Speaker 1: So in this video I'll show you how to use ChatGPT to support your qualitative data analysis specifically to support your thematic analysis. This is part one of a three video series in which we will code our data and then eventually develop themes from this data. So in this video I'll show you how to develop the initial open codes then in the following one we'll focus on revising these codes and organizing these codes in a process known as focus coding and then finally in the final video we'll talk about developing themes. At each stage we'll be using ChatGPT to to a varying degree so it's in some things it's a little bit better and some things some aspects of that analysis is still not very good and the keyword here is definitely support so it's going to support our data analysis. So let's get started. So let's get started. So let's get started. It's not going to do the analysis for us although it is an extremely complex tool it's an extremely complex learning language learning model if you just let it do your analysis chances are it will do something that looks all right but at a closer inspection it's not exactly what we want also I still use and I still prefer NVivo or any specific specifically designed software for data analysis over the internet. So let's get started. So let's get started. There we are. So if you haven't watched that one. And now we'll use both. We'll use ChatGPT and Microsoft Word to support our data analysis. For the purpose of this presentation, I'll be using example data sets. I'll be using interview transcripts with actors and actresses. So we have four interview transcripts and our hypothetical study that we're conducting or hypothetical goals of our hypothetical study. They are to explore possible factors that either positively contribute to their job satisfaction and in the long run to job retention or negatively affect factors, experiences, practices. We'll see what we find in our data that negatively affect job satisfaction and may result in these professionals leaving that profession. It is of course important to have some kind of a goal in mind. We will not need it that much at the first stage of developing initial codes, but it will be important later. If you're not sure why it's not important at the first stage, also have a look at my whole series in which I explain and I do the coding and thematic analysis. I don't want to repeat all that stuff now, but the goal is at this initial stage that we're about to start doing now, the goal is to develop detailed descriptive codes. So codes that will essentially serve the purpose of little summaries of our data, as I often say. This whole list, the total number of codes that we'll develop from this initial stage of analysis will serve the purpose of essentially being our table of contents, something that gives us insight into the data before we move on to organize these codes and eventually develop themes. So now let's have a look at my screen. We are looking at ChatGPT. If you haven't used it yet, don't worry. I'll explain everything. And also what I prepared for the purpose of this, I'll explain everything. And also what I prepared for the purpose of this, this analysis are the following documents. So here we have something I call just ChatBPT, for some reason, coding. Of course, I mean GPT. And here we have a list of prompts, a list of prompts or commands that will be useful. I'll explain why. So that's the thing about ChatGPT, it's intelligent or at least very complex. However, it does make plenty of mistakes. So in trying to develop this technique, and trying to collaborate, so to speak, on data analysis with ChatGPT, I developed these prompts that just help me, save me time. I can copy and paste them instead of constantly reminding it what I want it to do. You'll see that that's something that needs to be done. Now we have two transcripts for today. So for this video, I prepared two transcripts, interview transcripts, and finally our blank file with codes and themes. So this is where we'll be, first pasting our codes from ChatGPT, and eventually working on our codes, and then eventually working on our themes. So let's begin. So what we need to do first is instruct ChatGPT what to do. And here I will use my first prompt. So I already have it here. Let me copy it and I'll read it out loud in a second. So as you see, it says, you are a researcher. I'm basically just trying to be very clear about its role and what it's about to do. I will now upload an interview transcript and you will do what is called qualitative coding, specifically initial coding, also known as open coding. The text is an interview transcript. I do not want you to code the questions asked by the interviewer. I want the codes to be detailed and descriptive. I want you to apply codes to sentences or parts of sentences. And later when you develop a list of codes, I want you to be able to tell me what sentences or parts of sentences these codes were applied to. In other words, when I asked you to provide me example quotes for the codes that you create, I would like you to be able to do it. So this is important. This is, you have to have a certain way of communicating to this tool. Otherwise it will get lost. And it's very important to be thinking couple of steps ahead. That's why I keep saying that it needs to be able to show me the quotes because what I did before, I just asked it to code the data. When I later asked for quotes, it said it doesn't have