This presentation surfaces methodological tensions within Automatic Text Recognition (ATR) practice, the AI-enabled process of converting images-of-text into machine-processable data (Mühlberger et al., 2019), across cultural heritage research and institutions. It focuses on two main user communities - research libraries and digital collection users, while arguing for embracing friction as productive engagement with AI-driven systems: distinguishing access-oriented from research-oriented transcriptions, institutional and user expectations, and the foregrounding of documentation in producing ATR outputs and (re)using transcription models. In doing so, this presentation addresses an urgent need to discuss how individual libraries are integrating AI tools into curatorial, technical and bibliographic protocols (Terras, 2022: 144). Through consulting National Library of Scotland (NLS) curators and digital staff as to their attitudes, priorities and problem-areas regarding ATR, our approach is grounded in the values and priorities of the sector (Gooding et al., 2025). Six interviews were first conducted with NLS staff in 2024 and re-evaluated with the same participants in 2025, as part of an internally funded project. These interviews followed an action research methodology, defined as ‘a practice changing practice’ (Kemmis et al., 2014: 2) to further understand how participants orientated themselves toward emerging AI tools for collection processing. This was followed by group discussions with the ATR user community, specifically the membership of academic cooperative Transkribus (https://www.transkribus.org/). Transkribus was chosen due to it being the largest consumer-level ATR system and having a clearly defined community of practice (Terras et al., 2025). From this, a finite list of ATR practices was constructed by categorising latent themes and core ideas (Drisko & Maschi, 2015: 405). This followed Unsworth’s (2000) ‘scholarly primitives’ model for digital humanities research and core ATR practices to be mapped as Trading Zones across organisational contexts (Kemman, 2021: 39-58). This presentation demonstrates, as findings, how tension arising from disparate AI practice can be better negotiated, in turn ensuring that sectoral values of findability, accessibility, interoperability and reusability are maintained. Subsequently, it moves beyond external theorisations of curation and research, instead mapping where divergence occurs between infrastructural assumptions of automation, scalability and universality, and libraries’ status as trustworthy sources of data specificity, interpretive labour and situated analysis. By mapping ATR practice, research libraries are provided a visual resource that informs complex decision-making around internal resourcing of digitisation tools, model training and parameter setting, against the constantly evolving and uncertain world of AI development. Mühlberger, G., et al. (2019). Transforming scholarship in the archives through handwriting text recognition, Transkribus as a case study. Journal of Documentation. 75(50): 965-967. doi: 10.1108/JD-07-2018-0114/full/html Drisko, J.W., Maschi, T. (2015). Content analysis. London: Oxford University Press. Gooding, P., et al. 2025. The adoption of handwritten text recognition at the National Library of Scotland. In Navigating AI for Cultural Heritage Organisations, edited by Lise Jaillant, Claire Warwick, Paul Gooding, Katherine Aske, Glen Layne-Worthey and J. Stephen Downie. 187 - 207. UCL Press. Terras, M. 2022. The role of the library when computers can read: Critically adopting handwritten text recognition (HTR) technologies to support research. In The Rise of AI: Implications and applications of artificial intelligence in academic libraries, edited by S. Hervieux and A. Wheatley, 137–48. American Library Association. Terras, M. et al. 2025. The artificial intelligence cooperative: READ-COOP, Transkribus, and the benefits of shared community infrastructure for automated text recognition. Open Res Europe, 5:16. doi: 10.12688/openreseurope.18747.1 Kemman, M. 2021. Trading Zones of Digital History. De Gruyter. Kemmis, S. et al. 2014. The Action Research Planner. Springer. Unsworth, S. 2020. Scholarly Primitives. Available at: https://people.brandeis.edu/~unsworth/Kings.5-00/primitives.html
Joseph Nockels (Tue,) studied this question.