AI-Assisted Qualitative Analysis With Researcher Control (Full Transcript)

[00:00:00] Speaker 1: AI can make qualitative analysis faster, but it doesn't necessarily mean that it's going to be better if you use AI. I'm going to show you a software that can help you to analyze your qualitative data with the help of AI, but also help you to keep control over the data analysis process, reviewing AI suggestions, and also documenting your actions and decisions so that you'll be able to report what you did to your audience. And that is what I'm going to show you today. So as you can see here, I've opened my Invivo AI Cloud, and when you open, this is what you're going to see. You will see all your projects. I'm going to use this data to show you the step-by-step how to use this software to help you to analyze your qualitative data. So I'm focusing on burnout among primary healthcare physicians. I have my two research questions here, and then these are the demographics. And let me show you one of the transcripts. So this is one of the transcripts. So I'm going to have five transcripts, it's very short. I'm going to upload all the transcripts and show you how to use this AI-powered qualitative analysis tool to help you to make sense of your data, at the same time, documenting every step that you take in the process. So let me show you. So let's go back to Invivo AI Cloud. So this is what you're going to see when you go to Invivo AI Cloud, and I've already made a video to give you overview. So this one is step-by-step how to analyze your qualitative data. So first, what you have to do is to go to New Project. You click on New Project, and this is what you're going to see. It's going to ask you about the background information about your study. Why? So this is going to use that information to help to make sense of your data and suggest code for you so that you can review. So this step is very important. So in this case, I'm going to provide the title of my study. So I will just say, Experience of Burnout Among Primary Healthcare Physicians. So that's my topic. Then it's also going to ask you about your research problem, the background information about your study that will help the system to make sense of your data. So in this case, I have the purpose of my study. But in your study, you can provide a problem of the study, the purpose, and maybe the objectives of the study. You can provide all this information. But for this demonstration, I'm just going to provide the purpose of the study. So when you finish, you go to Continue. So based on the information that you have given to the system, it will suggest research questions for you. So here are some of the research questions that the system has suggested to me. So what should I do? You can add it to the questions that you want the system to know, or if you don't want, you can just leave it alone. So I think this one is consistent with what I want to find out. I want to find out the causes of burnout. So I click on Plus. And I'm mostly looking for solutions of burnout. So this one can also be part of my research question. And I think I'm not interested in this part because it's not consistent with the purpose of my study. And you also have an option to add any questions that you have. Let's say you don't like the way that maybe the second research question was constructed. You can delete. And then create your own. Let me bring my own here. I'm going to copy and paste it here. And then this is my research question. And then you click on Add Question. So the next step is to click on Continue. And then it's asking you which kind of data analysis method or strategy you want to use. So here, the only option here is Initial Topics. Reflexive thematic analysis will come later. So we don't have an option for that. So for now, we have only one option. So we just leave this place alone. And you go to Participant. Who are your participants? You don't have to provide this information if you already provided that information at the beginning. But if you want, you can do that. So this is my participants, Healthcare Physicians. And I could go ahead and click on Continue. Here, if you are working with your team, you can provide the email address of your team so that you'll be able to have access to your project or the projects that you are working on. So you can provide your email address and then you click on Add. If you don't have, you just can skip or you can click on Skip if you don't have any information to provide here. So at this point here, if you want the system, which is the AI software or AI model to go and search for articles for you that are related to your topic, you can check this one. If not, if you don't want the system to do that for you, then you can uncheck. You have to be very careful with this one because if you want it to also affect the credits that you have. So you always have been given a credit every month based on your plan to use. So be careful how you use the credit because when it's finished, you may have to also go and buy more credits. So if it's not needed, you don't have to use it. So this one, if you don't want the AI to search for articles that are related to your topic, you don't have to choose that. Say here, I don't want that. But if you want, the system can search for literature for you, and then it will be part, that literature will be part of the analysis. It will code all the literature as it's also coding the data that you have given to it. So when you finish, when you finish, you go to Continue. So this one, this is where you upload all your transcripts. So your transcript, you can upload a transcript in the form of PDF or a Word document or a text. Or if it's in a video form, you can also bring that information here. If it's audio, you can do that. And if you have a video or audio, you have an option to, for the system to transcribe for you, but you have