Data cultures: Contested meanings in a public cultural institution

Abstract

This paper maps the configurations of meanings surrounding data culture and examines its ongoing transformation. Using the National Library and Archives of Quebec (BAnQ) as a case study, it explores the interplay between practices and interpretations of what constitutes data culture within this public cultural institution. Rather than approaching data culture as an entirely new set of dispositions that organizations need to develop, I propose understanding it as a contested field of meanings. This field brings together heterogeneous elements—some grounded in long-established professional practices, others emerging in response to new digital demands—where divergent logics of action and values collide. Rooted in critical data studies, this paper offers empirical insights into the power dynamics within data cultures, conceptualized as complex arrangements of meanings, material apparatuses, and social practices. It identifies key factors that shape data culture at the BAnQ: organizational structures, professional values, institutional goals (such as artifact documentation, public accessibility, and performance optimization), governmental requirements, and data tools and ideologies. The formation and transformation of data culture at the BAnQ appear to be a dynamic process that requires aligning new practices with existing frameworks of meaning, while also exposing tensions and resistance among differing interpretations. From this perspective, data culture emerges as an arena of debate, leading to genuine disagreements over the data practices required to fulfill the BAnQ's broader mission. As the institution navigates the challenges of datafication, its approach to data governance becomes pivotal in balancing public service goals with the imperatives of data innovation.

In 2021, a member of the data-governance committee at the National Library and Archives of Quebec (BAnQ) ¹ remarked that there was “no data culture” within the institution. This statement did not suggest that the BAnQ lacked data; in fact, the institution manages vast amounts, ranging from metadata identifying archival collections and library holdings, to transactional, usage, and heritage data. Instead, the committee member was emphasizing the need for a more systematic and integrated approach to leveraging this data.

It may be tempting to attribute this challenge of developing a data culture solely to deficits: insufficient adherence to strategic vision, a lack of data literacy among employees, or inadequate interoperability between data infrastructures. While these factors certainly contribute to the challenge, I argue that an equally significant issue lies in a surplus—specifically, the multitude of practices that fall under the broad umbrella of “data culture.” This surplus complicates efforts to establish a coherent and unifying meaning for data culture within the institution, as it combines diverse—and sometimes contradictory—rationales, further blurring its contours.

The notion of data culture broadly refers to a set of dispositions—collective frames of meanings and repertoires of practices—that shape the production, use, and effects of data. Originally emerging in the 1960s with the rise of the computer industry, the term “data culture” was revived in the 2010s, coinciding with the accelerating datafication of societies. This paper aims to map the configurations of meanings surrounding data culture within a public cultural institution and to examine how they are currently transforming. What are the factors that shape the meanings of data culture at the BAnQ? And how is this configuration of meanings affected by ongoing changes in the apparatuses for collecting and valuing digital data? The analysis will explore the diverse resources, practices and interpretations of what constitutes data culture at the BAnQ. These interpretations are framed by operational goals such as artifact documentation, public accessibility, and performance optimization. They change in response to key dynamics currently affecting digital environments—such as datafication, discoverability, and platformization—leading to genuine disagreements about the data practices needed to fulfill the BAnQ's broader mission of public service.

The analysis is based on a case study of the National Library and Archives of Quebec from 2020 to 2023. It draws on documentary and discursive materials from meetings and policy reviews to investigate data culture, governance, and management practices within the institution. Grounded in critical data studies (Iliadis and Russo, 2016; Kitchin, 2014), the study provides empirical insights into the change and power dynamics within data cultures, conceptualized as complex arrangements of meanings, material apparatuses, and social practices. The article begins by examining the broader changes currently shaping data culture within the library and archives sector. It argues that data culture should not be viewed as a fixed, monolithic entity but rather as plural, dynamic, and constantly changing. In the empirical sections, the paper analyses the various meanings attached to data culture and identifies key tensions that arise from the integration of new data practices. As BAnQ navigates the challenges of datafication and digital innovation, its approach to data governance becomes central to balancing public service goals with the imperatives of technological advancement.

Changing data environments in libraries and archives

The National Library and Archives of Quebec is a public corporation overseen by the Quebec provincial government. Its overall mission is to preserve Quebec's cultural heritage while providing public access to its collections. The BAnQ was established in 2006 through the merger of three distinct entities. The National Archives, founded in 1920, are tasked with preserving Quebec's historical records. The National Library, created in 1968, focuses on supporting Quebec publishing and the conservation and exhibition of heritage collections. The Grand Library, opened in 2005 in Montreal, serves as a public library that provides on-site and online access to a wide range of documents for consultation and borrowing. These distinct fields bring together multiple action logics that shape data practices within the institution.

