What counts as a funder? Developing and applying an empirical typology of organisational acknowledgements

Bibliometric studies of funding acknowledgements have largely counted funders or sorted them into familiar types, taking for granted what a funder is. Yet acknowledgement and funding sections credit a far wider range of organisations and roles: facilities, data providers, employers, editorial services, and individuals. The boundary between a genuine research funder and these other roles has had little dedicated attention, even though how we draw it determines what any funding measure actually counts. Our presentation follows this question through three stages. First, developing a 26-category empirical typology of organisational roles in acknowledgement and funding text - built bottom-up from the Dimensions corpus, with human-coded category boundaries validated against hand-coded examples and inter-rater agreement; arriving at the categories is itself a core contribution. Second, applying it: in a pilot on two contrasting Danish institutions - Aalborg University (comprehensive) and Copenhagen Business School (specialist) - we classified every organisation mention in 250 publications from each. The two invert one another: 60% of Aalborg mentions are direct funders, against 57% non-funder roles at CBS, dominated by acknowledged individuals. A much larger study, covering all Danish research organisations across several time periods, is now under way. Third, a reflection on the role of AI throughout: different systems served different stages - Claude to seed and refine the categories and, with Gemini, to cross-check inter-rater agreement; Gemini to classify at scale; and Copilot in the literature search. We discuss what this means for analyses of this kind, including pitfalls such as output truncation and the need for human-coded validation.