to pay for it. It depends on the credit that you have. If you want to know your, the credit that you have, you go to your account and click on AI use, and then your credit will be shown here. So as you can see here, 1,472 credit is left for me for this month. And also 400 credit is left for me to, for the system to search for literature for me. And I have 10 hours of transcription. So if I have audio and I want to transcribe, I transcribe, I have credit to use, right? So you always have to check your credit before you take any action here. So I'm going to go ahead and bring all my transcripts. So let me bring the five transcripts. So I drag and drop them here. So I have my five transcript here. Now you can also manually bring any literature that is related to your topic, right? If you want, you can drag and drop. If you don't have any literature and you don't want the system to analyze any literature, you just leave that place alone and go ahead for the system to read and also develop or suggest quotes for you. So you see here is Aura read. So Aura is the AI system that is going to read your transcript and then provide you an output for you to review. So I'll go ahead and click on that. So it will take a few minutes for the system to do it for you. And then you have your information. So this is what the system is doing. The system has gone through all your transcripts and then summarize each of the transcripts and put them in a memo form. So if you want to view, let's say P5, what the system has done, you click on this one and then you see that the transcript is here on the left side and the memo is on the right side. And then the memo has the summary of the transcript and also the key topics. So you can see the key topics that was taken from the data and then any relevant information that it has observed, you can review that. So key quotations that are related to your research question can also review the key quotation. The system hasn't coded your data yet. This is just a brief information for you to know what the data is all about, right? It's just similar to when you are doing a manual coding, you have to familiarize yourself with the data by reading and also getting a big picture of what participants said. It's very similar. The system has read and this is what the system has observed. And the good thing here is that you can always edit if you want to. So you can see here that I can easily edit information here. You can always do that. You can review, add more topics or edit and do what you want to do with the information. When you are okay with what you see, you can click on the X and then you go back. You can review the next one if you want. After that, you can go to continue, organize, you click on that. And then now is the time for the system to really look at your transcript. Now the system wants to create demographic information. You have two options. One option is to manually create that demographic information or the second option is to let the system go through the transcript, especially if the transcript has all the demographic information. The system can analyze it and then provide you that demographic information. What I would do now is to click on analyze with Aura. And then the system will go through the transcript and then provide me all the variables and also the cases. So the cases are your participant. I have five participants here and then it has gone through and you can see that it has provided me the age, the role, the agenda, whether the person is working in his or her private practice, location in the U.S. and some of the information didn't show. This is why you always have to go back and review the transcript and trying to make sure that this information here is correct. Right. And if you don't like or you don't want any of the variable like organization, you can click on the X and then delete that. So if you don't want country, you can click on X, delete that. So you can always delete and also add a variable if you want and can also add a new participant or a case if you want. So you always have to review and then make sure that that information is right. If you review, you can also put the new information if this information is not right. The role, they are all physicians, so I don't need this one. So I click on X to delete. So that's what you're going to do. You can review that or you can manually put that information here if you want to manually do that. You don't have to click on analyze with Aura. You can just click on add cases, add a case or add variable, and then you'll be able to bring that information here and then run the analysis. Right. So you can do it manually or you can just allow the system to do that for you. But you always have to review. You always have to be ahead of the analysis. You have to be leading the analysis. The system suggests and you have to review and make sure that everything is right. Right. Because it can give you wrong information. So the next one that we have to do is to continue initial topics here. Before I talk about continual initial topics, I think what I would do is let me see. I already have demographic information here, so I'm going to make sure that the information here is right. So the P1 is African-American. So I click here and choose African-American. OK, so I'm done with that. The system didn't bring years of experience. Let me try and do it so that you see how it's done. Add variable. I'm going to say years of experience and then it's going to be numbers. So I'm going to click on add. And then here I'm going to put the P1 is 12 years, P2 is 20, this P5 is 30 years, P6, P4 is 2 years, P3 is 35, and P2 is 25. Right. So you see how easy that is. You can just provide your demographic information here. And when you finish, you go to continue initial topics. So this is where the system is going to go through all your transcripts and suggest topics