By data practices, we follow Ruppert and Scheel (2021) in referring to intertwined human and material agency, situated in specific knowledge regimes and “performed by actors […] in competitive struggles over authority, influence, and resources within specific fields of practice” (2021: 35). At BAnQ, these practices are more specifically enacted through activities such as archival data management, library cataloging, and digitization workflows, which involve classifying knowledge and encoding cultural heritage.

Since its establishment, the BAnQ has seen significant shifts in data technologies and the associated data practices—comprising tools, systems, and methodologies crafted for collecting, storing, processing, analyzing, and visualizing data. While these shifts reflect broader institutional changes, it has been significantly reinforced by the dynamic of datafication, the process by which ever-wider aspects of personal and social life (mediatized social interactions, behaviors, and transactions such as borrowing books) are transformed into digital data that automated systems can exploit. Examples of datafication in the library and archives sector include the recent shift toward using heritage data for training AI models (Neudecker, 2022), and the rise of AI-driven tools for metadata creation, document analysis, and content recommendation (Chen and Li, 2024). It led to the emergence of new professions and skills combining statistics, computational methods, information systems management, and business intelligence (Stark and Hoffmann, 2019).

Introduced in 2021, the BAnQ's new digital strategy framework embodies these transformations. A central focus of the strategy is the assertion that “innovation is accelerated by data exploitation and artificial intelligence” (BAnQ, 2021: 14). To this end, the framework seeks to advance the development of new data practices, aiming to “improve user experience” by leveraging data to better understand and respond to users’ “needs” and “interests.” The strategy also sets a goal to enhance organizational “performance” through the implementation of “business intelligence.” In developing these new data-driven practices, ethical standards remain a central concern, as demonstrated by the institution's stated intention to set benchmarks for “exemplary practices” (BAnQ, 2021: 14). Harnessing new methods for data production and exploitation at the BAnQ is thus driven by internal momentum, but not only. It is also shaped by mandates from governing policies.

Developing data culture has become a requirement of cultural policies in Quebec, driven largely by the framework of digital discoverability. In this context, fostering a data culture is directly tied to making cultural content more accessible online, whether through commercial platforms or cultural organizations’ websites. Enhancing the discoverability of Francophone cultural content has become a strategic objective of Quebec's cultural policy, as exemplified by the 2019 Franco-Quebec Mission on the Digital Discoverability of Francophone Cultural Content. ² The importance of data culture is further highlighted by policy statements from Quebec's Ministry of Culture and Communications, which emphasize the need to “support the cultural sector in evolving towards a data culture” (MCCQ, 2021: 5). ³

Enhancing data culture is also closely linked to digital transition policies. As a public institution, the BAnQ is increasingly mandated to supply data to a centralized governmental data-analysis apparatus. This requirement is driven by policies rooted in a platform-based model of interconnected data flows. Since 2021 the institution has been required to comply with the new Act Respecting the Governance and Management of the Information Resources of Public Bodies and Government Enterprises. ⁴ This legislation designates certain data held by public institutions as “strategic information assets of the governmental digital heritage,” mandating their “mobility” and “valorization”—terms introduced to emphasize their significance for government-wide purposes (Conseil du Trésor du Québec, 2021: 11). This framework aligns with the view of “government as a platform,” as outlined in an OECD report calling for enhanced data flows across governmental bodies to foster a more integrated and connected governance model (OECD, 2019). At the BAnQ, this shift implies strengthening both data management competencies and infrastructure, fostering the development of a data culture to meet these changing requirements.

Datafication, discoverability, and platform-based models of data flows are three key logics that shape the context in which professionals at the BAnQ are encouraged to embrace a new data culture. In the libraries and archives sectors, as in the workplace at large, fostering data culture necessitates transforming professional practices to align with a new paradigm of innovation through data. Yet these injunctions are weighted with normative presumptions that are rarely questioned. They are based on the following premises: the tabula rasa—the idea that professionals either lack any preexisting data culture or that their current data culture is inadequate and poorly suited to guaranteeing the organization's success in today's digital environment; the urgency of acquiring a data culture to avoid obsolescence; and the efficiency of this new culture in optimizing decision-making, thereby ensuring organizational growth and success.

These assumptions warrant critical scrutiny. Enthusiasm for a “new” data culture in organizations is often driven by the belief that a model based on the exploitation of digital data is both technically and economically more efficient. However, a body of scholarship in critical data studies has questioned and highlighted the limitations of data extraction, accumulation, and valorization logics. Van Dijck (2014) called “dataism” the assumption that large sets of data can offer objective, neutral, and reliable insights into understanding and organizing society. Recent contributions in the field have linked the expansion of the data analysis industry to the rise of data capitalism (Beer, 2018), suggesting that it affects both the workplace and everyday life in ways that can undermine justice and transparency (Couldry and Mejias, 2019; Morozov, 2017; Sadowski, 2019).

