A outcome of this project has been to develop an approach aimed at steering through the polarised positions emerging in the literature. Placing the insights from the empirical study in conversation with published LLM-based methods, the team developed a flexible yet structured set of guiding principles, together with a systematised process and suggested LLM prompts – referred to as an AMRAQ approach: AI-augmented middle-range analysis of qualitative data. This approach was then presented and further refined through four workshops with experienced qualitative researchers (n=18). 

AMRAQ draws on the principle of a human-in-the-loop process, which has demonstrated considerable promise as an augmentative approach that maintains researcher centrality while leveraging the computational capabilities of LLMs (Cook et al., 2025). Much of the analytical work happens 'offline', enabling the researcher to add nuance the LLM misses, actively guard against algorithmic biases, and maintain emotionally-sensitive engagement with the data. 

AMRAQ also uses a double meaning of middle-range. First, it occupies a middle ground in scale between small-sample, in-depth qualitative traditions and semi-automated, big-data methodologies – promising to enable qualitative studies with significantly larger sample sizes. Second, middle-range refers to the ambition of contributing to middle-range theories: explanatory statements filling the gap between grand narratives and minor claims about day-to-day processes (Cartwright, 2020; Emmel et al., 2018). Larger, heterogeneous sample sizes enhance the ability to iteratively test emerging theories through sub-group comparisons and counterfactual reasoning. 

An AMRAQ approach has the potential to make a transformative contribution to the methodological landscape but requires further testing both as a particular method and a test case for a range of similar approaches. 

An open-access version of our recommended AMRAQ approach will be provided here shortly, along with additional resources.