How to Use ChatGPT with NVivo to Build Themes Fast (Full Transcript)

[00:00:02] Speaker 1: Hello, everyone. The data that I'm going to use for this demonstration is about responsible innovation. So this is the purpose of the study. These are the research questions that I want to address in this study. And also, these are the demographics. There are 10 participants and these are their roles and information about their age range, gender, and also geographical location. So imagine that you have gone through your data using InVivo. So this is the data set. You have gone through all of the transcripts. You have all your containers for your research question. I see the research question one here, two, three. And then you have developed codes. And then for the research question one, you were able to go through everything and categorize the code to develop themes. So here we have 1, 2, 3, 4, 5, 6 themes. For research question two, which is the barriers to fostering responsible innovation, you have 1, 2, 3, 4, 5 codes. For this one, you don't have to go ahead and develop themes because there is not a lot of codes, right? So all these codes could be called themes. The reason why sometimes you develop themes is that you want to reduce the data. That's why you are categorizing all the codes to develop themes to address your research question. So in the situation where you don't have a lot of codes, then all of them become themes. You just maybe sometimes have to review them and making sure that the labels for the codes best reflect the significant information that has been extracted from the data and also address the research question that you have. So for this one, we don't have to do anything about it. For the final research question, which is opportunities for enhancing responsible innovation, we have a lot of codes here. So in some situations, sometimes it's very difficult to categorize the codes, right? This is where you can use AI to help you to categorize. So in this case, we're going to use AI to help us to categorize the codes that we have so that we'll be able to develop themes to address our research question. One thing that you have to think about is that, you know, we always say that when you give wrong information or inadequate information to AI or any AI tool, you may not get rich information. So make sure that all the labels that you have here, you have developed the codes so well that they are representing the significant information that has been extracted and also addressing a research question so that when you use AI to be able to know what exactly all these codes are for and to be able to really see what are the similarities among them and group them for you. Another way to enhance the use of AI is when you are about to give all this information to AI, you have to make sure that you define their codes, right? How do you define the codes? So what it looks like here, you can go to code properties and then you see the description. This is where you put what the code represents, right? So making sure that you define the code helps the system to understand what all these represent so that you'll be able to group it in an appropriate way. So these are the things that you have to know before you go ahead to think about using AI to categorize. So let's assume that we have done what you're supposed to do, right? So what you have to do is right click on any of the codes here and then you go to export and then you export list. So you have to find a place to export. So it will be code, so it can be a name, maybe responsible, innovation, and then I look for a place to save that information. So I save it on my desktop, right? When you finish saving, the next step is to go to charge GPT. So I have charge GPT here. So you may ask which model do you have to use? You have two main models, GPT-5 and also the legacy model, GPT-4.0, right? I'm using GPT+, which is the $20 amount. So you have two options. So I always recommend that when you are just using this AI tool to analyze your qualitative data, you don't have to use the thinking model. So under GPT-5, you have three options. You can choose auto. Auto means that the system will decide whether, based on the task that you have given to it, whether you want to use fast model or the thinking model, right? So auto, you give the system a chance to choose whether to be using a fast or thinking, right? So in this case, I will, if you are sure, me, I'm very sure that fast model will be the best, right? So I choose that. Another option is to choose the GPT-4.0, which is not a thinking model. It can do the same thing that the fast model can do, right? So in this case, I just choose the fast model. Then what you have to do is to attach the information that we save. So let me look, attach there. First, let me open it for you to see. So this is how this one looks like, right? So our focus is only the research question three, because we have all the codes. We haven't categorized them yet to develop themes. So I will upload that document here, right? And then that, now we're going to give the system instructions and see what we're going to get. So first give a little background information about your study, right? What your studies are about. So let me copy this one. So I'm copying the purpose of the study, right? So first, maybe you can start by saying that this is what my study is about, right? And then you paste the purpose of the study. So the next step is to tell the system what you want it to do, right? Now a little background about what the document that you have attached the spreadsheet is about, right? So you can say that the attached cell spreadsheet is output from InVivo 15, which contains a list of all the codes and themes developed under their respective research question. Now, I want you to only focus on categorizing all the codes generated under research question three, leading to the development of themes addressing the question. And then I can put the research question there. And another thing that you may have to provide is that you have to give a little information about how the theme should look like, right? So you can say that make sure each theme is between two to five words. I'm presenting the codes group and addressing the third research question. Okay. So in terms of the output too, you can