Qualitative research and AI. Two terms that make a lot of academics shudder when you put them together in the same sentence. Like, eww.
I think the conversations around AI and research, especially qualitative inquiry, have reached the kind of volume that risks distorting the sound. You know when someone turns the noise up so much that it just sends your ears all weird and you’re like, “Make it stop!”
That’s the point I’ve got to lately because so much of the conversation around AI in research goes around in circles. “You shouldn’t use AI for THAT, but it’s okay for THAT”. Then someone else comes along and switches the ‘thats’ around, so on and so forth.
We’re worried about what AI can do, what it might replace, the ever-growing nightmare list of what researchers should be worried about and how academic practice as a whole might change. Eek. Yeah, it’s a lot. No wonder we’re all a bit squeaky-bum-cheeks about it.
This is not our first new-tech rodeo
Qualitative researchers have adapted to a lot of technological change over the past several decades. In terms of being a bit (or a lot) rattled by a new techy thing, this is not our first rodeo, far from it.
Tape recorders and video for starters. Side note - I had a little diddy tape recorder when I first started out, it was so cute, with the mini tape in it, aww. Anyway, when researchers were able to record their interactions with participants, that changed how interactions could be documented and revisited. Then came the digital versions of that tech, which brought with them a whole host of new considerations.
Enter computer-assisted qualitative data analysis software - NVivo and Friends. That changed how researchers were able to organise and work with their data. Another side note - I’ve never really got on all that well with this stuff TBH, I love a highlighter and a sticky note.
Then the biggies - the internet, and then social media. Oof, they created a whole new range of things to think about. They were new research environments in and of themselves, as well as communication channels, which came with new ways of engaging with people and raised lots more awkward ethical questions.
Anyone who did research in the pandemic years will know all about the Zoom interview, love it or hate it (!). Here’s what my face used to do when someone said “Zoom interview” back in 2020: 😐. I’ve come around to it since, though. This kind of tech had an impact not just on participation, but on access and researcher-participant relationships.
“But, Elizabeth!” I hear you saying through the screen, “AI is different!” Yes, I’ll give you that. There are pretty significant ways in which this differs from what’s come before. It’s generative for starters, so it goes further than simply being some fancy thing that can store and organise information. Yes, this is different.
Thing is though, those previous technologies weren’t completely neutral, were they? They shaped the way we did research, and they made possible things that hadn’t even been conceivable before - for example, being able to research with people across the globe and offer new ways of looking at data.
I’m not saying here, “Calm down dear, we’ve seen this all before”. We haven’t seen this before, but qualitative researchers have always had to think critically about the relationship between methods and technologies, haven’t we? We’ve always had that responsibility to our participants and to our disciplines. In my view, AI brings that responsibility into sharper focus.
Perhaps one of the most important things we need to do is go right back to the fundamentals. Three questions to tackle here:
What is qualitative research?
What is qualitative research actually for?
What about that purpose remains important, regardless of the technology that’s available to us?
The definition and purpose of qualitative research
Let’s consult the OGs, Denzin and Lincoln. Here’s their 2018 definition:
Qualitative research consists of a set of interpretive, material practices that make the world visible. These practices transform the world. They turn the world into a series of representations… qualitative researchers study things in their natural settings, attempting to make sense of or interpret phenomena in terms of the meaning people bring to them. (Denzin and Lincoln, 2018: 10, emphasis added).
My understanding of this is that qualitative research is like taking a yellow highlighter to the world and saying, “Hey! Look at this thing here that you might not have noticed otherwise. Kind of interesting, right?!” People then gather round and say, “Oh, yes. That is interesting, let’s explore that further - where did that come from?”.
We try to achieve that through researching with people. We talk with them, we listen to them, we spend time with them as they navigate and reflect on their lives, we see them. As such, qualitative research creates a space to explore the sense-making that gives texture and rhythm to everyday life and consider how those things come into existence and change as time goes on.
In so doing, we are mindful of several things.
Firstly, the social and cultural scenery. We appreciate that people’s experiences are the culmination of histories, structures, cultures, values and beliefs. They don’t form in a vacuum; they’re not random, isolated events.
Secondly, qualitative research is not setting out in search of the Holy Grail. We’re not knights out on a quest to find the answer, and then we can all just gallop merrily home on our horses for tea and cake. We’re all about poking around in the ambiguity, contradiction and difference and figuring out why it all looks so messy. That’s fun, that’s geek-out time for us. And more importantly, it’s of unique value in terms of what it can highlight, challenge and change.
