Transcription
Now, some people will tell you to take each paper you read and write a summary paragraph and then edit all those paragraphs together, but this is a terrible idea. And it's a terrible idea because when you first start reading, you don't know how an individual paper fits into the broader context and there might be concepts that you don't yet understand. You also might not be able to tell if the paper is any good because not all published papers are. And this is one of the many basic, basic things that the AI advocates fail to acknowledge. A lot of published papers are just bad. So even if you trust the AI platform, which you shouldn't, you can't always trust the literature that it's summarizing.
There are countless videos online showing you how to use AI to write your literature review. All of them make it look easy, but all of the ones that I've seen get fundamental things wrong. And the problem, as with a lot of bad advice, is that it will seem to work at first, but cause major or even irreparable damage later. So the danger is that AI can help you produce something that looks like a literature review, but on a topic that you don't understand. And to the uninformed reader, it might look impressive, but there won't be any real insight into the topic. It will probably also contain serious mistakes that you might not notice if you don't have the expertise.
So, as I've said before, anything you submit as your own work, you have to be able to defend. Yeah, if you have a lazy examiner who doesn't read your thesis, you might be okay, but a good examiner will quite quickly figure out that you've used AI. They will find the parts where you don't understand what you've submitted as your work, and they will quite rightly give you a very hard time. If they discover that you didn't write major sections of your thesis, you will fail.
So can AI be used ethically and effectively? Well, yes, probably. But if you use it, you shouldn't be dependent on it. You should treat it like an enthusiastic but slightly incompetent intern, in that you can get it to do some work, but you need to check everything that it's done. And while it might make suggestions, you should never let it make decisions for you. And if you don't know enough about a topic to judge what AI is doing, you've got to fix that.
So I started my PhD back in 2003, which means I'm closer to the generation that searched the literature by going to the library and reading physical printed copies of journals than I am to the current generation with all of these AI tools. But despite all of the advances in technology over the last 21 years, the fundamentals have not changed. You still need a good knowledge and understanding of the literature to be able to write about it. And isn't that the point? Surely you want to develop your own expertise. You want to know the literature and have some insight of your own, rather than following some chat GPT generated template. And surely you want to develop your own skills so that you can write with confidence without being dependent on an AI platform.
So here's how to write a literature review without using AI. It might be slower, especially in the beginning, but you will be so much stronger for having done the heavy lifting yourself. But before we get started, if we haven't met before, my name is James Hayton. I'm a former physicist. And since 2010, I've worked full time coaching PhD students from all kinds of academic disciplines, in academic writing, and other general project management skills, with the overall aim of making your PhD a more positive experience. So if you want to know more about what I do, check out my website at phd.academy and sign up for email notifications so I can let you know directly when I publish new videos.
For this video, I'm going to assume that you already have some idea of what you want to study. But if you want to know how to find or rather develop a research topic, then leave a comment below and I'll cover it in a separate video. So I'm going to start with a fundamental rule, and that is to never write anything that you don't understand yourself.
Some people will tell you to take each paper you read and write a summary paragraph and then edit all those paragraphs together, but this is a terrible idea. And it's a terrible idea because when you first start reading, you don't know how an individual paper fits into the broader context and there might be concepts that you don't yet understand. You also might not be able to tell if the paper is any good because not all published papers are. And this is one of the many basic, basic things that the AI advocates fail to acknowledge. A lot of published papers are just bad. So even if you trust the AI platform, which you shouldn't, you can't always trust the literature that it's summarizing. And if you follow the standard advice of writing these summary paragraphs or taking key points and then paraphrasing them without understanding them, you will fill pages with content that looks like a literature review, but you will know deep down that you don't understand what you're writing about. And if you have any kind of imposter syndrome, this approach is just going to make it worse.
So we have to build some knowledge and understanding of the literature before we can write about it with any confidence. This means forgetting about writing, at least initially, and taking some time to just read. Now I can't give you an exact step-by-step process here, because it will vary depending on your situation, your current level of knowledge, the state of your field, and what you need to find out. So if you want a process that you can just follow from start to finish without thinking or using your own judgment, it doesn't exist whether you use AI or not. But there are general principles that you can follow and adapt as your needs change over time.
So generally speaking, there are different levels of knowledge that you need to develop. So let's start with the broad picture. So this is an understanding of the trends and debates and the kinds of problems people are working on. So we're not too worried about the details of individual papers here because initially we're just getting a low resolution picture of the field. And you can usually do this fairly quickly. So if you take a handful of recent papers on a specific topic and just read the introductions, you will probably notice that they all say similar things about the current state of the field. They will mention the same problems or ongoing debates. They will mention many of the same concepts and they will refer to the same key papers. For example, one of my projects involved looking at silicon nanoparticle luminescence, basically getting silicon to give off light. Every single paper said the same thing in the introduction, that there had been a surprise result, that others had reproduced it, but nobody knew the underlying mechanism. And if we could figure this out, then there was huge technological potential. Now, there was a lot of heavy technical detail behind it, and there were many different experiments and proposed theoretical explanations, but the basic story is easy to understand. And the key paper that everybody referenced was by Canham, a name that I still remember two decades later, even though I haven't looked at the topic since. I don't remember all the others, but I remember that one.