Building on these insights, this paper critically examines the ongoing transformations shaping data culture within libraries and archives. Rather than viewing data culture as an entirely new set of dispositions that heritage institutions are lacking, I propose understanding it as a field of contested meanings. At the heart of cultural analysis lies the idea that meanings are not fixed: they are produced, circulated, and negotiated within situated social formations (Williams, 1983). As Hall (1997) further emphasizes, meanings coexist, compete and are mobilized in contextually specific ways—always open to contestation and reinterpretation. This article examines how such meanings are coproduced through apparatuses for collecting and valuing digital data. These apparatuses are understood, following Foucault (1980), as heterogeneous ensembles of discourses, practices, institutions, and technologies that establish particular regimes of visibility, legibility, and value.

This perspective leads to an understanding of data culture as a dynamic interplay of heterogeneous elements—some rooted in long-established professional practices of data production and management, others emerging in response to new digital demands—where different logics of action and values collide. The analysis seeks to explore how the plural meanings and practices of data culture at the BAnQ are shaped by both internal organizational factors—such as diverse professional fields, established practices, and institutional mandates—and broader transformations in the digital landscape.

The BAnQ serves as a compelling case study for examining how data-related practices can vary widely within a single setting. This is due to its hybrid nature as an organization; it encompasses different professional fields—those of archives and libraries—each having a distinct relationship with data. Furthermore, the documentary sector, which has historically been a leader in digitizing information systems, provides a unique context for investigating how established professional practices in cultural heritage preservation and public access to knowledge are adapting to new data culture environments. By analyzing the impact of these changes, the paper explores how current data paradigms intersect with traditional frames of reference related to data production and management. The complexities of shaping data culture within institutional contexts reveal how heritage institutions must navigate and reconcile changing digital demands with their core missions. This points to a deeper, more fundamental issue: the need for critical reflection on the contested nature of data culture itself.

Situating data culture

Data culture spans multiple interconnected dimensions of cultural and technical phenomena. Prior STS scholarship has emphasized that digital infrastructures and practices are embedded in—and constitutive of—specific, situated social contexts. In 1995, Star introduced the notion of “cultures of computing,” highlighting the diverse symbolic and communal meanings embedded in computing as a set of situated practices (Star, 1995). Similarly, Law, Ruppert, and Savage argue that the digital is not a uniform field but is deeply implicated in the making of what might be termed “local” social relations (Law et al., 2010: 5). These perspectives inform the analytical approach adopted in the paper, which foregrounds the situated enactment of data cultures within particular institutional settings.

The expression “data culture” gained prominence in the 2010s, in conjunction with what Kitchin (2015) described as a “data revolution”—marked by the widespread design and integration of digital infrastructures and devices across all areas of activity, the exploitation of massive datasets enabled by increased computational power (big data), and the rise of an open data movement originating in public administrations. Initially popularized in professional and policy contexts, the term “data culture” was soon theorized within the field of critical data studies.

Its plurality of meanings partly stems from the concept of culture itself—a multifaceted and complex notion in the humanities and social sciences. At its broadest, data culture encompasses the experiential frameworks of a world increasingly shaped by data's cultural and technical aspects. From a production perspective, it refers to activities surrounding the creation, management, and transformation of data within organizational and professional settings. From a user perspective, it captures the experiences and practices of data publics and audiences. This paper focuses on cultures of production (Albury et al., 2017), specifically examining data-production cultures within the professional realm of the arts and culture sector.

Recent scholarship grounded in critical data studies has examined how data culture is being reconfigured across diverse settings, including the professional domain, public spaces of interaction, and the personal sphere. Key contributions in this field have focused on the effects of “local coding culture” in the ordering of data across multiple scientific disciplines (Bowker, 2000: 664), the cultural industries (Kennedy, 2015), the sociopolitical implications of data culture in smart cities (Bates, 2017), and the daily experiences with data and automation from the perspective of everyday data cultures (Albury et al., 2017, Burgess et al., 2022). Particularly relevant to this article is how Bates (2017) relates the concepts of data culture, data practices, and data governance, emphasizing the local specificity of individual sites of practice while also highlighting their interconnection within broader networks of technical infrastructures, public policies, legislation, and economic systems. In the Francophone sphere, research on data cultures leans more toward a critical perspective on information culture, intersecting with the fields of documentation, education, and organizational communication (Labelle, 2017; Lehmans, 2017). This body of work highlights the transformation of organizational models and the development of new data-related professional skills, while emphasizing the importance of a shared vision among organizational members and a critical awareness of the meaning behind their actions.

