Maybe it’s too broad to discuss AI replacing ALL qualitative research. I think AI could play a role in inductive/iterative approaches based on grounded theory/thematic analysis. This is what I typically see in HCI. Ethnography feels like a much more creative process. danah boyd often makes this hand motion of massaging data when she discusses her work. It’s a long-term, immersive process that requires a HUMAN to make meaning out of the messiness. I have a hard time believing an AI ethnographer could have the same touch.
I'm sympathetic to your distinction, but I'd argue the point I'm trying to make is a bit narrower: even if AI cannot do ethnography in any deep sense, it may still become good enough at answering types of Qs associated with qualitative inquiry. My argument is that, because AI's answers to these questions are not really qualitative research, organizations will begin using it instead (because it meets criteria they care about).
Hm ok, I see your point. In that case, I think your hot take applies more to qualitative work where the main goal is to look across a lot of data and extract common patterns (e.g., thematic analysis). I can see how that aligns with the positivist critique that qualitative work is prone to confirmation bias. AI could do that at a larger scale in favor of corporate/institutional narratives.
I feel like ethnography and adjacent qualitative methods (where you’re spending a long time with participants and trying to make sense of context-specific situations) has already been dwindling in favor of methods that are faster and easier to scale. The outputs are more nuanced and harder to generalize (though I would argue more interesting and fun to read). So maybe AI is just going to accelerate that shift 🤷🏻♀️
My thoughts are coming from recent conversations with anthropologists/STS scholars, as opposed to qualitative HCI
I guess I'm confused. What, exactly, is it about qualitative research is it that makes it undoable by AI?
I'm not saying I think AI *can* do it, mind, but I'm not sure what your argument is that it cannot. Maybe this is painfully obvious to those trained in qualitative research, but a clear articulation would be nice for those of us who are not!
My understanding of their argument is that qualitative research cannot simply be automated because the researcher is part of the method--- the essential thing being contributed is human interpretation. But frankly, this is just a definition. But I get where you are coming from... this tradition does not have a monopoly over a set of questions, and we will definitely get something that looks similar to qual research using AI (and that may be useful in similar scenarios), even if qual researchers themselves refuse to call it that
Ah, I read a bit sloppily! I was taking this to be your argument, rather than their argument.
I feel like the more persuasive argument would center on the matter of data collection, rather than interpretation? AI will only take you so far because people do not live in data centers (at least for now). But this has nothing to do with the qualitative/quantitative divide per se.
I feel this distinction is clearer from a quant perspective than from a qual perspective. Playing devil's advocate, I'd say that the collection _is_ the interpretation for much of qualitative research.
Maybe it’s too broad to discuss AI replacing ALL qualitative research. I think AI could play a role in inductive/iterative approaches based on grounded theory/thematic analysis. This is what I typically see in HCI. Ethnography feels like a much more creative process. danah boyd often makes this hand motion of massaging data when she discusses her work. It’s a long-term, immersive process that requires a HUMAN to make meaning out of the messiness. I have a hard time believing an AI ethnographer could have the same touch.
I'm sympathetic to your distinction, but I'd argue the point I'm trying to make is a bit narrower: even if AI cannot do ethnography in any deep sense, it may still become good enough at answering types of Qs associated with qualitative inquiry. My argument is that, because AI's answers to these questions are not really qualitative research, organizations will begin using it instead (because it meets criteria they care about).
There should be some kind of bound that if n = 81000, it’s not qualitative research
I absolutely agree, and that’s why I think that writing a letter saying qual research can’t be automated is the wrong point!
Hm ok, I see your point. In that case, I think your hot take applies more to qualitative work where the main goal is to look across a lot of data and extract common patterns (e.g., thematic analysis). I can see how that aligns with the positivist critique that qualitative work is prone to confirmation bias. AI could do that at a larger scale in favor of corporate/institutional narratives.
I feel like ethnography and adjacent qualitative methods (where you’re spending a long time with participants and trying to make sense of context-specific situations) has already been dwindling in favor of methods that are faster and easier to scale. The outputs are more nuanced and harder to generalize (though I would argue more interesting and fun to read). So maybe AI is just going to accelerate that shift 🤷🏻♀️
My thoughts are coming from recent conversations with anthropologists/STS scholars, as opposed to qualitative HCI
I guess I'm confused. What, exactly, is it about qualitative research is it that makes it undoable by AI?
I'm not saying I think AI *can* do it, mind, but I'm not sure what your argument is that it cannot. Maybe this is painfully obvious to those trained in qualitative research, but a clear articulation would be nice for those of us who are not!
My understanding of their argument is that qualitative research cannot simply be automated because the researcher is part of the method--- the essential thing being contributed is human interpretation. But frankly, this is just a definition. But I get where you are coming from... this tradition does not have a monopoly over a set of questions, and we will definitely get something that looks similar to qual research using AI (and that may be useful in similar scenarios), even if qual researchers themselves refuse to call it that
Ah, I read a bit sloppily! I was taking this to be your argument, rather than their argument.
I feel like the more persuasive argument would center on the matter of data collection, rather than interpretation? AI will only take you so far because people do not live in data centers (at least for now). But this has nothing to do with the qualitative/quantitative divide per se.
I feel this distinction is clearer from a quant perspective than from a qual perspective. Playing devil's advocate, I'd say that the collection _is_ the interpretation for much of qualitative research.
It makes sense that there would not be as clear a distinction between collection and interpretation in qual research.
It seems to me that this theme would be a much more productive line of argument than what the writers of that letter actually articulate.