specify whether you want it in the table format, you want the system to define the themes and also indicate the references. When you look at the Excel spreadsheet, there's also references and also files, right? It gives you the number of files and references. Do you want the system to also show that? You can also ask, right? So you can see here that when I click here, you can see all the files here and also the number of files connected to each of the codes and also the number of references. Do you want the system to create a table indicating the number of files for each of the themes and also definition of the theme? We can do that, right? So you can say that when you create a table indicating the number of files and references for each theme, including definition of each theme, right? So the sentence, the information might not, like the way I'm putting might not be perfect, but the idea is that I want a system to create a table and also indicate the number of files, number of references and for each theme, right? So let's see what a system is going to give us. So we click on enter and then let's see. So perfect. So we see how the system has created the theme. So we have a total of six themes and also a number of files aggregated is 23 and also the number of references is 29. You can always cross check and find out. So you can look at a table and look at the references and also the files connected to each of the codes and then you can aggregate them and then you can see whether the numbers are right. You always have to check, make sure that everything is right. You also have a definition here that is very good. So you see how if you provide a system a specific instruction, provide a system a lot of background information, you get a very good result that will be helpful for your analysis or your fed analysis, right? So when you get this one, you can easily copy and put it on the Excel spreadsheet or you can put on a document and you can even go back to in vivo. So when we go back to in vivo, you can even create themes, right? Based on what you found. So this is the first thing which is trust and accountability. So what are you going to do is that you right click here, you click on new code, you bring that information here, but you have to make it a little bit different. So in parentheses, you can say theme and then the definition has already been provided to you. You can copy and put a definition here just to remind you about what it represents and then check aggregate coding from children. You click on okay. And now, so when you look here, you have building trust and continuous accountability. So we have to look for building trust and continuous accountability. So building trust is here. You can drag and drop if it doesn't work. Okay. If it doesn't work, you can even right click and cut and copy. So we can right click on the second one, you cut and right click on the theme and paste. So this is the moment of truth. We're going to see whether the numbers are right. So it's three for files and four for references. So let's see. HHPT is right. We have three and four. That is good. So you're going to do the same thing by creating another container for the theme, which is collaborative partnership. Right. So right click here. You go to new code. You bring a theme here. You paste. You copy the description. You put it here and then you can check aggregate coding from children. Now what we have to do is to look for the code. So this one belongs to that. So you can drag and drop and then public private partnership. Let's look for the public private partnership and then let's put in there. So we have six files and 10 references. So here, what has it given us? Seven files and 10 references. You see the numbers might not be good. Right. So that's why it always says HHPT can make mistakes. So don't over depend on the result. Always cross check. Now we know that the number here should have been six, not seven. Right. Because we have that information here. You know, this one is right. Because when you look at this, click on the plus sign. Right. You have six here and you have one. Yes. I know the reason why the system said seven. The system was counting each of them. It counts a file once. Right. So let's say a file information was extracted and connected to this file. Right. For this code and the same file, the information was extracted and connected to this this code. Right. When you are calculating, aggregating, you don't say two. Right. Because that information is from one file. So a file is counted once. But for HHPT, a file is counted more than one. Right. Depending on how many times it's connected to a code. So you see the way of counting here is very different from the way of counting in HHPT. So you always have to check and make sure that you are following the standard way of counting. The reason why you have to follow this way is I saw that at the end, the number we're going to get here will not be above the total number of files. What is the total number? We have only 10 participants. Right. So here is supposed to be 10. Right. So if every information is extracted from every participant, you have to get about 10. Right. You shouldn't go beyond 10. But if you count a file more than one time, then the number here will be more than 10. So I think I would choose the way that InVivo does the calculation. Right. So that's it. So what are you going to do? You do the same thing. Right-click, create a container for the themes, and then drag and drop. And then you have your themes addressing your research question. So you see how HHPT can help you to speed up the process, especially when you're having difficulty categorizing your codes. Right. And sometimes when the codes are a lot, and impossible for you to do it based on the time that you have, you know, you can always use AI to help you. And also let people know, right, disclose that you use AI to assist, right, not dictate for you, to assist in the data analysis process. So I hope this one was helpful. If you have any questions, put your question in the comment section, I will be happy to address them for you. And don't forget to subscribe to my YouTube channel. I have good videos for you. And also you can suggest a video that I have to create, and I'll be happy to do that. So thank you for your time.