This paper builds on this scholarship by emphasizing three key aspects of data cultures: they are diverse and situated in specific social, material, and cultural contexts that shape their development; they involve a commitment to action; and they foster meaning-making. Based on these elements, I adopt the following understanding of data culture as a guiding framework:

Data culture is shaped by an interrelated set of collective repertoires of practices and frames of meaning, grounded in specific contexts and characterized by a particular sensibility and rationality constructed around data sets. These practices and interpretive frameworks influence the shaping of data, their status, and their effects on the world. They are informed by values, norms (both explicit and implicit), literacies, affects, and technical infrastructures.

This framework is not intended to fix or stabilize a singular meaning of data culture that would be valid across all professional practices or social contexts. Rather, it serves as an analytical framework—a set of heuristic dimensions—that can guide the investigation of how data cultures are enacted, negotiated, and transformed in situated contexts.

The analysis in this paper will focus on the relationship between repertoires of data practices and frames of meaning associated with data culture in the context of professional practices. The notion of repertoires of data practices refers to the structured actions, recognized methods, and routines carried out by professionals in libraries and archives to produce, handle, preserve, and manage data. Some of these practices are historically established and rooted in professional norms (e.g., librarians, archivists, data scientists) and institutional structures, while others emerge in response to new tools and requirements driven by changes in digital environments (e.g., AI tools, data analytics, discoverability algorithms).

Frames of meaning are the interpretive frameworks through which individuals and groups make sense of their actions, assign value, and prioritize specific data practices. This approach draws from frame analysis, a conceptual and methodological lens used to study how meaning is constructed, organized, and communicated through interpretive frameworks—or “frames”—that shape the perception of social reality (Entman, 1993). We argue that the framing of data culture at BAnQ is shaped by multiple, sometimes competing, and sets of values—ranging from personal fulfillment in one's profession (e.g., social interaction with library users), to organizational goals (e.g., document preservation), and to broader normative principles (e.g., cultural democratization).

Repertoires of data practices and frames of meaning are closely intertwined in shaping data culture. As new data tools and institutional requirements emerge within the BAnQ's already diverse landscape of practices, changes occur in the meanings associated with data culture. I approach data cultures at the BAnQ as dynamic processes, not only requiring the alignment of new practices with existing frames but also exposing disparities and resistance between different interpretations. From this perspective, the formation and transformation of data cultures can be seen as arenas of debate, characterized by tensions and negotiations among professionals. Analyzing the relationship between these repertoires and frames of meaning sheds light on the adaptations and conflicts that institutions like the BAnQ encounter as they navigate ongoing changes in digital environments.

Methodology

The primary data corpus analyzed in this paper originates from a partnership research project conducted at the BAnQ from 2020 to 2023, which focused on the governance of usage data (Casemajor et al. 2021, 2023). While case studies of data practices in high-profile commercial platforms have received significant scholarly attention, public sector organizations remain comparatively understudied (Kennedy et al., 2022). Nevertheless, prior research has demonstrated the value of institutional data ethnographies—for instance, Bates et al. (2016) examined the movement of meteorological data using a “data journeys” methodology, and Tkacz et al. (2021) studied data practices in a national disaster monitoring center in Brazil using data diaries. This paper builds on these methodological contributions and situates its approach within the field of collaborative and partnership research. As Schäfer et al. (2024) argue, such approaches enable the production of knowledge with communities and in context. In this regard, the paper contributes to the development of critical, qualitative methodologies in data studies.

The research project underpinning this paper included a survey of the BAnQ subscribers to gauge their perceptions of how their usage data was collected, used, archived, and shared by the institution. Rather than analyzing the survey data itself, this article examines the preparatory and subsequent stages of the survey, which included the following steps: (a) understanding the meanings of data culture at the BAnQ by engaging with the partnership committee members in discussions during meetings and workshops; (b) inventorying the types of data collected at the BAnQ; (c) investigating the methods of data storage and usage (both current and projected); and (d) reviewing institutional policy documents (such as the BAnQ's digital strategy for 2021–2028), activity reports from the past five years, as well as internal regulations on data management. Through these activities, I gathered a comprehensive set of documentary and discursive materials, derived from interactions over three years with 15 members of the institution from the general secretariat and legal affairs departments, as well as teams involved with the digital experience, communications, education, the public library, the national library, and the national archives.

