How AI Can Help With Mixed Methods Research Design (Full Transcript)

[00:00:06] Speaker 1: Hello, if you are interested in using AI for your research, using mixed methods, quantitative and qualitative, then you're in the right place. My name is Anmoye Brown, and I'm joined by Dr. Philip Adu, who has written several publications on qualitative research methods. So please, we would like to know where you're joining us from, wherever you are in the chat, just to write your name and where you're located. So to start us off, Dr. Adu, what are mixed methods research? What is mixed method research? You know, let's not assume.

[00:00:51] Speaker 2: Yeah, so thank you for your question. So mixed method means that you are using more than one research methodology. Most of the time you are using quantitative approach and also qualitative approach in a single study. That makes it a mixed method. So you always have to think about four things, right? So the first one you have to think about is sequence, which sometimes people call it timing. Are you collecting qualitative and quantitative data at the same time, or you are doing it sequentially? Are you collecting qualitative first and maybe end with quantitative? So the timing is very important. You also have to think about theoretical perspective that you are using. What kind of theory that philosophical paradigm that informs the approach that you are using, which is the mixed method approach. You also have to focus on the integration. How are you going to mix the two methodologies? Are you going to mix it during the data collection or analysis or during the data interpretation of the findings, right? Which of the approach would take a major role or will play a major role in your study? So these are the things that you have to think about when you are doing a mixed method study. Collecting qualitative and quantitative data within your study, but you also have to think about all these components as you are doing mixed method research.

[00:02:19] Speaker 1: It's very interesting, Dr. Odu, that you spoke about timing and the theory and so forth, but if I am having, I'm doing a piece of research or an evaluation and I use interviews and focus groups and I use a survey and that's it. Is that a mixed methods approach?

[00:02:40] Speaker 2: You have to think the name mixing, right? So you are not only collecting qualitative and quantitative data, you have to think about how are you going to mix the two, right? So just collecting qualitative and quantitative data doesn't necessarily mean that you are doing a mixed method study. You also have to think about, do you have qualitative research question? Do you have quantitative research question? Are you looking at the phenomenon from two methodological point of view or are you looking at the phenomenon? One is looking at one part of the phenomenon. Another thing is looking at the second part of the phenomenon. One example is, let's say you want to measure the satisfaction or you want to study the satisfaction, student satisfaction in a classroom. You can do a mixed method study, right? The quantitative will focus on measuring the level of or degree of satisfaction. The second one, which is quantitative, will be interviewing them to find out the experience of satisfaction. So you see how you have to be very systematic in terms of choosing the approach. Just choosing two data collection strategies doesn't necessarily mean that you are doing a mixed method study.

[00:03:58] Speaker 1: So very interesting. So from my understanding, a mixed method study goes beyond just the data collection method. It also looks at your research question and what aspects you're measuring. So I think that's a very important takeaway. It's not just saying, oh, I'm doing a focus group and a survey. So it's mixed method study. There's much more to it, such as the timing, are you going to do it at the same time and your theory. Okay. Thanks, Dr. Odu. I'm going to welcome John. John is from Kampala and I also see Benith in Uganda and Isaac, I think the African delegation came out today.

[00:04:42] Speaker 2: Yes.

[00:04:44] Speaker 1: I can see them. Prosper from Malawi. Thank you so very much for joining us. And if you have any questions or comments, please put them in the chat. We want this to be an interactive session so you can get the most out of it. It's not a script just between Dr. Odu and I. So Telahun, please put your question on explanatory in the chat so we can answer it. And Trudy is speaking about doing a convergent study. Does that ring any bells, Dr. Odu?

[00:05:18] Speaker 2: Yeah. So the convergent study means that you are collecting qualitative and quantitative data and you're trying to bring them together. Most of the time it's also called concurrent triangulation design, mixed method design. It's concurrent. It means that you are collecting qualitative and quantitative data at the same time and trying to analyze them separately. But you're also going to look at the phenomenon from two methodological point of view. That's where the triangulation looking at. An example is let's say you want to explore the experience of participants, right? You can collect quantitative data, find out their perception about the experience, and then qualitative data trying to talk about the lived experience, expressing themselves, giving their thoughts, their action, their decisions. And then you are going to look at the phenomenon from these two methodology point of view. So convergent. They are coming together to explain or help you to understand the phenomenon.