And this brings us to the next key point. Not all the literature is of equal value. Most papers will have very little impact on the field, but there will be a few that have disproportionate impact, perhaps because they made some major discovery or invented a new technique or developed a new theory. And these get cited far more than anything else. And if everybody else is referring to these people, you need to know who they are and what they did that was so important. And this brings us to the next rule. You should always go back to the original source and check what they actually did and said. Because when somebody else summarizes a paper, unless they quote directly, there is always a subtle change. Now, in some cases, the secondary source explains the concepts better, but I've seen so many instances where they get it wrong, and then other people have cited the secondary source. So it's always a good idea to go back to those key original sources and try to understand what problem they addressed, what they did that was new, and what influence this had on the field. And although it's important to take some time to try to understand these sources, you maybe don't need to get too lost in the detail because you can always come back to these sources multiple times throughout your PhD. And if you come back to a paper a few months or years later, after reading a lot more, after conducting some of your own research, you might have a very different understanding of it.
So a quick side note here. Years ago, I saw a professor on Twitter saying that they only ever read anything once and I wish I had saved it because it was such a perfect example of the kind of misleading bullshit that you see online from people who really should know better. Any professor who says they only ever read anything once is either a genius, is lying, or is really bad at their job. So you can and should come back to the most important sources and read them multiple times because your perspective on them will change as your expertise grows.
So as you build this initial picture of the field, the general trends and debates and the key influential figures, you will inevitably come across concepts, theories and techniques that you don't understand. And it's worth taking some time to try to understand some of these. Now you will need to be selective and prioritise here to avoid getting overwhelmed. And part of this is accepting that you can't know everything and there will always be gaps in your knowledge. This can be a little scary as you've probably come through an education system where you're graded based on how much of the syllabus you know, but at PhD level, we have to let go of that way of thinking because the amount of literature and knowledge is endless. You will never know everything, but you can identify some key concepts that you need to understand for your project.
One of the most important aspects here that is so often overlooked is developing an understanding of common research and analytical techniques. And this is crucial, not only for your own research, but to understand and assess the literature that you read. So many videos I've seen about literature reviews just take what the paper says without any critical thought, but you need to be able to look at how they conducted their research and reach their conclusions. Now it takes time to build this knowledge. So again, it's important to be selective, either based on the kind of research you want to do, or the most common techniques used in the field. And it might be worth finding good authoritative sources on specific techniques. So if you're doing IPA, for example, maybe you should get a copy of the book by Smith, Flowers and Larkin. Now again, you don't need to get lost in all of the detail, but you can get an initial understanding of the basic principles, and then keep going back to that source whenever you need to. Investing time in that kind of basic understanding will then make it possible to understand the papers that use those techniques.
And then finally, we will have specific papers that are highly relevant to your own particular study. Now these might be papers that look at the same or similar problems but from a different perspective, or perhaps they serve as a foundation for part of your research if you base your methods on something someone else has done. And there will probably be a relatively small number of these, but you need to know them really well.
So to recap, as a foundation, you need a basic knowledge of the general trends, problems, and debates in your field. And you can get this from the introductions of a few papers, or from a good published literature review. You need to know the big influential papers and what they did that was so important and how that influenced the field. And you need to understand the most important concepts and the most commonly used techniques. And you need to know the most relevant literature for your specific project. And for all of these you should build up a collection of key high-quality sources that you can go back to whenever you need. And the key word here is quality, because it's not about the number of papers in your bibliography, but the standard of the sources you rely upon. And if you understand even a handful of important, high-quality papers, it is far better than having extensive notes or summaries of hundreds of papers that you don't really understand.
And then, with that strong foundation, you can start to look in more detail at the literature around specific issues of interest, depending on what you need at any given time. And the way you should approach this depends on what you're trying to do. So if you need to do a systematic review, then maybe you need to follow PRISMA guidelines, but maybe you're looking for a solution to a specific problem, or different methods that have been used to measure something, or for the latest cutting-edge results. To find these, you might need to follow a more organic process, trying out different keyword searches or chasing up references in the bibliographies of other papers, or just by talking to people, because universities are full of people with useful expertise. But however you approach it, there will be dead ends, and you will have to sort through a lot of papers that are not useful to you, or maybe they're just not useful to you right now. But that's the nature of academic research.
Now you might be able to supplement this process using AI, but you cannot delegate the process to AI. You need the ability and the patience to search and filter the literature yourself. It is a fundamental skill that you must have. It's slow and sometimes frustrating, especially at the start, but you will get better and better at filtering through the literature to find the best and most relevant sources as your knowledge grows. And in part two I will talk about how to turn that knowledge of the literature into writing, but if you have any questions on this video then please leave a comment below because I have skipped over some points quite quickly. And if you'd like to know when I publish part two please head to my website at phd.academy and leave your email so I can let you know directly when it's available. And if you found this video useful, please share it with somebody who needs it. And if you found this video useful, please share it with somebody who needs it. So that's all from me, thank you so much for listening and I'll see you in part two.