A secondary data corpus, consisting of documentary and media-related materials, was used to contextualize the case of the BAnQ along two dimensions: (1) data industries implications—specifically, the services provided by companies such as Microsoft that the BAnQ uses for its data management; and (2) public policies and administrative directives that are applicable to the BAnQ as a public institution. To further situate the BAnQ within the Quebec arts and culture sector, I drew on three group interviews conducted in 2021. These interviews involved a total of 16 cultural professionals (managers and project leaders) from the library, archives, performing arts, book, film, and music sectors. The aim was to identify the key stakeholders and principal issues associated with data culture in this professional sector. ⁵

A thematic analysis was conducted on the primary data corpus to identify statements related to the following themes: (i) the meanings attributed to the notion of data culture at the BAnQ; (ii) the variation in these meanings across different departments and professions within the institution; and (iii) the BAnQ's institutional initiatives related to data cultures and to the coordination of related activities. These statements were subsequently analyzed using three organizational goals that shape digital practices at the BAnQ: documentation of artifacts, public accessibility, and performance optimization. These three goals were identified by situating the statements within the context of the BAnQ's mission and digital strategy. The final phase involved examining how the meanings associated with data culture at the BAnQ change in response to the shifting dynamics of digital environments.

Various data practices, one common vision?

Understanding how organizational factors shape the formation and transformation of data cultures at the BAnQ first requires characterizing the data resources the institution produces, manages, and leverages to meet its objectives. Second, it entails examining the diverse data practices among its professionals and the management's efforts to unify these practices into a shared vision of data culture.

Mapping data resources and valorization practices

The BAnQ defines its data resources as “institutional data,” referring to “the set of data produced or received by the BAnQ . . . necessary for . . . documenting the achievement of its missions, conducting its operations, making decisions, and reporting on its activities” (BAnQ, 2022a: 4). Four main types of data (Table 1) are integrated across various data practices within the institution.

Transactional data are generated during commercial or financial transactions, such as payments to a supplier. (2) Identification metadata for items in library collections and archival fonds include descriptive information about the content (e.g., title or author) and administrative details (e.g., copyright). (3) Usage data are generated though user interactions with a digital system such as consulting a record on the institution's website or borrowing a document. (4) Heritage data encompass digitized cultural artifacts such as historical documents, records, manuscripts, maps, prints, posters, photographs, and postcards.

Table 1.

Main institutional data resources at the National Library and Archives of Quebec.

Type of data	Characteristics	Examples	Possible exploitation
1. Transactional data	Commercial or financial transactions	Payment to a supplier	Measure of organizational performance
2. Identification metadata	Descriptive and administrative information on the collections	Document title, copyright	Enhancing discoverability
3. Usage data	Interactions with a digital system	Document borrowing by a subscriber	Discoverability, personalization of services
4. Heritage data	Documents, records, manuscripts, and artworks in digital format	Text file, image file	Training AI models

The first three types of data are stored in a data warehouse—a centralized infrastructure designed to collect, manage, and analyze vast datasets from diverse sources within the organization. In this warehouse, borrowing data—records of items borrowed by users—serves as a key asset, since borrowing histories are retained rather than routinely purged. As of 2023, the BAnQ boasted nearly 400,000 subscribers, with over 100 million operations (including borrowings and returns) recorded in the data warehouse since the inception of the public library. In other jurisdictions, such as the European Union, frameworks like the General Data Protection Regulation (GDPR) prevent this kind of data accumulation.

Identification metadata and usage data are primarily used in strategies to enhance the discoverability of cultural content. Transactional and usage data are leveraged to support organizational management and optimize performance. As for heritage data, it is stored in specialized digital repositories designed to ensure long-term preservation and accessibility. When used as a training corpus for AI, the value of these heritage artifacts lies not in their individual significance as documents or historical works, but in the vast amount of textual and visual data they provide for automated analysis. ⁶ In this context, heritage institutions are increasingly called upon to contribute data for training generative models—such as those behind systems such as DALL·E and ChatGPT. In France, a report from the Artificial Intelligence Commission (Commission de l’intelligence artificielle, 2024) explicitly identified heritage data as a strategic resource with significant potential for use in francophone machine learning systems.

All four data types are regarded as strategic assets that drive innovation. However, the presence of multiple data resources, diverse practices, and differing objectives across the BAnQ's various departments presents a significant challenge: what does it mean to develop a unified data culture in this context? These questions have led the BAnQ to initiate consultations, bringing together its different entities and departments to forge a shared vision of data culture.

Organizational challenges in forging a shared vision

In 2020, while developing its new digital strategy, the institution established a task force dedicated to advancing its data initiative. Comprising 10 managers and staff members, the group formed a data-governance committee to address the legal, technological, and organizational challenges of data management. The initiative highlighted the need for better coordination of activities, with two primary goals: first, to forge a common vision of data culture at the BAnQ, and second, to ensure that this vision was embraced by all staff members.