for you, for you to review and making sure that everything is right. Topics that are based on your research questions. Right. So I click on that. There's an option for you. Do you want AI to generate initial topics so that you can review? Do you want to provide your own topics for the system to go and look for quotations or evidence in connecting to your topics? Do you want both AI doing the generating topics for you and you also creating topics for the system to go through? Let's try the first one. So I click on start. Based on the number of transcripts or the quantity of the data, it will take some time. Here you can see that it's very fast. It takes seconds for me to get my result. Right. So now this is just an overview here and also the gap the system has identified. So you can review and see whether this information is correct. Right. But what is more important are the topics and also the definitions of the topic. Right. So looking at this, this is one of the courses of Bernhardt. Right. And then this is also definitions. So you have to decide, is this one addressing my research question that I have? Yes. Administrative burden is part of the courses of Bernhardt. If it's addressing, I will check, I'll confirm. One of the reasons I like about this software is that everything it provides to you, they are all suggestions. You have to review and accept or reject. As you are accepting and rejecting, the system is documenting your actions. So you'll be able to communicate the main action that you took as you are using AI to help you to make sense of the data. So as you can see here, changing workplace demands. So sometimes you have to look at the definition and make sure you also have an option to change the topic or the code, the name of the code. You can always click here and change it if you don't want it. Right. I think it's quite accurate. So I'll check here, confirm. And so you go to and just confirm and also reject if you don't want. Let's assume that this, I don't like this one, I reject. And then to be deleted, early career stress, I will accept that. So this one is more about the solution of burnout is the second research question. So the participant talk about attending church and also volunteering. So faith and community. So I will say practicing faith and engaging with the community. So this is what I want it to be. So I type and accepted that. So the option here is that you review, you review the labels and also the definitions and see whether it's really addressing a research question. If it's addressing a research question, you confirm. If it's not, you can reject. You can also adjust. You can confirm, you can adjust and also confirm. Right. So that's what I've done here. Family time. Family time. I will say having time with family and then I check here. So I expectation. I check that. I check this one too. I check that. I check this one. Personal coping. Okay. So anything that is quite personal, like prayer, reflection, positive thinking. Okay. Personal coping. What about, what do you think? What should I put here? Should I say utilizing or using personal coping strategy? What do you think? Is this one? Okay. Let me hear from you. Check here. And as I'm doing this, if you want to learn more about how to use this software, you can contact me. We can meet one-on-one and I'll be happy to provide you support. And it's quite straightforward. We can have a conversation and address any question that you have. If you want me to come to your organization and present or train your colleagues about how to use software like this to help to analyze your data, how to make sure that you evaluate all the AI output, making sure that they are correct. I'm the right person for you. I'll be able to show you step-by-step unique techniques that you can use. It's very simple and straightforward to help you to make sense of your data. I'm happy to help. So contact me and also don't forget to subscribe to my channel. So I'll go ahead and accept the rest and then we move ahead to the next stage. So a question may come up. What if the system gave you not a lot of codes? What should you do? There's another option. You can go to, you can chat with your document using their system and it will be able to create more codes for you or topics. Or let's say you change your mind and okay, maybe I have some five topics that I want the system to go through the data and identify information that are significant and connect to my topics. What should I do? There's an option for you. You can just chat with the system, tell it and it will do it that for you. So here are the SEPs that are linked to the codes that you have reviewed. Now, so you see all the SEPs here, here you don't have an option to change anything, just to review. So you can see the administrative burden, we have two quotations linked to that. And then we have four quotations linked to having time with family. And then you can just review and see the kind of evidence that are connected to their data. Okay. What if you want to take some of the quotations out? There's another option that I'll show you how to do that. So I'll go ahead and click on done. And then this is what you're going to see, right? So I finished my coding, right? But we haven't finished yet because we don't have your themes, right? But I finished the coding, I have my five sources here. I have 12 codes that was developed, and then you can go to map. So the map will show you all the codes and how they are connected. So how they are connected. So you can easily double click on any of the codes and review the quotations that are linked to the codes. And then you can, you know, if you want to know more, you can always click on open source and it'll take you to the source, right? So as you can see here, it's taking me to one of the sources. So let me close here and go back again. We come back to this. So let's say you