Academia is a uniquely demanding world with its specialized language that most people find unnatural. In addition to that, there is a prevalence of English in academic texts, which isn’t the native language of many researchers like me. Years ago, at the start of my PhD journey, I faced the daunting task of writing a literature review in English. Despite putting in my best effort, I needed a native-speaking colleague to review the text before submission. While it wasn’t as terrible as I’d feared, it still required significant corrections.
My Experience with LLMs
Having gone through that experience, the thought of writing my thesis – exponentially longer than my review article – was terrifying. I assumed I’d need to invest in professional proofreading services when the time came. Fortunately, I wrote my thesis during the era when ChatGPT had become widely known, and I can proudly say I made extensive use of it!
Interestingly, when I browse online discussions, I notice a persistent reluctance to embrace AI tools in academia. Many seem to believe we’re better off without them. The concerns typically center around AI’s potential errors and the fear that relying on these tools will not only spread misinformation but also spell the end of authentic writing. I understand the skepticism – these Large Language Models (LLMs) do have their share of hallucinations, and for highly specialized fields (which most researchers work in), they often lack the depth needed and can provide incorrect information. However, I believe that rejecting AI now is equivalent to refusing to use calculators in favor of mental calculations or choosing encyclopedias over computers. In my view, we need a paradigm shift, focusing on educating researchers about the optimal use of LLMs.
LLMs for Editing Text
While LLMs might hallucinate when providing references, they excel at text editing. As a non-native English speaker, I’ve found that conventional correction tools like Grammarly aren’t sufficient for improving the overall flow of academic writing. Don’t misunderstand – I’m not advocating for blindly trusting AI to generate your text. Rather, I’m suggesting that having AI polish your best draft preserves your original text and ideas while enhancing their presentation. This is essentially the role that book editors and thesis proofreaders have always played. As someone studying biology rather than English literature, it’s perfectly normal that I’m not completely proficient in English writing (at this point, I’m not even fully proficient in my mother tongue!). I must say that ChatGPT did an outstanding job editing my thesis. I would provide it with complete paragraphs, specifically requesting it to revise the scientific language while maintaining the original ideas and content. It truly was a lifesaver.
LLMs for Literature Reading
While I didn’t have this advantage during my PhD thesis work, I’ve since had the opportunity to experiment with several tools that perform remarkable things with PDFs. You can simply upload your PDF of interest to get instant summaries, and even engage in a conversation with the paper – asking about main findings, limitations, new techniques used, and key conclusions. These tools dramatically reduce the time needed to process a research paper. Again, some might resist using these tools because it feels like you’re not “actually” reading the paper. I understand that perspective. But consider this: how many papers are sitting in your “To Read” folder? If you could quickly get a summary to determine whether an article is relevant to your work, and then read the important ones in full, you could significantly reduce that backlog and clear your mental space. Most of these tools even allow you to chat with multiple papers simultaneously, which is invaluable for literature reviews. Let’s be honest – how many times have you spent countless hours reading papers just to write a single sentence in your thesis or paper? These tools help overcome this challenge. Additionally, since you can chat with the paper, they’re incredibly useful for journal clubs – you can discuss the strengths and weaknesses of the paper and bring those insights to discussions with your colleagues.
Some notable examples of these tools include SciSummary, SCISPACE, PaperGuide, and PaperPilot.
LLMs for Project Troubleshooting
With the growing sophistication of LLMs, combined with the tools mentioned above, I see AI as valuable for discussing protocols, suggesting techniques to explore, identifying gaps in the literature (whether for a review article or planning a new project), assisting with grant writing early in your PhD journey, or helping interpret your results.
LLMs for Helping in the VIVA Preparation
The pinnacle of a PhD student’s journey is the VIVA. One of the techniques that helped me prepare for my VIVA was sharing my thesis’s abstract, objectives, and conclusions with ChatGPT, and asking the LLM to suggest likely questions. Remember, you don’t need to trust the answers blindly, but you should be able to address the questions it generates. You can even challenge the AI to come up with the most difficult questions possible. If you spend time answering the most challenging questions both you and the AI can devise, since you’re the expert on your thesis topic, I’m confident nothing will catch you off guard.
The Dark Side
I understand why many professors are wary of LLMs – what was once a straightforward grading task has become a hunt to catch students who spent thirty seconds generating essays using ChatGPT. Solutions to this challenge are emerging, as evidenced by the growing number of software tools that can detect AI-generated text.
When I was in elementary school, my literature professor required us to write three essays per trimester. In one class, clearly as frustrated as many educators are now, she called out students who had copied essays from online blogs. This was roughly 15 years ago – there have always been cheaters. Time will ultimately reward those who put in genuine effort and expose those who think they could game the system.
It’s also time for professors to get creative and use AI to their advantage. They could ask students to incorporate AI into their projects, generate and proofread AI essays, or use AI to create diverse, creative exam questions.
Conclusion
The world is evolving rapidly, and there’s never been a better time to be a student and benefit from what AI has to offer. But, at the risk of sounding cliché, with great power comes great responsibility. Using AI as a shortcut won’t solve your problems – it will likely create more. However, if you learn to leverage AI and AI-derived tools effectively, you’ll produce higher-quality work and be better prepared for life’s challenges, all while saving valuable time.
I wish you a smart and responsible journey with AI and would love to hear how it has enhanced your research experience.
Cheers,
Mariana
Think this could help someone you know? Spread the word!