Human-in-the-Loop (HITL) as an Epistemic Imperative in Algorithmic Knowledge Organization

Noruzi, Alireza Human-in-the-Loop (HITL) as an Epistemic Imperative in Algorithmic Knowledge Organization. Librarianship and Information Organization Studies, 2026, vol. 37, n. 2, pp. 7-18. [Journal article (Paginated)]

Preview	Text HITL.pdf - Published version Available under License Creative Commons Attribution. Download (999kB) | Preview
Preview	Text (English Translation) HITL_En.pdf - Other Available under License Creative Commons Attribution. Download (333kB) | Preview

English abstract

The exponential growth of information worldwide has rendered entirely manual cataloging, classification, and metadata creation economically and logistically unsustainable. Conversely, the rush toward fully autonomous, AI-driven knowledge organization systems has proven equally problematic, as these systems are beset by algorithmic hallucinations, contextual blindness, and the implicit encoding of historical biases. Consequently, the discipline of library and information science finds itself trapped in a false dichotomy between the unscalable bottleneck of human labor and the unreliable opacity of machine autonomy. This editorial argues that Human-in-the-Loop (HITL) in knowledge organization is not a concession to technological limitations, but rather a strategic cognitive partnership that leverages both machine capabilities and human nuance. This paradigm transcends a mere quality-control mechanism, emerging instead as a foundational philosophy for knowledge organization in the algorithmic age. Drawing upon empirical evidence from evaluation studies conducted at the Library of Congress, the German National Library of Economics (ZBW), and the Shanghai Library, this study demonstrates that while automated systems perform poorly in subject heading assignment and classification number allocation, with accuracy rates below 50%, HITL workflows can achieve near-perfect accuracy (approaching 100%). The application of this paradigm across five core domains—namely cataloging, classification, indexing, metadata, and ontology—is briefly addressed. The editorial concludes by analyzing the key challenges of implementing HITL in knowledge organization, including intervention timing, training data quality, and organizational adaptation, and argues that the future of the field lies not in human versus machine, but in human-directed AI systems—an approach that guides the enduring profession toward a renewed and empowered purpose in the age of artificial intelligence.

Item type:	Journal article (Paginated)
Keywords:	Human-in-the-Loop (HITL); knowledge organization; artificial intelligence; epistemic imperative; metadata; cataloging; algorithmic bias; classification
Subjects:	H. Information sources, supports, channels. > HM. OPACs. I. Information treatment for information services > IA. Cataloging, bibliographic control.
Depositing user:	Dr. Alireza Noruzi
Date deposited:	26 Aug 2026 11:46
Last modified:	26 Aug 2026 11:47
URI:	http://hdl.handle.net/10760/48480

References

Dayal, U. (2026). Why Human-in-the-Loop Is Critical for High-Quality Metadata? Digital Divide Data. https://www.digitaldividedata.com/blog/human-in-the-loop-metadata

Kasprzik, A. (2025). Transferring applied machine learning research into subject indexing practice. In: Balnaves, Edmund et al. (Eds.), New Horizons in Artificial Intelligence in Libraries. Berlin: De Gruyter Saur, 2025, pp. 199-212, https://doi.org/10.1515/9783111336435-015

Library of Congress. (n.d.). Exploring Computational Description: Investigating how machine learning can help with cataloging. The Library of Congress Labs. https://labs.loc.gov/work/experiments/ECD/

Olson, M. (2026). Operating the franchise: vendor consolidation, algorithmic mediation, and the remaking of academic librarians as platform administrators for digital capitalism. Information Research, 31(2), 348-367. https://doi.org/10.47989/ir31262945

Potter, A., & Saccucci, C. (2024). Could artificial intelligence help catalog thousands of digital library books? An interview with Abigail Potter and Caroline Saccucci. The Signal: Digital Happenings at the Library of Congress. https://blogs.loc.gov/thesignal/2024/11/could-artificial-intelligence-help-catalog-thousands-of-digital-library-books-an-interview-with-abigail-potter-and-caroline-saccucci/

Shi, Z., Zhang, Y., & Wang, P. (2026). Research on a new intelligent cataloging model for special collections based on multimodal large language models. Journal of Information and Management, 11(1), 52-63. https://jim.library.sh.cn/EN/Y2026/V11/I1/52

Downloads

Actions (login required)

View Item