want to know more about this code, you double click on it. And then you see that this code is linked to other codes, right? And then if you want to know the source, you click on open source and it'll take you to the document. Here too, you can even export this document and it will have all the codes connected to it. Let me export it and let you, so that you can see what I'm talking about. Click on export document and let me open the document. So this is what you're going to see. So as you can see here, all the codes that are connected to the document, they are here, right? All the codes. So you can review them and maybe send it to your supervisor or colleague. If you want the person to review that. Aura is the name of the AI tool that run the analysis, right? So let me close it and go back. And here too, you can always edit. If you want to edit here, you can also, you can edit by clicking on it and typing in the information here. You can also manually code here. You can select information that is significant, and then the system will suggest a code for you. So to click here. So you see what I did. I first click on code, select the place that is significant, the system will give you three options. You can type the code, you can let the system suggest a code for you, or you can use existing codes, right? To do the coding. So as you can see here, you can even manually code your data here. So let me close it and go back. And you can also export this visual representation. This is showing that, it's just showing the thicker the connection, the higher the degree of connection between the codes. I'll give you more information concerning that. It's quite useful when you have arranged them based on the research question. This one is difficult to explain, but let's go to analysis. So when you go to analysis, this is what you see. These are all the codes that were developed, and then this is finding out the connection between the codes and the variables. So in this case, we have age, and then you can see that administrative burden is one here. This means that one quotation is linked to this group, age group, and also linked to this code. And if you want to know more information, you can click on this one, and then it will show that code here, and you can even go to the source too, right? Click here and go back to the source. So you see how things are so connected. Let's say I want to change it to gender, and I want to see in terms of the codes that I have here, what is happening with how are they connected to the gender, right? So as you can see that here is two, and the early carrier stress is two here. You click on that, and then you can see two quotations here. So now this is how you can also export if you want to export this one. You can also click on case here. So this one shows the number of cases. So let me show you. The number of cases means that the number of participants that are linked to a specific code. So you can see there are one participant who is a male is linked to this one, and one participant who is a female is linked to this one. So SFs is a number of significant information that was extracted and linked to the group here, and also linked to the code, right? So if you have any other questions, let me know. Put in the comment section, I'll be happy to address them for you. So code co-occurrence. So this means that this is like, is there any relationship between the codes, right? So you can see that workload pressure has a connection with time with family, right? And it's two here. So as you can see, they're showing the relationship. And then if you want to know more about that information, you click on the number, and then the quotation will be under here for you to see. The same thing if you want to know the cases, what participants are linked to what code, you click on cases, and then that information will show here. You can see here that three participants are linked to this code and that code, right? So that's how you can also explore this one. If you want to deselect some of the codes, you can do it here. You can deselect and it will not show on the table, right? So you have a lot of options here, and you can export if you want to export. The next one is SFs. So you can see that if you want to see all the SFs, you can click on SFs and get all that, you know, all the information the system has extracted. And then if you want to know the source, you can go to open source and then it will take you there, right? So what next? Codebook. So let me show you the codebook. So as you can see here, this is the codebook. So the codebook has all the codes and also the definition of the codes and a number of SFs. You can export it if you want to, if you want to search for a specific code, you can search for it here and then it will be able to show, right? So what next? Concept. Concept here is empty now, but you can, this is where you can group your codes and that specific concept. So think about concepts as teams, right? If you want to group your codes into team, this is the place that you can do it. You can manually do that. You click on create new concept and then you'll be able to select specific codes that are linked to the concept. So let me give you an example for you to see what I mean here. So I click here and then I want to say, do it this way. I would say team underscore research question. I just want to indicate that it's research question one and maybe I'll type having high workload as you can see here. And what I have to do is I can select the ones that are linked to having high workload. So after selecting things that are related to this theme, you click on save and then it will appear here, but it's another option. You can chat with the system and a system can go through and categorize the codes into forming themes for you. You can do that for you, right? Should we do that now? Maybe let me show you. So go