ChatGPT & MAXQDA: Qualitative Data analysis with AI

In this video, I show you three and a half ways how ChatGPT can help you with your analysis of qualitative data in MAXQDA.

Here are the prompts I used in the video:

1) summarizing an interview:
Write a summary of the following interview. Use an academic language that is commonly used in social science research. Describe: 1) Who is this person? 2) What does this person contribute to the research question “How satisfied are the research participants with their work-life-balance?".

2) summarizing a code:
These are text segments from various interviews about work life balance. The segments belong to a category called "child care". Provide a summary of the text segments that belong to this category. Highlight the commonalities among all statements but also describe some exceptional cases.

3) create a code system based on color codes:
These are important text segments from a qualitative reasearch project about Work Life Balance. Considering the content of these segments, what could be good categories to categorize these segments for further analysis? List them as an ordered list using hierachies of main and subcategories where appropriate.

4) define a code and provide feedback:
These are text segments from various interviews about work life balance. The segments belong to a category called "child care". Provide a definition for this category, provide a good example from the input segments and point out any segments that might not fit this category and explain why they do not fit.

Recommended viewing:
I regularly watch videos by @aiadvantage and highly recommend his e-book "ChatGPT Cheat Sheet"

Update:
Since creating this video MAXQDA has released version 22.6 which includes the great & new AI Assistant. This allows you to create AI-generated summaries within MAXQDA. All precautions mentioned in the video still hold true. Also, copying and pasting from/into ChatGPT allows you to customize your summaries more.

Contact:

You can also find me on LinkedIn

More about:
my freelancing services (English)

the three deadly sins of MAXQDA that you should definitely avoid! (German)

[muellermixedmethods]

After I had finished editing this video, I noticed something. Here is what ChatGPT said about it 😅

me:
how bad is it, if I say ChatGTP instead of ChatGPT in a YouTube video?

ChatGPT:
It is a minor mistake, but it might confuse some viewers who are familiar with the correct name "ChatGPT". However, as long as the rest of the content in the video is clear and informative, this mistake should not significantly impact the overall quality of the video.

[mikahundin]

1. Using ChatGPT to Summarize Coded Content: The presenter suggests that after reading interview data and doing initial coding in MAXQDA, researchers can copy the text into ChatGPT and have it generate a summary. This can save time and effort compared to manually writing a summary.

2. Summarizing Coded Segments: Similarly, researchers can export coded segments from MAXQDA into a spreadsheet, then copy that into ChatGPT. The AI can then provide a concise summary focusing on commonalities as well as any exceptional cases within the coded data.

3. Generating Code System Ideas: By taking the content of important coded segments and providing that to ChatGPT, the AI can suggest new code ideas that may help describe the data more effectively. This allows the researcher to develop a code system grounded in the data but enhanced by the AI's insights.

4. Defining Codes with ChatGPT: For any existing codes, the researcher can copy the coded segments into ChatGPT. The AI can then generate a definition and example for that code, as well as identify any segments that may not fit the code properly. This helps the researcher refine and improve their code definitions.

5. Maintaining Researcher Control: The presenter emphasizes the importance of the researcher maintaining control - only using ChatGPT for summarization, idea generation and definitional tasks on content the researcher has already reviewed. Blindly trusting ChatGPT's output is risky due to potential hallucinations.

6. Internal vs. Formal Use: The presenter recommends using ChatGPT's capabilities for internal tasks like writing memos and summaries, but cautions against directly incorporating any AI-generated text into formal research outputs, as this could lead to academic integrity issues.

7. Embracing the AI Assistant: Overall, the presenter sees great potential in using ChatGPT as a powerful research assistant within the MAXQDA qualitative analysis workflow, as long as the researcher remains vigilant and maintains appropriate control.

[abcufcanada]

I had to laugh at first because of ChatGTP, but your prompts really woked. so thanks a lot <3

[Louisianish]

I use MAXQDA in my job and was looking for a good video to share with colleagues if they ask about applications of AI in qualitative data analysis. And this video is perfect!
AI shouldn’t be a replacement for human qualitative analysis, but it has the potential of being a powerful tool and assistant that can help make the analysis process more efficient, as long as people don’t rely on it without confirming the text it’s producing. As you said, hallucination is a common issue with large language models and other types of generative AI. If you’re heading up a research team and are considering incorporating AI into your workflow, it’s important to understand how this technology works before you decide to get rid of half of your research staff because you assume the AI will just do all the data analysis for you and do so with 100% accuracy.

[RobinMario]

The main problems when considering whether or not to use A.I. to analyse qualitative data are privacy concerns. Safe use mens that the a.i. must be able to run nativly on your computer (offline-mode). Using chatgpt to analyse actual interview, especially those containing sensitive infirmation is highly unethical.

[abdo.]

He probably also used chatgpt to write the script of this video