Qualitative research isn't one single thing, of course. There are different ideas about what constitutes a good explanation. We're not all trying to do exactly the same thing, and we don't necessarily agree about how knowledge is produced. What we do share is a commitment to taking human experience and social life seriously, including the context in which they unfold.
Third, pardon the pun - completely intended - we’re not all robots sucking up facts about people’s lives then regurgitating them out onto the page. We understand - and don’t shy away from - the fact that our own assumptions and experiences shape how we make sense of those of others. Yes, reflexivity - it’s something we embrace rather than shuffling about awkwardly and kicking it back under the sofa when we think no one’s looking.
Fourthly, qualitative research itself is not a homogenous “thing”. Our different traditions have different emphases, zooming in on particular parts of social life: lived experience; social processes; culture; power; discourse.
So, how do we take this into the age of AI, and what endures?
What AI DOESN’T make irrelevant
When we look at qualitative research in this way, a few things become clear, which emphasise just how important the human eye and mind are in qualitative inquiry.
Context, interpretation and judgement
Firstly - context. We’ll often hear people picking over the bones of a debate online, where someone says, “Well, that was taken completely out of context"!”. The not-very-nice practice of ‘clipping’ - whereby a short extract is taken from a longer conversation and circulated without the surrounding material that gives it meaning - has really brought this to the fore in recent years.
Context matters. It really does, because something somebody says or how someone behaves does not mean the same thing in every setting. A quote from an interview isn’t self-explanatory. Yes, it might be an accurate transcription of someone’s words, but that doesn’t explain what led them to speak them out loud.
Being appreciative of context involves carefully considering the circumstances in which people said what they said or behaved as they did, examining the beliefs, values, relationships and assumptions that set the scene.
Now, AI can suggest potential patterns and how things might be connected. But identifying a pattern is not the same thing as being able to make sense of its significance within a specific context.
Now, I'm not suggesting that AI can't produce interpretations. It absolutely can and some of those suggestions might be genuinely useful. They might even prompt us to consider something we hadn't thought about before and plenty of people have shared with me that AI has helped them with this very thing.
Imagine a study with PhD researchers about supervision, in which several of them say they feel supported by their supervisors. What does ‘supported’ really mean? AI could probably suggest a few things: Is support about receiving detailed feedback? Is it to do with being given the freedom to work independently? Is it about having access to someone during difficult periods? Feeling intellectually challenged? Having concerns taken seriously?
BUT - producing a plausible interpretation (or several) isn't the same as establishing that it's a good one. Does it make sense in the context of this particular study? Is it consistent with your methodological approach and underlying assumptions? Is there sufficient evidence to support it? What other interpretations might be possible?
These are questions that require academic judgement. And even if AI has helped us develop an interpretation, we're still responsible for deciding whether we can stand behind it. AI might contribute to the thinking, but it doesn't take on that responsibility for us.
This goes beyond merely extracting information from the data that’s been collected; we’re not digging holes on the beach looking for buried treasure whilst an AI robot stands next to us doing the same, both of us furiously digging and the fastest one to get to “the” meaning is the winner. We’re building meaning.
Human researchers are needed here - an output from an AI does not settle the questions of what we should pay attention to, how to understand that, what competing interpretations might be available and what the evidence can support.
This involves human interpretation, and it also involves judgement. Judgement about what we do with our interpretations, both as we’re developing them, and afterwards - because that too is a key part of qualitative research.
Qualitative research is about practices that change the world, and that involves making judgement calls. What analytical avenues do we pursue now? What does the data support? Who might disagree with our take on it? What are the limitations of our claims, how far can we really take them, how much mileage have they got? Judgement is still needed - and humans are needed for that - whatever AI has to contribute, we then make the call about what to do with it.
Ethics and accountability
Ethical responsibility is a key strand in the debates around AI and qualitative research.
In terms of confidentiality, what happens when sensitive research material is shared with an AI system?
What exactly are participants consenting to in terms of what happens to their contribution?
Where is information stored now, and who gets access to it?
Whose perspectives shape what an AI system might privilege?
What needs to be disclosed about how AI contributed to the research process?