The primary challenge of this data initiative was reaching consensus within the organization on unified principles to guide data exploitation. One factor explaining the significant differences in data-related activities at the BAnQ is that the institution comprises three separate entities (the national archives, the national library, and the public library). Although these entities share core values—such as providing access to knowledge and producing high-quality reference data—their data practices are shaped by professional cultures that are closely related yet distinct, leading to variations in norms and tools used. For example, librarians use data formats such as MARC (Machine-Readable Cataloging) for bibliographic normalization, whereas archivists adhere to other archival description standards like ISAD(G) (General International Standard Archival Description). Moreover, the data infrastructures they rely on are distinct: the system handling consultation requests for archival documents in the national archives operates independently from the one used to manage document loans in the public library.

Another factor contributing to these differences is the role of staff members, which is closely tied to their specific organizational functions. The data-governance committee included representatives from central departments, each having distinct relationships to data: the legal service focuses primarily on compliance with laws protecting personal information; the digital experience department is pivotal in advancing the organization's digital transformation strategy. This department, along with the communications service, is most supportive of leveraging data for strategic use. However, even within these departments, further distinctions can be made between individuals with “data profiles,” as categorized by Labelle (2017)—professionals who possess “advanced use and knowledge of data,” such as data analysts, marketing agents, and information system managers—and those with “nondata profiles”—individuals who may occasionally engage in data entry but for whom data exploitation is not central in terms of decision-making.

These distinctions underline the potential for divergent and sometimes conflicting visions of data exploitation within the same institution, which a member of the BAnQ data-governance committee has labeled as “impediments.” The definition of a data culture is therefore the result of a process of negotiation among different perspectives, as Lehmans (2017: 20) duly notes. Given the diverse practices and viewpoints on data culture at the BAnQ, an internal consultation was held to forge a collective vision. This consultation featured workshops with approximately 30 key professionals from various departments including legal, information technology, finance, and human resources. The aim was to identify principles that would garner “strong adherence from the participants,” noted a committee member. However, as of 2023, establishing such a culture remained a significant “challenge,” according to the same person.

Although the internal consultations held at the BAnQ were not directly observed as part of this study, it is still possible to pinpoint key areas of tension regarding data exploitation based on ongoing discussions in the cultural sector. These include the ethical considerations of automating certain professional activities and the resulting implications for the workplace. Additionally, profiling users by tracking their behavior in digital environments brings up legal issues related to the protection of personal information, as well as ethical issues concerning the perception of publics as audiences to target through their data profile. Disagreement may also involve struggles for position between different professions within the organization—for example, between communications or marketing professionals, whose work entails a certain familiarity with the exploitation of audience data, and librarians or archivists, whom a committee member described as “more conservative” with regard to the exploitation of usage data. Finally, hierarchical tensions may exist between a vision of data-led innovation primarily promulgated by management and the perspectives of the rest of the employees.

The data practices at the BAnQ are diverse and heterogeneous due to various organizational factors. Yet despite differences across the BAnQ's departments, professional cultures, and staff profiles, all are expected to contribute to the overall mission to provide democratic access to culture and knowledge. In the next section, I will examine how changes in data environments influence the BAnQ's operational goals through shifting frames of meaning.

Changing frames of meaning

Ongoing changes in digital data collection methods introduce new processes, priorities, and decision-making frameworks at the BAnQ. These shifts may influence the interpretative frameworks professionals rely on to understand their actions and assign value to data practices. As these repertoires of practice change, their interrelated frames of meaning change as well: where is continuity maintained, and where do more significant transformations emerge? Table 2 summarizes the three interpretive frameworks of data culture identified at the BAnQ. These frameworks, structured around the institution's core objectives—documenting artifacts, ensuring public access to archival fonds and library collections, and optimizing performance—have changed over time, giving rise to new challenges.

Table 2.

Framings and shifts of data culture at the Bibliothèque et Archives nationales du Québec

Frames of data culture	Shifts	Issues
Documentation of Artifacts
Indexing, Classification, Cataloging, Creation of Metadata	Open Data	New formats for metadata structuring, cross-departmental cooperation
	Automated “Companion” Tools	Development of new interprofessional skills, enhancement of data-processing infrastructures
	Generative AI	Provision of training data, compliance with intellectual property law, monetization of documentary heritage
Public Accessibility
Services Making Collections Available to the Public	Discoverability	Development capacities for new tools, service customization, privacy in data usage, diversity preservation
Performance Optimization
Service-Optimization and Strategic Decision-Making	Organizational Performance Metrics	Availability and reliability of activity data, hierarchical tensions
	Platformization	Enhanced data flow, data pooling

Documentation of artifacts

Data production within documentary systems is a key frame that shapes the meanings of data culture at the BAnQ. It encompasses activities such as indexing, classification, cataloging, and creating identification metadata that capture the characteristics, provenance, and usage rights of items in the library collections and archival fonds, as well as digitizing these artifacts, thereby producing heritage data (BAnQ, 2022b). Three major developments have recently impacted the BAnQ's documentary practices: open data initiatives, the development of automated “companion” tools, and the advent of generative AI. While some of these align seamlessly with the institution's guiding principles—such as producing high-quality reference data—others have introduced tensions, especially in relation to the datafication of heritage artifacts and the use of data resources for generative AI.