[00:06:24] Speaker 1: Okay. All right. And welcome, Trudy. Welcome. Thanks for your great question, Trudy's in Barbados. And question coming in, how about triangulation? Is triangulation a method? Is it different from mixed method?

[00:06:41] Speaker 2: It depends on how you label it. We saw that under mixed method, we have different types. And one of them is concurrent triangulation mixed method design. So concurrent means that you are collecting qualitative data and quantitative data at the same time, right? And then you want after analyzing both, you want to look at the issue, interpret the findings and looking at, you know, interpret the findings and helping you to understand the phenomenon. So it's like looking at the phenomenon from two different angles in terms of methodology. So that's where the triangulation comes in. So normally, you know, sometimes you really want to get into the quality of the results, right? And we all know that using one methodology might not be able to reach that kind of quality. So you have to look at it from other methodological point of view and try to say, okay, are they all saying the same thing, right? Is there any kind of difference? If they are saying the same thing, this means that the issue exists. All this information is really reflective of what participants were giving us, right? So triangulation is also one of the type of mixed method design.

[00:07:57] Speaker 1: Okay, great. Thank you for that interesting question and answer, Dr. Ado. But we did say this live stream is about the use of AI. And since we have laid the background now, we have defined what it is. Oh, we have someone on stage.

[00:08:15] Speaker 2: Yes. So can you ask us a question, Tila Ham? I'm hoping I've mentioned your name right. I think we cannot hear you. So maybe you have to work on your audio. Maybe let's go to the next one. I think we have... Can we hear you? Or can you hear us? Okay. I think maybe later. Okay. Okay.

[00:08:59] Speaker 1: Any other questions? Yes. So we're getting into the AI aspects because this live stream is about how you can leverage AI for your mixed method study. So Trudy has a question, and that is, how can she, I hope your pronoun is she, use AI almost as a research assistant to cooperate the themes from her thematic analysis?

[00:09:29] Speaker 2: Yeah, I think you can do that. I think that what you can do, you can use charge GPT. And you always have to, since it's an assistant, don't assume that they know everything about a study, right? So you have to introduce the study to the assistant. How do you introduce a study? Provide some background information. What are you studying? What is the purpose of your study? What is the research question? And also ask for clarification or ask the system that the system wants you to clarify anything about your study before you ask it to analyze your qualitative data for you. After having that conversation, then you can upload your transcript, right, and ask the system question. You can ask it to go through the data with the research question in mind, identify information that are significant, and extract all of them for you. You can also ask the system to develop labels or codes based on the significant information. You can ask the system to categorize the codes to develop themes, right? Then when you have a theme, you can compare the themes that you have gotten from the AI to what you have done. And you know, sometimes it's on the AI can give you some suggestion about how to frame some of your themes, right? There might be something that you didn't look at based on the research question. The AI kind of points that information out. You can even go further and say that I did manual coding, and these are my themes that I came up with. Can you compare my theme to your theme and then try to reconcile them? And the system can do that for you, right? So it's like the help is limitless. You can ask any question and then see what you're going to get and then use the information that you think is going to be useful. Yes, so it can be used as an assistant for you for the qualitative portion.

[00:11:24] Speaker 1: Okay, great. So you spoke of the qualitative analysis, right? But how can AI tools help in integrating both the qualitative and quantitative approaches in research design?

