Speaker 1: The first step of doing a systematic literature review is coming up with a review question, like what do you actually want to know about the world and how can you phrase that as a simple question. You can write down all of the questions you want and then choose from the best one or a combination but I like to go to ChatGPT and use them as like a sounding board and a research assistant so that they can help me really sort of refine what I actually want to do a systematic literature review on. So here we are, we head over and we say, help me define a systematic literature review research question about beards and their smell. Maybe that's what I was interested in. My beard smells lovely. It smells like Australian sandalwood at the moment. Beautiful. It says a systematic literature review research question should be specific blah blah blah. And then it comes up with one. How do microbial communities in beards influence blah blah blah. And it gives me kind of a first start. The one thing I found about any AI that you're asking, it makes a lot of assumptions about what you want to know. So I highly recommend that you go in and you sort of like re-prompt it and you say, I like this bit, but I don't like this bit, or this bit's good, but you're a little bit off on this area. That is how you kind of use this as a research assistant as like a sounding board for all of your ideas. Then once you've got a research question and you need to spend probably most of the time of the first bit of searching on this because it's so very important. Come up with a definitive but broad, and I know that is so contradictory, but you need to come up with something that is focused enough that it will give you sort of like a good outcome but not too broad that all of a sudden, you know, like you're dealing with thousands and thousands of papers. So that is the challenge, and use ChatGPT to get that balance. Now, you can also use frameworks. There's different frameworks that you can use which will help you with this first sort of like step. And I just asked ChatGPT. I'm familiar with some of these, but some of these were new to me as well. I said, what frameworks for a systematic literature review can be used for this question? And it says Prisma, it used Cochrane Handbook for systematic reviews, it's got the Joanna Briggs Institute Methodology, Spyder and Pico. One of the most famous ones arguably is Pico where you say, okay, I've got this P, population, I've got this I, intervention that I'm looking at, I've got this C, comparison of all of the things that I found and O, outcome. Then what happened when they did these things? And quite often the C stands for comparison because it's a quantitative measurement of comparing it to say like a placebo if you're doing a lot of health stuff or another sort of intervention. So that's how we use frameworks to start thinking about our research question. What population are we gonna look at? What intervention are we looking at? What comparison, if any, are we gonna look at? And we're gonna look for the outcomes within those systems and structures that we set in place. So that's step one. Step two, actually, is what defines a literature review from a systematic literature review? Let's get into that. This is so very important for a systematic literature review because we need to know what methods we are going to use to filter all of the different stuff that we're gonna come across. We wanna know stuff like what procedure are we gonna go through to find the literature. We wanna know what keywords we're gonna use, what semantic search terms we're gonna use in certain databases to find the literature. Now, I like to head over to something like Search Smart. This will give you sort of like the best databases to search for your systematic literature review. And so all you need to do is look for scholarly records or clinical trials if you want, put in the subjects or the keywords and then sort of like define whether or not you want systemic keyword searching, backwards citation, forwards, all of that sort of stuff and also non-paywall databases and you click Start Comparison and it will go off and give you all of the different databases that you can look at. Then, keywords. Keywords are so very important because we often find research based on how they're described like in the abstract or the title. So be very specific with your keywords. By the way, I have another video, go check it out here, where I talk about how to find all of the literature that you'll ever need using different approaches, AI, Boolean searches, old school keyword searches, and that video will allow you to find everything you need in your systematic review. But databases are very important. Where are you gonna search? what keywords are you gonna search for, what semantic search questions, and that's new for this sort of like era of AI because it allows us to actually just put our research question into a database and have it sort of understand that question and give us results back. So now we're on to the exciting part which is finding the research papers. The one thing I like to do first and foremost, and that's only possible now because of AI's semantic search. I love it so much. Let's head over to the three tools that I think you would wanna use. The first one is Elicit. Ask a research question. Beards and, ooh, not bears, and smells. Let's see, that's not really a research question, but let's see what it comes up with. But it's that sort of stuff that you need to sort of like thinking about. Like, is that a keyword combination that you want to put in all of the databases or not? Whatever you decide using your meat brain. So, here we go. Here's all of the different papers that I could talk about. Brilliant. The next one is consensus. Beards and smell. Then we can go off and find all of the papers here using that sort of semantic keyword search as well. And we've also got size space. I can go here, beards and smell. And this is where I like to find all of my stuff using keywords and semantic search. So making sense, oh, this hasn't really done too well with beards, beards and issues, blah, blah, blah. So overall, you can see that we've got a little bit of discrepancy between what these pick up. So it's very important, I think, that you try a few to see what works best for you. And then finally, we gotta head over to something like Google Scholar, and we wanna say, okay, what keywords are we gonna put in? This isn't semantic search, this is just putting in beards and smell. And we can use Boolean operators to make sure that we're actually gonna get the papers that are relevant for us. So we can go beards, and then and, because we want and, smell. There we are. So then we're gonna come up with all of the smell and beard articles that it's going to come up with. The smell report, shame and glory. Only the beards, even after beards became merely rather than daring, the rather radical, oh my God, I don't like this one. The British Journal of Sociology, come on now, you can do better than that. But that is where you