TechX: Ethically using AI tools like ChatGPT as your guide? – Dr. Akanksha Mahajan
We are all aware that ChatGPT, an AI tool introduced in November 2022, passed the United Nations Medical Licensing Examination (USMLE) some time ago [1] and ever since then, it has also earned a position as a potential research assistant [2]. However, as medical students, future medical practitioners, and researchers, you must not get swayed by these tools because while they greatly reduce your headache, they always carry serious ethical implications. So, firstly, let us focus on avoiding those ethical implications before we dig into the “mentorship part” of these AI tools.
● Avoiding Plagiarism: AI tools can generate content, but using it verbatim in research without proper attribution or as a replacement for your work is ethically problematic. Beginners should understand that ChatGPT is best used as a guide, not a ghostwriter.
● Maintaining Originality: ChatGPT can assist in structuring and brainstorming ideas, but it should not replace a researcher’s unique voice or critical thinking. The authenticity of one’s insights is key to valuable research, and over-relying on AI diminishes this originality.
● AI as a Supplement, not a Source: ChatGPT doesn’t have access to real-time, verified databases or peer-reviewed research, meaning it may not always generate accurate or up-to-date information. Beginners should cross-check AI-sourced content against reliable academic sources to avoid introducing errors or biases.
● Risk of Propagating Errors: AI may confidently present inaccurate data or outdated concepts. Users should always corroborate findings with empirical data to ensure their research remains evidence-based and credible. This habit safeguards against the unintentional spread of misinformation.
● Recognizing AI’s Limitations: AI-generated content may contain inherent biases based on training data, potentially skewing outputs. Researchers should be cautious, especially in fields that require cultural sensitivity or nuanced understanding.
● Interpreting AI with a Critical Eye: Beginners should interpret ChatGPT’s output critically. Even if AI provides a well-structured response, researchers need to adapt and critique it to fit the specifics of their study.
● Ethical Diligence in Learning the Research Process: ChatGPT can provide shortcuts, but ethical research requires understanding the “why” behind each step. For example, when generating keywords or structuring a review, users should treat AI suggestions as starting points, understanding the rationale behind each AI-suggested step rather than accepting it passively.
● Long-Term Skill Development: Ethical use of AI also considers the researcher’s growth. Beginners should use AI to learn and not bypass crucial aspects of research training, which ultimately makes them more competent, independent researchers.
Now, let us move on to the utility of AI tools as a research mentor!
I feel that review articles are the perfect place to start your research journey as a beginner. Review articles are an ideal starting point for beginners in research due to their accessible structure and skill-building potential. They provide a broad understanding of a topic by synthesizing a range of studies, allowing beginners to see patterns, identify gaps, and gain a comprehensive knowledge base in their field of interest. By working on a review, new researchers develop essential skills in literature searching, critical analysis, and organizing information.
Unlike original research, reviews do not require primary data collection, making them a low-risk, high-learning opportunity. They also offer exposure to different types, such as systematic reviews, narrative reviews, and meta-analyses, each with unique methodologies that enhance versatility. Through this process, beginners learn to define research questions and formulate hypotheses, setting the stage for future studies.
Review writing also requires negligible financial investment. Moreover, publishing a well-constructed review can establish credibility and attract collaboration opportunities.
Overall, review articles offer a structured, affordable, and comprehensive entry into research, building foundational skills and confidence for more complex projects down the line.
Now I will try to provide a brief description of some of the review articles and some ethical ways to utilize ChatGPT for writing them.
Narrative reviews:
Narrative reviews do not involve any kind of analysis, they simply summarize various findings about a particular topic. When you truly master them, identifying gaps becomes way simpler. As a beginner, if you struggle with finding topics, you can ask for suggestions from ChatGPT about any recent research in the area of your interest and focus your narrative review on that. Now, when you have narrowed down a topic, refining your search strategy on PubMed becomes the next challenge and you can certainly use ChatGPT to help you come up with suitable keywords for searching the articles on PubMed.
Systematic reviews:
When you master narrative reviews, systematic reviews become a piece of cake. Systematic reviews are like a qualitative analysis of the available data.
This is an example of how ChatGPT can help you during the process of working through a systematic review:
Now, you can also ask ChatGPT to “elaborate” any point that you did not understand and if that still does not help, just write an input “explain in layman’s language in detail,” and witness the magic unfold.
Meta-Analysis:
To be very honest, this was something that always fascinated me even more than original research papers. Why? Because I feel learning the art of conducting metanalysis is even more essential than conducting loads of studies on whatever limited sample size you get. Let us say you are curious about some drug’s efficacy that recently started gaining popularity or something. But you checked research papers and some mentioned positive results in terms of its efficacy and some did not. So, this is finally when metanalysis comes to your aid. Meta-Analysis helps you quantitatively analyze the data from a group of studies about a particular topic that have been conducted so far. When it comes to quantitative analysis, for most of the medicos, understanding the various tools or software is the hardest part. I tried taking tons of tutorials on “RevMan” and despite that, remained so lost every time I tried using it, so, this is how ChatGPT can help you, you can ask ChatGPT to give you the set of exact commands that you are supposed to be using and how you need to use them etc.
These are just a few examples but I hope you have understood my intention behind this article. And I am hoping that AI can help you “ethically” as long as you don’t have a human researcher to mentor you! 😉
References:
1. Alkaissi H, McFarlane SI. Artificial Hallucinations in ChatGPT: Implications in Scientific Writing. Cureus [Internet]. 2023 Feb 19; Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC9939079/
2. Mondal H, Mondal S. ChatGPT in academic writing: Maximizing its benefits and minimizing the risks. Indian Journal of Ophthalmology [Internet]. 2023 Nov 20;71(12):3600–6. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC10788737/
