by Christina Silver, Director. CAQDAS Networking Project
I was invited to give the keynote at the MAXDAYS 2026 Europe virtual conference that took place on the 17th March. I was grateful for the opportunity to talk about the relationship between methodology and technology, which is something I’ve spent much of my career thinking about. This is a brief overview of my talk. You can also access the recording, if you’d rather watch/listen the whole take and hear the Q+A discussion.
The title of this keynote was “Are we losing our (qualitative) minds? Possibilities for critical thinking, reflexivity, and interpretation in GenAI-assisted thematic analysis”.
In the title ‘qualitative’ is in parenthesis to highlight whether, by inviting Generative AI into our analytic practices, we might “lose our minds” both in general cognitive terms, and in relation to the integrity of our qualitative methodologies. In other words, does relying on GenAI make us less critical, less thoughtful, more passive or lazy? And does using GenAI for qualitative analysis undermines what it means to work qualitatively?
My short answer is we might, but we needn’t.
Generative AI can’t do critical thinking, interpretation, or reflexivity in the same way we can, so these remain human tasks and responsibilities. Yet its use doesn’t automatically diminish critical thinking, reflexivity or the capacity to interpret – because it depends how we use it.
Before discussing the methodological aspects of my argument, I raised the ever-important ethical backdrop – as I always do. Many researchers when considering the ethics of using GenAI for qualitative work, focus on research integrity issues such as privacy, bias, transparency, and due diligence. These are critical, but in addition are broader ethics: socio-political and environmental concerns, including who develops AI, how training data are sourced, implications for intellectual property, authorship and so on, as well as the socio-political and environmental consequences of LLM use. These issues impact researcher’s choice with respect to AI use, and are legitimate reasons not to use GenAI for qualitative work. Indeed, they influence my own choices, as I’ve written about elsewhere. Nevertheless, as an educator, engaging with GenAI is partly a pedagogical duty – my role as current Director of the CAQDAS Networking Project means I have a responsibility to understand how it works in order to support informed debate, but I never do that without discussing ethics, which frame all discussions and uses.
We know there are many different responses in qualitative spaces to the rise of GenAI and what it means societally and methodologically. Some adopt eagerly, others reject on principle, whilst many are uncertain what it means for their work, and are grappling with how to deal with it. It can be helpful to reflect on theories of technological determinism, instrumentalism, and reflexivity to consider what’s happening in the space. More determinist thinking positions technological development as inevitable – GenAI exists, therefore we must use it, and this helps understand the rush we observed in some quarters to use GenAI and develop new tools. Instrumentalism in contrast, emphasises human intention and control in how we use tools, and reflexivity that tools always have consequences and must therefore be considered in context. All three perspectives are visible in current debates, including in claims that GenAI capabilities render existing methods obsolete or that identifying patterns in text automatically constitutes “theme development”.
In many ways, current debates mirror those we saw in the early 1990s, when CAQDAS-packages first emerged. One suggestion then was that CAQDAS-packages homogenised method, which I didn’t subscribe to at the time. However, now, because of the nature of GenAI and the ways it is being used and promoted in some quarters, the risk that qualitative analytic methods are being flattened, homogenised, or sidelined is real. For example, many discussions focus on GenAI doing ‘thematic analysis’, yet I question that what GenAI can generate are “themes” in the ways methods conceive them. The fact that a “theme” means something different across methods is important here too. Analytic methods in qualitative spaces are diverse, conceptually grounded, and shaped by epistemological perspectives, and none of this is replaced simply because GenAI can identify patterns across large volumes of data.
This is why the Five‑Level QDA way of thinking is so important now. Emphasising those analytic strategies – comprising research objectives and analytic plans – drive software tactics (the use of tools) is the core principle of the method Nick Woolf and I developed. Although we agree that tools may inform method, tools are not the architects of method. I do not believe analytic methods developed over decades disappear “in a poof of magic dust” just because new capabilities exist. And so, in this keynote I urged researchers to remember the underlying principles of the type of qualitative research they are doing. Answering questions including: What is the point of analytic methods in your practice? and what role do tools play for you? helps identify what appropriate choices about using GenAI (or indeed not using it), looks like in different contexts.
To illustrate this, I provided an overview of a framework I’ve developed over the past few years in teaching about GenAI – which will be published soon in a forthcoming chapter (see references). In my own work, using this framework (the Why-When-How-What-Does framework) continually, to enact techno‑methodological reflexivity, enables me to make appropriate decisions about GenAI use, not on a global basis for all projects, but on a case-by-case basis, for each project with its idiosyncratic needs. Because there is no one-size-fits-all answer to whether and how to use GenAI for qualitative work.
I finished up making some comments about my thoughts on how the developers of MAXQDA have implemented GenAI capabilities into the program. The focus was on MAXQDA for this keynote because the audience was the MAXDAYS conference, organised by Verbi Software. You can access those specific comments by reading the full script, available here.
I concluded by returning to the starting question and the central message of my talk: we do not have to lose our (qualitative) minds if using GenAI, but it requires active responsibility – constant reflexivity, informed decision‑making, and being in control of analytic processes – to ensure we don’t. GenAI may be reshaping the landscape, but it need not diminish our intellectual or methodological agency. In the end, the quality of qualitative research remains firmly in the hands of researchers.
AI use: I use Generative AI tools to research my teaching of them. However, no Generative AI was used in preparing this talk, or in writing it up.