can go and actually find all of this information. And so semantic, keywords, databases, and Boolean operators to have a look at what you're excluding and including in your search is very, very important. So that is the step three. Yeah, step three, that is searching for the paper. And now we need to filter and screen and read. Once we've ended up with a load of papers from our searching based on the criteria and the methods we set out in step two, we've now got like an exclusion and inclusion protocol where we need to say, okay, we've got all of these studies, Which ones are we going to include and which ones are we going to exclude? And it's a really sort of like simple process of just filtering. This is why you need a load of papers at the top. Put loads of papers at the top and then they have to filter down to the useful papers down the bottom. And it may only be a small fraction of all of the papers you found, but this is what a systematic review is all about. It's about making sure that we include the papers that are relevant for your research question and not just like general themes, which is like a normal literature review where we just sort of say, oh yeah, there's this theme and this theme and this theme. No, this one's much more focused, so we need to filter it. I like to use the Prisma flowchart to work out which ones I'm getting rid of and keep track of the ones I've got rid of and how much I've filtered it down. So a Prisma flowchart looks like this. We've got identification in the top here and then we've got records identified through database searching. In this case, they had 96. and then we've got other additional identified through other sources, and this was none in this bit. Then they removed duplicates, so there was two that were the same, so they removed one of them, and then they said, okay, we've got this many in screen, 95, and eligibility, full text articles assessed for eligibility, there was only five, and all of these were actually excluded because it didn't meet their criteria that they'd set out in part of their exclusion or inclusion criteria. So you can see we've got like examines treatment, not prevention. So this was like obviously like a health study where they were looking at treatment and not the prevention or something. So that was most of them, that was 52. Then one was pediatric, one was irrelevant. Oh no, loads were irrelevant, 37 were irrelevant. So you can see we've gone from 96 all the way down to five at this point. And then full text articles not included. Well, there was none there, which is great. but here we've got four which studies included in quantitative synthesis or a meta-analysis was only four, they got rid of 92 of them because they didn't meet the specific search and exclusion and inclusion criteria that they set. That is so important and that is very, very typical of a systematic review. So now it's about taking those special studies that you found and getting all of the important stuff out of them. you should read them, especially if there's only four. You should read them from end to beginning. No, don't read them like that. Read them however you want, normally with abstract, then to conclusions, then to introduction, then to method, anyway, you get the idea. Do you know what, actually, I've got another video on how to read like a PhD. Go check out that one there. It's much better than what I just said. But now you need to read them and you need to start thinking about how these studies are influencing your research question sort of response. Are they for it? Are they against it? Do they give you a new insight? Is there something sneaky in there when you look at them all together that is surprising? It's those sort of things that really should be sort of milling around in your head. We're not looking for any sort of definitive stuff just yet, but we just need to read, analyze, refine, understand, all of those stuff. Those words are very important, put them there. But now, we've got a couple of new ways that we can actually talk to all of our documents. So one place I really like is docanalyzer.ai and what you can do is upload your documents and tag them as, in this case I've got literature review, you can see I've got one, two, three, four, five, six here. So then we can go to labels and we can go chat with these six documents. And the one thing I love about docanalyzer is that it doesn't like try to make stuff up. If it doesn't understand what you're asking or it can't identify it in the documents that you've given it, it will just say, hey, I don't really know, can you give me a bit more information? It doesn't sort of like BS its way into chat, which I really like. So, for example here, it says to identify the important parts of the document, I would need more specific keywords or topics of interest. That's what I want from an AI, something that isn't just gonna make stuff up. Another thing you can do is head back over to size space, And in SciSpace, you can actually get results from my library. So if you put those very specific studies that you've filtered and found into your library, you can then ask it questions across that library, which I think is really, really fantastic. So not only do you read it all, if you can, if it's a sensible amount of papers, but then you can start chatting to all of the documents together in something like DocAnalyzer and SciSpace, and then you can get sort of further connections, further deeper inquiry into things that maybe you have missed. Or maybe there's just a question, you've read them all, and there's a question sort of in your mind. You're like, actually, does this apply to all of the papers or not? Put it into something like this and it will search across all of your documents. I absolutely love, I'm doing this today, Chef's Kiss, it's my new favorite thing. Chef's Kiss, yum, yum, yum, yum, yum. But doing that means that you're not gonna miss out on anything because you're going to use old school tactics by just reading, read, read, read, read, read, and new school tactics by using AI, AI, AI, AI. Together, they are the perfect combination, yes. And then it's all about writing it up, making sure that you actually talk about what your research question is, the methods you've used, the filtration criteria, and the exclusion and inclusion criteria, the keywords you search for, then what you've found, how they all sort of like relate together, and the outcome. What is the outcome of this literature review? Does it support your research questions? Does it give you a new insight? That is how you write this. That is the structure. It is so very sort of systematic. A systematic literature review has to be systematic, otherwise you'll just end up being completely lost in all of the papers. Oh, so many papers, so many papers. Filter them out, find the good ones, write it out. Brilliant. All right, if you like this video, Go check out this one where I talk about how to write an exceptional literature review with AI. It's going to be a great sort of addition to what you've learned in here. Go check it out.