John Kevin/iStock/Getty Images Plus
Walt Whitman wrote, “I am large and contain many people.” In qualitative social science, this is true both as a celebration of what makes us human and as a warning about the limits of using artificial intelligence for data analysis.
AI can emulate the pattern discovery of qualitative research in the social sciences, but it lacks a discernible human perspective. This is important because in qualitative research, it is important to clarify the researcher’s position, that is, how the researcher is involved in the research, in order to increase confidence in the research results.
A technology like ChatGPT, trained on vast human knowledge, constitutes a many without a self, rather than a self with many. By design, these tools cannot have the single, explainable perspective needed to foster trust: location information.
For overworked faculty and students, using ChatGPT as a research assistant is an attractive alternative to the arduous task of manually analyzing large amounts of text. Although there are many qualitative research methods, a common approach involves multiple cycles of creating meaning within the data. Researchers tag pieces of data with “codes” that describe explicit expressions or implicit meanings, and through additional cycles group them into patterns. For example, if you analyze interview transcripts in a study on college attrition, you might first find codes like “financial need,” “first-generation status,” and “parental support.” In another cycle of coding, these may be grouped into larger themes centered on family factors.
Although this is an oversimplification, it is clear that this kind of pattern discovery is a key strength of today’s open AI tools. However, using AI in this way overlooks the influence of researcher identity and context in qualitative research.
There are four main reasons why jumping on the AI train too soon can cause problems for your future qualitative work.
Researchers are just as important as research.
Good qualitative research has something in common. It is a rejection of the concept of objectivity and an acceptance of the subjective nature of interpretive work. They acknowledge that their research is influenced by the researcher’s context and background. This idea of careful consideration of positions is not entirely standard across a wide range of social sciences, but it is gaining momentum. The rapid adoption of AI tools in research makes it especially important to highlight the complexity of how researchers engage with their work.
AI is not neutral.
We know that AI can hallucinate and generate false information. But even if it wasn’t, another problem is that technology is never neutral. It is always full of the creator’s bias and experience. In addition to this, AI tools combine and derive information from a vast array of viewpoints on the internet on a particular topic. If you agree that positional clarity is key to the credibility of qualitative research, then you should seriously pause before introducing AI to large-scale analysis in interpretive research. . Experts admit that we don’t know how AI will make decisions (black box problem).
The introduction of AI tools can have a negative impact on the training of new researchers.
Just as educators may be concerned that relying too much on AI early in the learning process may undermine understanding of the fundamentals, this also has implications for the training of new qualitative researchers. . This is a more important consideration than the reliability of the results. Manual qualitative coding builds your skill set and provides a deeper understanding of the nature of interpretive research. Furthermore, clarifying how you, as a researcher, will influence your analysis and being able to act on it is not an easy task, even for experienced researchers; It takes reflection and patience, but many may find it not worth the effort. It is nearly impossible to get new researchers to understand location without going through the process of manually coding the data themselves.
Unlike human researchers, AI cannot protect our data.
Researchers are not the only thing missing when using open access AI tools for data analysis. Institutions require safeguards for information provided by participants for research studies. While it is certainly possible to include disclosures in consent forms regarding the use of data within an AI platform, the black box element makes it difficult to provide truly informed consent to participants about what is happening with their data. you can’t. Offline options may be available, but they require computing resources and knowledge that are out of reach for most who would benefit.
So, is the use of AI in qualitative research reliable?
When used to audit research results, AI has the potential to act as a quasi-research assistant or add further credibility to the qualitative research process, but should be applied with caution in its current form. There is. Of particular importance is the recognition that currently AI cannot provide the context and situatedness needed for qualitative research. Instead, useful applications of AI in qualitative research include providing general overview information and helping organize your thoughts. These complementary tasks and similar tasks can help streamline the research process without denying the importance of the connection between researcher and research.
Even if you can trust AI, should you use it for qualitative analysis?
Finally, there is a philosophical discussion. If we have AI that can do qualitative analysis in an acceptable way, should we use it?Like art, qualitative research can be a celebration of humanity. When researchers’ self-awareness, important questions, and robust methodologies combine, the result is a glimpse into a rich and detailed subset of our world. What makes these studies worth writing about and worth reading is the context and humanity that the researchers bring to the table. As we reduce the role of qualitative scholars to AI prompt generators, our passion for investigating human experience may fade with it. Studying humans requires a human touch, especially in an open and interpretive way.
Andrew Gillen is an assistant professor in the Northeastern University School of Engineering. His research focuses on engineering education.
