I have just published an article (attached) that examines the use of ChatGPT for qualitative data analysis. In my opinion it would be difficult to use this approach for what Braun and Clarke (2022) now call Reflexive Thematic Analysis, because they rely heavily on an initial process of open coding, whereas ChatGPT uses a query and response format that addresses the content of the data more generally.
If you could use AI strictly for coding, then I suspect it would be much less disruptive than using it to replace coding. For two attempts at this latter strategy, you can look at the contrasting implementations of ChatGPT in ATLAS.ti and MAXQDA.
ATLAS.ti attempts to automate the entire coding process by using a single command from ChatGPT to generate a complete set of codes. I have tried this, as have several of my colleagues, and all of us came away disappointed at best. The problem is that the program generates hundreds of codes which then need to be checked and categorized by hand.
MAXQDA takes a very different approach by using ChatGPT to summarize various aspects of the coded data, such as all the content associated with a given code. The obvious limitation here is that you have to have your data fully coded before you can apply ChatGPT.
Maybe someone will find a use for ChatGPT in the coding process, but so far I have not heard of it.
When we talk about the Thematic Analysis (TA), we talk about data, tools such as MAXQDA and the steps to analysis. I think AI can help us in part of tools and maybe some data as well. But data gathering from real world is our priority.
- Badges
- Science topic
- Similar topics
- Correction