[00:11:39] Speaker 2: Yeah, so I think, you know, we just have to start from the proposal stage, right? Imagine you have a topic and you know that you're going to use a mixed method design, but you are not sure because we have different kinds of mixed method design. So you can have a conversation with Chachi, Piti, or Cloud AI or any AI tool. As I said, the beginning is giving the system some context. What is your study about? What do you want to find out? What is your population? And giving the system all this information, and also ask the system to clarify, it wants you to clarify anything. After that, you can ask it to generate or give you an idea about the research design, that mixed method design that is appropriate, the kinds of questions that you have to ask, and the system can help you to do that. So it's all about having clear communication with the AI tool, providing a context, and then continue to ask questions until you get what you want, right? And also, the good thing is that I've created a tool, AI tool, that can help you to determine the right methodology in terms of the mixed method for you. I can use it and then see what you're going to get. So that has been very useful for a lot of my students, all my clients. So I can share that information with you so that you'll be able to use that.

[00:13:07] Speaker 1: Okay. So before you do that, you know, we have quite some room to cover. If you're just joining us, we're speaking of how to use AI to assist with mixed method studies. And if you want to see that tool that Dr. Adu is speaking of, that can help you select the right method for your study, just enter in the chat, share, share, and then we know that you are interested in having that tool. I see we have someone on stage who would want to say something. Jacqueline, can you hear us?

[00:13:51] Speaker 2: Yeah, I don't understand why the audio is not working now. Let's try another person and see.

[00:14:01] Speaker 1: Yunusa. Okay. All right. So keep your comments and questions coming in the chat. If you're interested in seeing Dr. Adu's AI tool that can help you to select the right method for your mixed method research, write the word share. Okay. All right. So we keep going on. Earlier, Dr. Adu, you were speaking of certain AI tools such as chat.c and cloud.ai that can assist with your qualitative data research. Are there any other tools that you recommend, especially for mixed methods?

[00:14:51] Speaker 2: Yeah, I think, you know, there are a lot. Apart from chat.c, we have the cloud.ai. I can share my screen and then I can talk a little bit about that first screen. Let me see. So cloud. Yes. So the cloud.ai, it's just like chat.gpt, right? You ask the system questions and you'll be able to get some good information, right? So you just, you know, as I said, try as much as possible to give a background information about your study, right? And then ask the system if it wants you to clarify anything before you ask them, okay, can you help me to determine the right research method approach? So you can use the cloud.ai. And I think that you can also use this tool that I've created and that chat.gpt is called mixed method decision, design decision tool, right? And I put the link in the chat box or the comment. And I can also share the document with you about all this information about what I'm talking about after our conversation. So when you go there, you can ask the system some questions. And I think I did something that I can share with you. So this is the first, let's go up here. This is the first question that I asked the system, right? I said, I'm doing a study about University of Ghana students. I didn't bring, let me correct this one. So student experience of online course. Before you proceed, is there anything I should clarify? So now you see that I've given the system some information about my study, and I want the system to make sure that it understands what I want to do. So the system asks me some questions. What is my research question? What is the mixed method design I want to use? What is the priority? And all this information, right? Then I was able to answer some of the questions. You don't have to answer all, just answer the things that you think that you'll be able to answer. And then the system was able to provide me research question that I can focus on, right? And so this one is a qualitative research question. Also, it provides me the mixed method research question, and then quantitative research questions, right? And then it also suggested the mixed method design. This one, it wants me to use concurrent triangulation design. And then it provided a justification why it is the best compared to other mixed method design. And then it gave me how am I going to prioritize in terms of which one is going to play a major role in your study, right? So it looks like because you are using triangulation, the two methodologies, they are all going to play equal role in your study. Because triangulation, you are looking from the issue from the two methodological point of view, right? So they all have to have equal role. Integration, how are you going to integrate? It's saying that you can integrate your both methodologies during the data analysis and also interpretation stage. So you see how this software can provide you an idea about your design in terms of the sequence, the type of design that you have to use, the priority, the integration, and also the purpose, right? And I think that's where this tool could be very, very helpful for you. And as I said, if you want to get access to it, I'll put it in the comments section. And if you have any question, you can ask me, and I'll be happy to address it for you, right? And there's another tool that I also recommend. It's called Social Studies Simulator, and I will put the link in the comments so that you can get access to it. This one is so interesting. So let's assume that you already have an idea about your mixed-method study, right? But you also have to make sure that you have anticipated the challenges that you're going to face and then things that you can action that you can use to overcome all these challenges, right? This is where the simulator comes in. It's just like you are doing a parallel study, but you are not actually doing a parallel study, where you tell the system, this is my mixed-method design. This is my population. Can you help me to make sense of the kind of data I have to collect? And if in case I call it the data, what kind of analysis should I use and what kind of results will I get, right? So the system is so intelligent that it creates artificial data. We call it synthetic data, and run the analysis, and then you'll be able to get an idea about the tools that you have to use, and then in terms of qualitative, the teams that may come up when you call it your data. So it's like you have done a little bit of synthetic parallel study just to get an idea about how the kind of information that you're going to collect from participants and how you're going to analyze and also present the result. And then it also gives you information about lesson learned, right? What lessons have you learned from this simulation so that you'll be able to take into that concentration as you are doing the actual study? So these two tools will be very helpful for you. The mixed-method one and also the social science studies simulator will be very helpful as you are doing the mixed-method study.

