Preserving algorithmic systems: a synthesis of overlapping approaches, materialities and contexts

• Published in: Journal of documentation Vol. 79; no. 6; pp. 1380 - 1392
• Main Authors: Hodges, James A., Trace, Ciaran B.
• Format: Journal Article
• Language: English
• Published: Bradford Emerald Publishing Limited 24-10-2023; Emerald Group Publishing Limited
• Subjects: Accountability; Algorithms; Archives & records; Archivists; Authors; Computers; Data collection; Definitions; Digital preservation; Literature reviews; Mathematics; Preservation; Professions; Recordkeeping; Social Problems; Social Systems; Synthesis; System effectiveness; Transparency; Socio-technical systems; Algorithms; Digital preservation; Algorithmic culture
• ISSN: 0022-0418; 1758-7379
• DOI: 10.1108/JD-09-2022-0204

PurposeThis article aims to advance a multifaceted framework for preserving algorithms and algorithmic systems in an archival context.Design/methodology/approachThe article is based on a review and synthesis of existing literature, during which the authors observe emergent themes. After introducing these themes, the authors follow each theme as manifest in existing digital preservation projects, starting with algorithms' earliest conceptual starting points and moving up through themes' eventual implementation within a complex social environment.FindingsThe authors find current literature is largely divided between that which addresses algorithms primarily as computational artifacts and that which views them instead as primarily social in nature. To bridge this gap the authors propose that “the algorithm,” as the algorithm is frequently deployed in popular discourse, is best understood as not as either the algorithm's technical or social components, but rather the sum total of both.Research limitations/implicationsThe study is limited by its methodology as a literature review. However, the findings point toward a new framing for future research that is less divided in terms of social or material orientation.Practical implicationsCreating multifaceted records of algorithms, the authors argue, enables more effective regulation and management of algorithmic systems, which in turn help to improve their levels of fairness, accountability, and trustworthiness.Originality/valueThe paper offers a wide variety of case studies with the potential to inform future studies, while contextualizing the studies together within a new framework that avoids prior limitations.