access to that data. I know it's crucial to straight away say that these codes have to be linked to specific quotes. Otherwise it will start giving you imaginary quotes and saying these could be example quotes and so on and so forth. You don't want that. You want to have access to quotes. So this is why, as I said, we always have to keep reminding it everything and be couple of steps ahead. And now here's the text to be coded. Now we'll paste our texts or interview transcript. Let's start with Al Pacino. And we're gonna paste it here and just let it do the magic. So just press enter or click on this icon. As you see now, it's developing the different codes. It's quite random because every time you asked it to do things, it may have a slightly different approach. You'll see that in my future prompts. I'll keep saying that I want you to use the same exact approach. This is pretty good. It's not bad, but I want to have more codes. I want them to be longer. So here are the codes it developed and here are the quotes for each code. But I already have my prompt here that I'll use. And I'll read it out loud in a second. Please develop more detailed codes, please. That's the key word. I'm always nice. I already told you, just in case. You never know when the machines will come after you. So please develop more detailed codes. I would also like the codes to be a bit more descriptive and please separately list quotes that show all sentences or parts of sentences coded with each code. So here we are. What it's now doing is slightly change the approach, which I actually like more as well because straight away we have quotes. And this will be important at stage two when we work on developing these codes a little bit further. So these codes are okay, but they are not perfect. And I would like to, oh, it's kind of weird. It's now mixed all these approaches. I would like more codes, please. I would like more codes without please this time. So it's thank you. And it's giving us more codes. So you kind of have to see, and if you're not happy, if you want more, it's always better. If you follow me, you know that I always say too many codes better than not enough codes. So just use it to the extreme and get as many codes as possible. And this will be characteristic to our whole work with this tool. You constantly have to verify it, constantly have to ask it for more and so on and so forth. So let me just copy this so it has our codes. I'm gonna start copying this into our documents with codes and themes. I'll indicate that this is Al Pacino. This is important. I'm gonna use color coding as well. So I'll explain in a second. So I'll first copy and paste. Remember, there is a limit to how many words you can paste here. These are very short transcripts. If you have a long interview transcript, it's likely that you'll have to break it down into a couple of parts. For each part that you're uploading, you'll have to first include that prompt that I told you about. So it still knows what it's doing. That it still knows it's for the same purpose and it's supposed to do the same thing. So I know where number five went, but let's see. It kind of did a little bit of a mess out of it, but it's all right. It's just for the purpose of presentation. So that's fine. So the goal is to make sure that it's all nice and clean. You have some codes, you have some quotes. I would ask it for even more, but not for now. So for now I'm fine with it. Normally I would probably keep insisting on slightly more. There is no, as you know from my videos, set number of good number of codes or correct number of codes, but the more the better. So as you see, this is color coded. So now I have Al Pacino. In the future, in the second video, I will work on revising some of these codes and perhaps developing some more detailed codes based on these quotes, but it's still saving me time. I didn't have to read it. I didn't have to think of the codes. It's saving me plenty of time. And now what we'll do, we'll repeat the process for the second actor. So we're gonna open first our prompts again, and this time we'll use prompt number three, which is exactly the same thing. So again, I'm reminding it, everything, all over again, you're a researcher and I'll upload all this stuff. What I added in this one, however, is I want you to use exactly the same approach and the same format as you did above. This is important because as I said, it's likely to just mix things up and use a different approach. And it's still likely that it will, but at least it will be similar. So now let's copy and paste our Christian Bale. Our second source. Copy and paste, press enter. Now it's developing the codes again. Before we continue, I just wanted to remind you that there is a lot of services that I offer through my website. So feel free to explore that website. If you need a more detailed, more tailored approach, there are a variety of one-to-one tutorials that I offer during which we can look at your coding or your data, try to develop codes for it, or I can evaluate your initial attempts to coding or address any other aspect that has to do with qualitative research planning and implementation. So again, feel free to explore this website and see if there is anything relevant to your needs. As you've noticed, if I ask it to develop more codes, it's likely that it is gonna put the quote somewhere else. So you constantly have to keep an eye on it and just decide if you're happy with it or not. At