[00:20:52] Speaker 1: Okay, great. Thank you so much, right? And Conrad is asking how to avoid being marked as plagiarism. When using AI.

[00:21:03] Speaker 2: You said how to avoid what?

[00:21:06] Speaker 1: Being flagged as plagiarizing when you're using AI.

[00:21:11] Speaker 2: Yeah, I think so. I always say that use it as, first, you have to review the tool as an AI assistant. It's not doing the research for you. So it's like giving you ideas about what you're going to do. You're not going to copy and paste things, right? So it's just the same as meeting an expert, a methodology expert like me, and meeting me and asking questions, and I'll give you some ideas. Or you give me information about your study, right? And I provide you, oh, I think based on this information, I think maybe concurrent mixed-method approach would be the best, right? Or sequential transformative mixed-method approach would be appropriate. And you ask me, oh, what is sequential? And I give you all this information, right? So it's like information gathering. You are not going to copy and paste anything. You are just getting ideas so that it will help you to be well-informed as you are doing your research. If you look at the tool like that, you will not have issues with copy and pasting, right? And you just have to see it as gathering ideas for you to do your research.

[00:22:24] Speaker 1: Right. And this touches upon other ethical considerations, right? So what else should one keep in mind ethically when using AI for their mixed-method study apart from just copy-pasting without citation?

[00:22:41] Speaker 2: Yeah, I think that, you know, ethical where you have to make sure that when you collect your data from participants and you want to use the AI to do the analysis, you have to make sure that you take away all identifiable information that is in the data so that, you know, participant information will be protected, right? And also, you just have to make sure that this system is fallible. This means that it can make mistakes, right? So you always have to be critical. You don't take the information and run with it. You have to ask some questions. You have to let the system do some self-evaluation based on the output. Can the system reflect and give you some ideas about whether this information is appropriate, right? So engaging in that kind of conversation about self-evaluation is very important. And always you can cite a source, especially when you use ChachiPT to analyze your qualitative data. You can tell people about how you communicated with ChachiPT to get the themes, to address your research questions. So being transparent in the process will be also helpful when it comes to ethics and also doing research using AI too.

[00:24:02] Speaker 1: Okay, great. Thank you. Keep the questions coming and comments. And if you want more detailed information, we'll also be having a workshop where we delve further into some of these ethical issues and also to give you more examples of how to use AI. So Conrad is on stage. Conrad, can you hear us? Conrad? Okay.

[00:24:34] Speaker 2: Yeah, we are not hearing them. All right. So sorry about the technical issue.

[00:24:41] Speaker 1: That's fine. I think persons are writing, they're active in the chat. And one question is, how can AI assist in managing large data sets in mixed methods research?