The first notable change in documentary practices at the BAnQ is related to open data. It involves producing datasets in a structured, standardized format that is freely accessible under an open license. Opening datasets was identified as a key priority in the BAnQ's new data-valorization policy. It is not merely an institutional initiative but also a mandate from government policy. As a public corporation, the BAnQ must adhere to “open government” directives that emphasize transparency and efficiency. In 2023, the BAnQ released five datasets on the governmental platform Données Québec (BAnQ, 2023). The datasets include a list of documents published in Quebec with ISBN identifiers and a list of films available at the BAnQ. The skills required for open data initiatives—such as knowledge of documentation standards, classification, and cataloging—are closely aligned with the expertise of librarians and archivists, and thus do not significantly disrupt existing practices. Additionally, providing free access to institutional data aligns with the BAnQ's mission of ensuring public access to records and collection contents. While open data initiatives present challenges, such as adopting new metadata formats (as noted by Lehmans, 2017, and Pinède, 2009è), the overall integration of open data practices into the BAnQ's documentary framework has not caused significant tension.

A second change in documentary practices at the BAnQ is linked to the development of “companion” tools designed to assist librarians and archivists. Powered by AI, these automated tools utilize identification metadata and heritage data in technologies for text recognition, content analysis, semiautomated cataloging (e.g., generating descriptions of works), and predicting the physical condition of documents to aid in conservation efforts. ⁷ While the BAnQ experimented with prototypes of these tools, their adoption in routine practices, apart from OCR for text extraction, is not yet widespread. Here too, these tools do not drastically alter documentary practices but rather enhance them through technological support. However, they introduce challenges, such as the need for new professional skills and greater cross-departmental collaboration, as their development requires expertise in both documentary practices and data science. Additionally, they pose challenges for data infrastructures, necessitating updates to data warehouses, modifications in data processing methods, and access to advanced computational tools, often requiring the pooling of resources.

A third transformation in documentary practices concerns the advent of generative AI. Automated tools like ChatGPT, which rely on large language models, require vast document corpora for training. Heritage institutions are increasingly called upon to provide this training data, exemplified by France's 2023 Digital Commons for Generative Artificial Intelligence project, launched by the Public Investment Bank. ⁸ This new approach to valorizing documentary corpora involves the datafication of heritage for industrial and commercial purposes. This shift creates tensions within the traditional frameworks used by librarians and archivists. On one hand, there are tensions around balancing innovation with intellectual property laws. On the other hand, new forms of monetizing documentary heritage for AI training bring additional complexities. For heritage institutions to participate in commercial partnerships with AI companies, as promoted by France's initiative, they must assign a monetary value to their collections. However, such financial operations—such as assessing the annuity value of national documentary heritage—lack well-established frameworks. Furthermore, this monetization strategy can conflict with open data principles, particularly when exclusive access to training data is granted to AI companies, allowing them to secure competitive advantages. This tension is reflected in the BAnQ's new data-valorization policy, which includes a clause allowing the default openness of institutional data to be overridden in favor of a commercial agreement with a third party, provided there is “serious and documented justification” (BAnQ, 2022a, 2022b). This could potentially undermine the ethos of open data, which advocates for equal and free access to data for transparency, innovation, and collaborative development.

Overall, documentary practices are changing with new paradigms of data valorization, positioning data culture as a catalyst for technological innovation and industrial application development. Yet, tensions emerge as heritage institutions strive to balance innovation with their legal and ethical responsibilities as stewards of cultural heritage. Reconciling these often conflicting priorities create a complex landscape for institutions to navigate.

Public accessibility

Public access to collections and fonds is another key frame that shapes the meaning of data culture at the BAnQ. Indeed, national libraries and archives are dedicated to democratizing knowledge, providing communities with resources to explore their historical and cultural heritage. The recent parameters of online content discoverability have introduced new mechanisms of ranking, findability, and recommendation to enhance public access to collections and fonds. But there is considerable ambivalence within BAnQ—and across many cultural institutions—about the implications of personalization in algorithmic recommendation systems.

