Video: What Researchers Should Know Before Using AI Tools: Q&A with the Expert 

The discussion on responsible use of artificial intelligence (AI) tools in the research landscape is not new. However, questions on AI use keep emerging. In this video, our expert, Dr. Sunaina Singh, addresses some of the key doubts raised by early-career researchers in this interactive Q&A session.  

You will learn about self-plagiarism, plagiarism, AI detection score, and much more. Discover how to use AI tools ethically in research. Here are some questions answered:  

• My co-author uploaded a blind manuscript on ChatGPT for suggestions. Will it be problematic from a plagiarism point of view?  
  From a plagiarism point of view, it wouldn’t be a problem. As long as you have not copy pasted the output given by ChatGPT and used it. Then it could be a problem. But if you just saw the output and you have edited it, then it should be fine. You may just use ChatGPT to get some ideas and brainstorm. That is perfectly fine and there’s no problem of plagiarism.
  
• What are the key differences between ethical AI, trustworthy AI, and explainable AI?
  Explainable AI would be where the end user knows what are the training parameters, how was that particular model trained on what dataset, and how many parameters, and basically everything you want to know about the tool is available to you. Maybe they even give you a fact sheet on how to use the tool and things like that.
  
  Between ethical AI and trustworthy AI, probably it’s a very fine line. Because trustworthy is more about the user being able to trust that. So that could be based on the brand. You know the principles they follow, the guardrails they follow. So for example, as I believe, and I could be wrong, but on the basis of what I’ve read, Open AI is more about innovation and pushing and having ads and pay, everything has to be paid and things like that. Whereas Anthropic is more cautious, which is why they didn’t go ahead with the Pentagon deal.
  
  So for ethical and trustworthy AI, you will have to really read about the developers and understand their policy. Follow their marketing campaigns, are they more on marketing or do they actually feel responsible? Do they follow responsible use? The exact answer to this probably, I don’t really know, but I think explainable AI is slightly different from this.  
  
• Is there a single AI-based tool which can use for literature search, writing, and plagiarism check?
  To my knowledge, I’ve not used too many of them, but I think Paperpal. It helps with writing and definitely does the plagiarism check and there is a literature search option also. So you could check with that. Maybe even Trinka does that. So there are a couple of at least two, three, like it’s a suite of applications. So they will have about two, three options like that. So Paperpal for sure has these three features that you mentioned.
    
• What is the difference between plagiarism and AI detection score?
  These are two totally different things and very important to know. A plagiarism score is a very absolute and exact score. It will tell you the similarity of your text with previously published text. So if you have plagiarized or if there are a couple of similar words or sentences, it will detect 70% similar, 30% similar, etc. So up to 15 or 20% similarity is considered acceptable in certain parts of the paper. Anything more than that means you have plagiarized it. So that is very clear cut.
  
  AI detection score is a gray area because it will detect the possibility. It is not an exact score. It is a possibility that this text may be generated by AI. There’s an 80% chance that this was generated by AI. So it is not saying for sure, but chance or 70% chance or 15% chance.
  
  Now, the problem is, these are not reliable. But unfortunately, a lot of journals and a lot of professors use it. It creates a lot of false positives. The thing is that AI generated text sounds like human text. So anything you write and put through it, it will give you some score. So there are other ways in which you can monitor if somebody has used AI, but these AI detection tools are not very reliable.
  
• Is digital storytelling still relevant?  
  Oh, yes. Of course! Why not? I think it’s more relevant now than ever before. Storytelling in any form will always be important, especially in science communication. And now with so many digital tools, definitely.
  
• Can we use AI to help us locate the right methodology to use for the data?
  Sure, you can ask questions like that, but don’t completely rely on the answer. It may give you a couple of ideas. So those ideas, maybe you can work on it and something that you were not thinking about, it may give you that idea. And then you can research about it, read about it, and then you can hit upon what is the right methodology. So definitely ask questions for that inspiration, but don’t believe it blindly without reviewing it.
  
• Is there any difference in percentage for accepting paper in the cases of plagiarism and self-plagiarism?
  I would say it is similar, because self-plagiarism is also plagiarism. So the publisher could claim a copyright suit or something like that. If you want to get the paper accepted, in both cases, it will be a little difficult. Self-plagiarism may have a slightly higher chance of acceptance. But that’s only because you’re asking the journal. But both cases are unethical. So try to avoid plagiarism at all costs. Paraphrase it completely and you can get professional help or use an AI tool to paraphrase.
   
• Can we use AI for chapter review of literature in thesis to summarize paragraph grammar or give only important points in a previous article?
  AI tools can generate a chapter review of your literature. It will also give you a summary. But don’t use that entire summary. You should then read it and then figure out how to structure that summary. Try writing it in your own words and then you can polish the grammar. My point is that while you ask it to generate a summary, what happens is we just blindly say, “Oh, what a nice, beautiful summary. I’m going to copy paste this in my thesis.” That is what you should not do. So read it, improve on it, add some text, delete some text, and then polish it for language and then it’s fine.

You will understand the key problems of using general tools like ChatGPT, Claude, Perplexity, and other generative AI tools for research work. Our expert also recommends alternative tools like Paperpal, which are meant to serve academic researchers, and highlight how they are more beneficial than general AI tools.