[00:24:56] Speaker 2: Yeah. So I think that what you have to think about is you always have to think about how large is your data, right? And if your data is, I think they use it as that, they measure it based on like number of words or number of characters, right? And if you are using chatGPT, I think you'd be able to analyze your quantitative data and qualitative data. If you have, let's say, 10 interviews transcript, and then you have maybe a survey with about 5,000 or maybe 2,000 people, you'd be able to run the analysis. And I think that when it comes to, we call it context window, right? When it comes to how much information you can give to the system to get what you want, it all depends on the tool that you are using. And sometimes they change the requirements. So my suggestion is try to explore the tool with your data and see whether it will be able to take that information. And then if it will be able to take that information, you can ask the system some question to get what you want. But so far, based on my experience, chatGPT can handle most of the data that researchers use. And also the Cloud AI has bigger context window. If you want, you have a huge amount of data. Like if, let's say, you want to put in the number of pages, right? So if you have about 130 pages, right, in terms of if you want to translate your data into pages, Cloud AI will be very helpful. For chatGPT, maybe something lower, like maybe 120 or 100 pages, right? 100,000, sorry. 100,000 pages, you'll be able to use chatGPT. So it's all about trying it out and see. When it comes to interview, if you have more than 10 transcripts, then you have to upload your information in batches for you to analyze. So you have to upload, chatGPT allow you to upload up to 10 transcripts for you to do the analysis. So you can add, upload the first 10 and ask the system questions. And then you can upload the second 10 or the third or something. So breaking them into batches will be very helpful for the system to conserve.

[00:27:26] Speaker 1: Okay, useful tip. If your data set is huge, just upload them in batches. So that's a good practice when using any AI tools. And speaking of which, say you have survey, open-ended responses from your survey. How can you use AI to analyze this data effectively?

[00:27:53] Speaker 2: Yeah, you know, it's so easy now. chatGPT has a system called, you know, they have incorporated a system called Python programming, where it can, you know, when you interact with chatGPT, the system will, information will be communicated to another programming tool and analyze your data for you and then bring you your result. So, so that the information will be accurate, right? So you can easily use chatGPT, you just have to upload the open-ended responses and ask the system a question. And always, I always say, give the system a lot of information about the data. Don't assume that it knows everything, right? It didn't do the research with you. If you don't give detailed information about the data, it will assume and provide you with general information. So always make sure that, especially if you have data with qualitative, like open-ended responses and closed-ended responses, right? You have to indicate each of the variables, what is the definition of the variables. If the variable, a person gets five and another person gets one, what does it mean? You have to explain it to the system, the scoring. Sometimes you also have to think about some of the instrument that you may be using, some of the items in the instrument, you have to do reverse coding, right? So one will be chained to maybe five and four will be chained to maybe three, right? All this information should be provided to the system and ask the system to do the reverse coding for you before you do the analysis. So don't just upload a document and start asking the system a question. You can get wrong information, right? It's not like a magician. You feed, it's like garbage in, garbage out. The wrong information that you give to the system, you may get wrong information, right? So always be specific and provide all the background information. Ask the system whether it has any question for you, right? And then you can go ahead and have very good conversation and get your information that you want.

[00:30:10] Speaker 1: Right. So my key takeaway is AI does not relieve you of your duty to check, validate, and also to make sure that the information you give is correct then and the context. Conrad, you're back on stage. Are you able to hear us? Okay. I think not. Okay. Jaquim. All right. In the meantime.

[00:30:47] Speaker 3: Yes. Yes. I can hear you. Sure.

[00:30:51] Speaker 1: All right. Go ahead, please. Welcome.

[00:30:55] Speaker 3: Thank you. Thank you so much. Actually, this topic is interesting for me because I'm currently conducting a mixed method study. And before yesterday, since I'm analyzing my data, I tried to use AI. And the last point which Professor Philip just put forward regarding the need of providing enough data so that we can have the right information. I really experienced that because I just introduced that.

[00:31:48] Speaker 2: Oh, I think we lost maybe him. I think he will come back. So I think he was emphasizing on the need to provide a lot of background information before you ask a system to analyze. Right. And as I always say. Oh, go ahead. Yeah, we lost you. Hello.

[00:32:20] Speaker 1: Yeah.

[00:32:20] Speaker 2: Yes, we can hear you now. Yes. So what happened was that when I asked it to code the data for me, it was not able to code the data for me.

