We’ve said it before, and we’ll say it until the robots come home: We are bullish on AI in research.
Marvin has leaned heavily into advanced AI to add context to user conversations and to speed up qualitative data analysis. Technology can never replace the brilliance of a trained researcher, but it can help.
Co-Founder & CEO Prayag Narula recently shared his expertise about AI in research in an episode of the Rosenfeld Review.
Takeaways from AI’s Role in Qualitative Research
- How we already use qualitative research in business
- Technology legitimizes qualitative research processes
- How AI augments research and make it easier to uncover insights
- AI challenges of the past vs AI challenges of today
Skip over to listen to the podcast, or read on for snippets of Prayag’s side of the conversation.*
Decision-Making with Qualitative Research
One thing that really frustrated me was that qualitative research is one of the primary drivers of decision-making for all organizations. And I would argue it’s the primary driver of decision-making for anybody on a personal level as well.
All executives, companies and organizations take decisions based on the conversations that they have had — the qualitative data. But there is no centralized place where people are looking at this. Even if the research is being done with the sole objective of making decisions based on qualitative research, that’s happening in multiple places in the organization, not just design research.
A big part of pricing research is qualitative research, and a big part of strategy research is qualitative research. You talk to any of the big management consultants, they spend most of their life doing qualitative research. Sometimes they call it qual research, sometimes they don’t. But at the same time, there is no scientific method that’s being applied at an organizational level toward that qualitative research.
One of the prime drivers for me is that all research is valid, all research is scientific. It’s not oh, that’s just that one is anecdotal and quant research is more data, quote and code. That’s not the case in social sciences. We have been doing qualitative research as a primary method of science discovery for six decades.
But even today, if you ask most people in an organization (even sometimes the researchers!), they’ll go on to say data and quantitative research is what’s “right.” And I want to change that.
Qualitative Research Methods Are Valid — and We All Use Them
I want to change this thinking that qualitative research is anecdotal or not scientific.
I call it the “tyranny of math.” When we talk about math, then it’s valid. If it’s not math, it’s somehow not valid.
One of my research friends put it, he said the best way to make your research team more valid is to add two decimal points. So instead of saying six out of ten people think this, you say 60.0% of people think that.
Let’s look at Oracle for example. In the CEO’s biography, he talks about how having tools that Oracle built made the processes seem much more legitimate. He was of the opinion that processes need to follow software instead of software following processes.
I hope that that’s the case now with qualitative research. And it’s not just tools like Marvin. There are other tools in the market that are giving qualitative research much more validity.
Having a software stack that prioritizes qualitative research is extremely important in making it valid and breaking through the tyranny of math. And honestly, it also makes it a lot more useful. Your data is valuable, and you need a tech stack dedicated to it.
Introducing AI to Your Qualitative Research
I am very excited about the use of AI technologies in research like ChatGPT. We have already started to use them in Marvin with automated tagging, notes and synthesis. It’s really, really powerful.
It certainly won’t replace human researchers — we will always need trained researchers to conduct good research.
But it gives them another layer on top of everything that they are already doing. It gives them another set of tools to make their work easier, but at the same time, it gives them additional context: “here is what AI says about this research.”
You have another data point to use. Obviously, AI is not going to engage with the data like the researcher would, but it does give you a superficial understanding of the bigger picture, and it could be really helpful in some cases.
AI would give you one interpretation of the data that you might not have considered.
The big advantage of using AI as your copilot and having AI be working right next to you and assisting the researchers is this sense of, “am I missing something?” Because honestly, one of the biggest fears of research is there is an insight that’s staring in my face, I just didn’t look at that data or just never asked that question.
And AI doesn’t completely remove that, but it does give you an additional sense of whether there’s an insight that you never considered. What are the odds? I don’t know. They’re probably low, especially in the early days of large language models. But it’s definitely time to include that kind of technology into our research practices.
AI in Research: The Next Generation
AI challenges have shifted in the last five years. We used to worry about nuances of multimodal qualitative research and wonder how AI answers questions based on that?
Large language models are trained on so much data — they have something like 4 billion data points in terms of their training. So they have the concept on a large part, if not all, of human knowledge. But how do you take that and limit it only to your own research?
That seems like an easier challenge to tackle than the one from five years ago. So I’m very hopeful that there is a big advance coming in the use of AI and in qualitative research.
We are at the forefront of that, and we have been able to solve these problems at a smaller level already.
This is a really exciting time to be a researcher — especially one who’s excited about the potential of AI.
There’s much more to come.
Listen to the Rosenfeld Review at AI’s Role in Qualitative Research for the whole story.
* All commentary reflects Prayag’s views on AI in research in February 2023. The transcript was created by Marvin with a human writer jumping in to add context and copy edits as needed.