Obviously, right? We will be asking these questions as researchers who want to make a positive difference to people, mindful of the fact that doing that involves diving into things that are sensitive. AI has brought these issues into sharper focus but it hasn’t removed our responsibility for them. The buck still stops with us, the humans.
We are responsible and accountable for the research we do. We still need to explain how we arrived at our conclusions and why we consider our claims to be reasonable. We still need to think carefully about the consequences of our choices for the people whose experiences form the basis of our research and how we represent them in our findings. We’ve always done this, that responsibility existed long before AI - but now we’re having to look at how we might need to adapt in the age of AI.
When we strip off all the fat, when we cut back all the excess, when we declutter - what is the essence of our work? Context, interpretation, judgement, ethics and accountability. So, where do we go with that now?
How might qualitative research genuinely change?
In my view, what will need to change about qualitative research in the age of AI are the things we do in service of our purposes and responsibilities. It ain’t what you do, it’s the way that you do it. Thank you, Bananarama. 🎶
In all seriousness, though, I think two things are an important starting point.
(1) Get to grips with the new tech
Yes, we probably need to get down with the kids here, grasp the nettle, and take a closer look at the tech that’s emerging all around us. We don’t need to become experts with every AI tool that the influencers are pushing on us (goodness, NO!😉) but we need to know enough about it to figure out how suitable (or not) it is for our research and the implications of using it.
That’s going to help us avoid throwing any potentially helpful things in the bin under an overly simplistic, “Ban it all, it’s ALL unethical” stance that a lot of shouty people on social media have been filling my DMs in recent days with. It’s also going to help us carefully think through where we do want to draw the boundaries around it, because we are going to have to do that at some stage. Even if those boundaries might shift in time as society and technology develops, we do need a starting point.
Talking to each other about this is going to be important, because thoughtful conversations with other researchers give us the opportunity to work through the messy questions together, rather than just shouting our opinions at each other across the internet. Just sayin’. 😉
(2) Get back to writing
Whilst we’re getting clued up on the tech, I think we need to put writing back at the centre of intellectual development as qualitative researchers. Writing is a way of developing knowledge, it’s thinking on the page.
We often have to write before we decide what we really think about something. It’s only through putting those words in front of us that we can look critically at them and figure out if that’s what we really believe.
This is about the intellectual work that happens as we try to articulate our ideas. When we're struggling to explain something, when we write a sentence and realise it doesn't quite capture what we mean, when we have to go back and rethink an argument because something isn't adding up.
Struggling to write a sentence or a paragraph tells us something, right? We’re stuck for a reason. We try to figure it out. We think about it some more. It helps. It also gives us faith in our own intellectual abilities. It boosts our confidence because it proves we can figure a lot of stuff out on our own if we just give ourselves the time and space to do that. As tough as that might be in a world where we can get an instant answer.
To that point, when we start to use AI as a ‘thinking tool’ to bounce ideas around and help us get clear on what we think, are we beginning to erase writing, one of the important parts of the thinking process? What happens to our thinking when we do less of the writing? Are we handing over opportunities to develop our own understanding? That's something we need to pay very careful attention to.
On the other hand, could AI sometimes support the writing process without taking over the thinking - for example, by helping a researcher interrogate an argument they've already developed? I don’t know yet, I haven’t made my mind up on this one. I’m still a bit sceptical about it. Mmm.
Conclusions
So, summing up - jeez this ended up being way longer than I thought it would be, there goes my Saturday morning, FML 😂 - technological change doesn’t render qualitative research obsolete and it doesn’t make its fundamental intellectual purposes irrelevant.
What it has done is highlight that qualitative research remains - as it always has been - an evolving practice that’s adapted to a range of social and technological changes.
In terms of what adaptation to AI looks like, we don’t have to unquestioningly grab every tool available, nor do we have to close the door in AI’s face, politely saying, “No, thank you. Goodbye”.
The question qualitative research needs to ask now is perhaps this: How can we adapt to AI while remaining clear about what qualitative research is trying to accomplish?
To be continued.
Reference
Denzin, N.K. and Lincoln, Y.S. (2018) 'Introduction: The discipline and practice of qualitative research', in Denzin, N.K. and Lincoln, Y.S. (eds.) The SAGE handbook of qualitative research. 5th edn. Thousand Oaks, CA: SAGE Publications, pp. 1–26