[00:32:26] Speaker 3: It was not able to code the data for me. Was that when I asked it to code the data for me, it coded the data not only from the participants, but also from me, from the interviewer. And it became very long for me to analyze. And it gave me another work for me to put it in my chapter about data analysis. So the point maybe what I've learned now from you is maybe I'll need to refine or maybe provide more information to the AI so that it can provide me the right content for my analysis. I think so.

[00:33:22] Speaker 2: You are so right. And I think that a lot of people make the similar mistake that you made. So especially the interview transcript, you have to indicate what you said like interviewer and what participants said. If you put everything together, the system cannot differentiate between what you said and what participants said. And that's where it analyzes everything. It sees everything as transcripts and it analyzes it for you and provides you all the information. So yes, it's so true providing some detailed information and also background information will be very helpful for the system to provide a rich result. So thank you for sharing.

[00:34:06] Speaker 1: Yes. Thank you so very much for your input. Is there anyone else to take the stage? If not, please. Okay. All right. So this was a question from Conrad. As researchers, we can collect our own quantitative data. Can we use ChatGPT to replace SPSS to run analysis?

[00:34:31] Speaker 2: This is a very good question. I think you can do that, but it all depends on your institution, right? It's taking a little bit time for people to accept the use of ChatGPT. A year or two ago, people thought that, oh, no, ChatGPT cannot be good in math. So it will provide you wrong information because it's just predicting the next word. But it has improved the system in such a way that when you interact with ChatGPT, the system also communicates with another system, which is Python, to run your analysis for you. Right? And also Python is very good in quantitative analysis to give you all the information. So when you tell the system that, oh, I have this data, can you determine the correlation between these two variables? ChatGPT doesn't address the question. It sends that information to Python programming. And the system runs the code for you and then gives the answer to ChatGPT and ChatGPT Connect gives you the information. So you see how, yes, it will reach a stage where people might not use SPSS anymore. They might use ChatGPT. If SPSS, they don't incorporate AI, like chatting with the system to run the analysis, maybe things might change. Maybe they are working on it. But for now, if, you know, your institutions are okay, you can use ChatGPT asset data or qualitative or quantitative data analysis to help you to make sense of your data. Right? And another one that, second one that I can also share is called Julius AI. It's also very good in analyzing your quantitative data. Right? So the good thing about this one is that it uses more than one AI tool. Right? It's used ChatGPT or we call it GPT4 or you use Gemini. You use Cloud AI, some of the models to help you to make sense of the quantitative data. Right? So, yes, it will reach a stage where if they don't improve the SPSS to incorporate AI, many people will not use AI to like Julius or Cloud or ChatGPT to make sense of their quantitative data.

[00:37:06] Speaker 1: All right. Thanks for giving us your thoughts on the future, what the future holds. Okay. So I think we're right up on time. I don't see any other burning questions in the chat. So, Dr. Adwoa, I give you the final word before we close.

[00:37:25] Speaker 2: Yes. So I will be sharing, let me quickly show you. Let me quickly show you this too. So I'll be in the comment section. You can search, you know, the platform that you are using, whether YouTube or LinkedIn. I'll be sharing this PowerPoint for you to, you know, whatever I talk about, I put it in about eight pages or eight slides for you to just get a summary of what we discussed. And if you have any questions, please let me know. I'll be happy to address them for you. And also I provide methodology consultation. So in case you want to get information about how to analyze your data with ChatGPT or any other AI tool, you can contact me and I'll be happy to address them for you. And also thank you for your time. And also thank you, Anne, for all the questions and also the moderating this successful event.

[00:38:25] Speaker 1: You're most welcome. I did say you have the final word, but I just have to pitch in that in October, we'll be having a series of workshops, October 21 to 23. There's the link there, the QR code, you can scan it. And this is where we'll go in more detail. I see a question coming in on how to use ChatGPT as a core searcher. We will give you answers to that question and much more. So join us for the workshop in October where we can go in more detail and take care.