here and say that, can you categorize the codes to develop themes addressing both research questions, right? And a system can do that for you. You can type it in, but here, let me use a different software to dictate for me so that I will be faster, right? Can you examine each of the codes and categorize them to develop themes addressing the research question one and two. So for each of the themes, always start with theme underscore RQ1. If it's addressing the research question one, if it's addressing research question two, create theme underscore RQ2, and then categorize them so that we can get a theme addressing the research question one and two. So that's what you're going to type. Because I want to be fast, I just use a system to transcribe and this is what I was saying, right? I have the information, you can sometimes edit, sometimes it doesn't get everything that you said. So I will click on enter, right? So you see here manual. Manual means that before the system takes action, it will ask you for you to approve. If you say automatic, then a system will do some changes without prompting you, right? So I like manual because I want to see what a system is doing. So I go ahead and click on enter and let's see what it will do for us. So it's reading the code book and this is what a system provided for us. And then if you click on approve, or you can also edit too, right? You can here, you can edit and then make some changes before you approve. That's what I like about this system. You always have an option to review, reject or edit to modify it or approve, right? So let's approve and let's see what it will do for us. So now that I approve, this is what I got. So now I'm going to ask the system to put everything under the concept, right, window. So you just type it in. It doesn't have an option for you to talk to it, but I'm using a different system so that it will be faster. Can you put all the output about a theme into the concept window? So this is what I was saying and then sometimes I have to make, so let's see, it looks like the system didn't do exactly what I want, but you can always, you can always make edits when it comes here, right? So here it's asking you what it has to do. Yes. So it's also providing a suggestion. Yes. This is what I want a system to do for me. Also it's asking whether I want it to be, the format to be changed from this one to that one. I will say, okay, let me approve and let's see what the system will give us. So it looks like the system gave me only one theme, addressing the research question, right? So you don't want that. So you can always further communicate with the system, right? So I can say that, so can you make a revision of what you did for me? This is what I want you to do. Can you generate at least three themes for each of the research questions? I'm getting my transcription here. So you always can change your mind. You are having conversation until you're finished. I don't want this one to turn on, so I will reject that. Okay. Okay. So you see that the system has generated three themes addressing my research question, right? And then I will, if you want, I can make the revision less by organizing the concept into these six groupings. Okay. I will just say, go ahead. Okay. So it has done that. You can review and make sure that everything is right. If you are okay, you click on approve, and then the system will make the changes here. Approve. Approve. So you see how when I'm approving, it's still, it's making some changes here. Approve. You can put it to automatic and let's see, I'll put it in auto so that it will be faster because this means that you have to approve each of them. So you see that as you are doing, it's doing things here for you. You can always go back here to even make changes. If you want, you can click here and delete a concept, or you can click on a concept and add more codes to that. So you can, you always have an option to review and provide that information. I will say, just go ahead. You just have to respond to what the system is saying, and you always have to review, refine the research question two concept. Maybe refine for now. Okay. I don't want to waste your time, but I think that the learning here is that you just have to continuously communicate with the system until you reach what you want, right? And that's what you have to do, review, and if it's not making sense, you have to delete or ask the system question to clarify, or you can code it manually yourself. You can also manually create your theme here and then group them based on what you want. When you go to audit trail, it has all the interaction that we have had. So you see that I have hundred interactions, right? Including what I rejected, what I confirmed, everything is here, right? And so this is the pure audit trail. You can show it to your supervisor what you did and what AI, the role AI plays, if you want to show that. And there's a place to, when you go down and press of one of these, it shows you the model that the system used to run the analysis is GPT 5.4 mini. And also the system prompt, the prompt the system use as you are asking it to summarize your data for you. So this is a prompt the system use to do the summary, and you can review if you want to. So you see how transparent this process is. You have all this information there. The time is also there, the time that you did, how long it took, everything is there for you to review. You can export it if you want to. And then you can use this as a way of, if you want to use it to help you to communicate what exactly AI did and what you did in the analysis, you'll be able to provide that information. So we finished with memo, we go to cases. So, so cases is just showing the demographic information, right? This is what I have for you. I hope this one was helpful. Don't forget to subscribe to my channel. Thank you so much for your time.