Artificial intelligence has unquestionably left its mark on the world of academic research, with its ability to analyse vast datasets and perform calculations at scale. But when it comes to qualitative data gathered through interviews, observation, archival searches, or other methods used throughout social science, education, and market research, AI’s usefulness is yet to truly be determined.
As the debate continues over the opportunities and risks posed by artificial intelligence in research, a new paper by UCL School of Management researchers Daniel O’Sullivan and Davide Ravasi suggests that AI’s ability to help interpret and theorise qualitative data is potentially greater than we might expect.
Forthcoming in Strategic Organization, the paper sets out how GenAI can help qualitative researchers explain recurring patterns within their data in terms of more general forces and mechanisms that underlie these observations – and offers a protocol that researchers can use to do so. This approach is described as a form of ‘disciplined imagination’, where possible explanations for an observation are advanced, evaluated against a range of criteria, and gradually refined over time.
Traditionally, these steps have depended heavily on human judgement, intuition, insight, and creativity. But the researchers argue that generative AI can augment this process in four important ways:
Firstly, it can expand imagination by rapidly generating a larger number of possible explanations than a researcher might develop independently, drawing from multiple domains within the social sciences. Because AI systems do not become attached to a particular interpretation, they can also produce multiple and even contradictory perspectives on the same observation, helping researchers avoid becoming anchored to their initial assumptions.
Secondly, AI can enrich imagination by developing emerging ideas into more detailed theoretical accounts. A tentative insight that might otherwise be discarded can be expanded into a more coherent explanation, often drawing on related theories and concepts from across the academic literature.
Thirdly, AI can broaden discipline by helping researchers evaluate explanations against multiple criteria simultaneously. Rather than considering factors such as plausibility, coherence, methodological fit, and topical relevance one at a time, researchers can use AI to assess how competing explanations perform across a range of dimensions.
Finally, it can scaffold discipline, helping researchers apply criteria more consistently by, for instance, offering arguments both for and against particular interpretations. This can help researchers identify weaknesses in their reasoning and refine the evaluative standards they use.
At the same time, the authors of the paper warn about well-known limitations and encourage researchers to resist treating AI-generated outputs as authoritative. They underline how the approach they propose “should not be interpreted as a prescription for the automation of theorization, but rather as a framework for thinking more carefully about how emerging technologies can support one of the most elusive, yet most consequential, activities in qualitative research.”
“GenAI, as we understand it,” – they note – “certainly cannot, and should not, replace human contribution to the theorization of qualitative data.” Rather than rejecting this technology entirely, however, they invite other researchers to redirect their efforts from debating whether or not generative AI should be used for qualitative research to developing guidelines and practices that leverage its capacity to structure and augment interpretive processes that are otherwise constrained by our cognitive capacity.
Prompt-making and sensemaking: Using generative artificial intelligence to interpret empirical patterns in qualitative research was published online in Strategic Organization on July 31, 2026
