically analyze its suggestions.
- Paraphrasing: Paste sentences or passages and request paraphrasing for alternative wording. Verify accuracy.
- Developing skeleton drafts: Give ChatGPT an outline or bullet points to produce a rough first draft for you to then refine and expand.
- Simplifying complex explanations: Have it rephrase convoluted sections more simply without losing meaning. Critically review its versions.
The above demonstrates some potential applications—experiment to uncover what works best for your needs, while maintaining ethical usage.
Finding a place for generative AI in your scientific writing
ChatGPT can enhance scientific writing, particularly regarding presentation refinements when applied carefully.
However, misconceptions exist about its actual capabilities. Maintain responsibility for your content’s accuracy, reasoning, and ethics. Supply core ideas yourself and selectively utilize ChatGPT’s output after thorough verification.
Approach ChatGPT as an intelligent assistant, not an authoritative expert. With proper diligence, it can become a useful aid for scientific communication while retaining necessary human guidance over the process.