the moment, it's looking pretty good and it's actually giving me more quotes than it did the last time. So I'm just gonna start copying and pasting. This is our codes and themes file. Let's say it's gonna be red. And this is Christian Bale. So let's copy and paste. And now we have eight codes, which is definitely not enough. Not sure why it stopped. Let's try continue generating. And it continues generating. So there are more codes, not enough. Let's ask for more. So let's go to our prompt list. Please develop more detailed codes, blah, blah, blah. I would like the codes to be more descriptive. So exactly what I used before. So that's how it works, back and forth. You have to monitor it, just treat it as some child or somebody who's struggling to focus. It helps to think of it this way. And let's have a look and let's see what it does for us this time. Importantly, and I said that already, importantly, we will be working on this. We'll be reading this in stage two, which is video two of this series. We want to work on this. We don't want to just rely on what it develops. We want to work on this. Continue generating. We'll be reading these quotes, which is already easier. It's kind of organized. It's easier than reading. We'll be reading these transcripts and deciding for ourselves. We'll be reading these quotes and we'll be developing even more codes, detailed codes from these quotes. As you know, I like detailed and descriptive codes. And this seems to be not exactly what it does. Sometimes it generates codes that are pretty detailed and descriptive. Sometimes they are kind of abstract and not exactly the way I like them. But for now, I'm going to have to accept what it did. Two more codes. You know what? Let's first copy and paste these. It's a bit messy now again. Let's copy and paste these. And now let's ask for more. I would like the codes to be more descriptive, longer and more detailed. And also, please include the quotes for each code underneath. So we're going in circles. However, it's kind of, it's just giving me descriptions. That's how it's descriptive. But the point of this is that I'm going to be, I'll be happy with this as well. I'll take it as it is. I'll also copy and paste these because they are still based on the same dataset. So the more the better, as I said. So I'll use these as well and I'll paste them next to the codes that it developed before. And I'll just have a little bit more work to do, but at least I'll know that I have lots of codes to work with. And that's important because maybe it missed something in the initial approach. Maybe it worded something better or differently in the second attempt. So, so I'll take all of that. All of them. It doesn't matter. They are, there can be several codes that are exactly about the same thing. And this is something that happens anyway in data analysis. If you're using software or you did it manually, you have to verify later manually and you have to check it manually. So I'd rather have too many. If there are duplicates, I'll just delete them. But I'll just, it doesn't have to be that accurate is what I'm trying to say here. So I'll just take all of these. So I'll copy and paste them here. They are my red color. You can see they're red. You can see there are lots of quotes, lots of codes and it's all red. So, so now we have Al Pacino in green. Now we have Christian Bale in red. I won't be doing all the other participants. It's already a long video, but, but this is what we're doing. This is initial coding. It's descriptive. It's just a little summaries of what's happening in the data. I'll continue to do that. And in video number two and in the future steps, what needs to be done, as I said, is, is manual work. So there are always going to be a lot of things that need to be done. So I'm going to start doing manual work. We'll start reviewing these codes, possibly working on these duplicates, restructuring some of them. We'll use the quotes that it gave us to create more codes. So I like my codes, as you know, I like my codes detailed. So I'll use some of the quotes to just add more detailed, more descriptive codes. And the reason we are using color coding here is that we'll be mixing all of these codes. We'll be merging them and mixing them up. At some point where they come from. So that's very important in data analysis. And this is it regarding the initial stage of thematic analysis. The initial stage involving initial coding using ChatGPT. As you can see, it does speed things up. It helps a lot. I really like to see what it does, what it helps us with. It's still, like I said, nowhere near what professional software can do for you, but it's much better than manual work or coding. Using Microsoft Word only. So I hope that you learned something new. If you did like the video, please comment below. Maybe you have some ideas. Maybe you've already tried to implement this for data analysis. I'm always excited to see how creative you guys are. Every week I see some creative approaches to things that I would normally do differently. So let me know. It's a constantly evolving field. I'm sure there will be lots of developments in terms of ChatGPT. I'm sure there will be another AI when it comes to coding. So it will be a common topic on this channel. Finally, look for the next videos in this series if they have been recorded at this point and share if you know somebody who can benefit from this instruction. I'll see you in the next one. Bye-bye.