At the BAnQ, “discoverability” encompasses three aspects: findability—the capacity to find what one is looking for; ranking—the ability of content to appear in a data stream without specific searches; and recommendation—the suggestion of specific content to users based on content similarity or user profiles and behavior. The BAnQ's new digital strategy prioritizes “further personalizing services . . . based on data and usage statistics that reflect [users’] habits and preferences” (BAnQ, 2021: 11). Personalization algorithms shift the focus from broad public access to highly customized, data-driven recommendations based on user profiles and preferences. The ramifications of this shift go beyond professional adaptation; they also redefine the relationship between cultural organizations and their audiences, who are increasingly segmented and targeted according to consumption patterns (Beer and Burrows, 2013).

While personalization can improve access by tailoring recommendations to user preferences and behavior, a member of the BAnQ's data-governance committee emphasized, “we don’t want the recommendations to be too personalized.” This ambivalence highlights the tension between the efficiency of algorithms and concerns about their broader impacts. Many professionals believe that personalization can optimize the match between content and audience, as reflected in the BAnQ's aspiration to “provide the right information, in the right way, to the right person, at the right time” (BAnQ, 2021: 11). This vision assumes that usage data are comprehensive, objective, and transparent—a tenet of dataism, which suggests that faith in the ability of data to reveal hidden truths and provide insights that might otherwise, be inaccessible through traditional methods. It is also reflected in the belief that “behavioral data let publics speak” and can empower “silent users,” which was common among professionals in Quebec's arts and culture sector interviewed for this study. These perceptions resonate with issues identified by Hallinan and Striphas (2016), who examine the shifting meanings of culture in the context of the growing presence of algorithmic recommendation systems, such as those used by Netflix. They highlight how, within an algorithmic culture paradigm, the dominant approach to recommendation systems has shifted toward more predictive models tailored to individual taste preferences.

However, various professionals expressed concerns that overly precise algorithmic personalization could undermine the broader mission of cultural democratization. What they challenge is not the idealized potential for perfect optimization between user and content, but rather the implications of excessively powerful profiling and taste prediction mechanisms. The perceived risk of a filter bubble effect—whether real or imagined—serves as a significant deterrent. The BAnQ staff members are keen to avoid confining users to narrow, highly targeted recommendations and instead aim to broaden horizons by offering unexpected and diverse discoveries. A frequently cited example to avoid is Amazon, originally a book retailer, whose approach is viewed by librarians as limiting the diversity of cultural offerings.

To sum up, the new parameters of discoverability have created tensions between the pursuit of efficiency and the need to maintain diversity in cultural offerings. While personalized recommendations can improve access by aligning content with user preferences, they also risk narrowing the cultural experience. This contradiction highlights a key challenge at the BAnQ: striking a balance between the advantages of personalization and the broader goal of cultural democratization.

Performance optimization

A final key frame of data culture at the BAnQ centers on organizational management. In this context, data culture is closely linked to strategies aimed at optimizing performance through data-driven decision-making. It is grounded in a principle of industrial efficiency, whereby the datafication of professional activities seeks to streamline and rationalize the organization's services. From this vantage point, institutional data are regarded as strategic assets—critical resources essential for achieving management objectives and sustaining a competitive edge. This framing of data culture is currently undergoing two major shifts: the first is the increased tracking and measurement of organizational performance; the second is the growing lure of platform-based models.

The datafication of activities for management purposes at the BAnQ is primarily aimed at optimizing services and resources, a goal strongly supported by upper management. As the CEO suggests, discoverability is treated as “a question of operational efficiency” (Grégoire, 2023b: 8), which can be quantified using performance indicators, such as the frequency of online consultations of heritage documents. These management practices leverage earlier decision-making systems, notably Enterprise Resource Planning (ERP) software, as highlighted by Flichy (2013). Although these systems have been in use since the 1990s, their development has accelerated significantly over the last decade, driven by new AI data-analysis tools that enhance strategic decision-making.

At the BAnQ, this orientation is reflected in management's directive to establish a “datamart.” According to the CEO, “From now on, project production is based on the use of dashboards, and targets are set for each of them” (Grégoire, 2023a: 14). Microsoft's Power BI system is used to facilitate data-driven activity reports, enabling the BAnQ to “make better decisions with data,” as Microsoft claims. However, defining precise indicators to achieve these targets is a significant challenge, complicated by concerns over data reliability. Moreover, relying on activity data to drive decisions top-down can create tensions within the organization, as employees may feel that management prioritizes key performance indicators over the meaningfulness of daily tasks and workplace quality. These concerns are exacerbated when the validity of performance indicators is debatable, heightening concerns over their impact.

The focus on data culture as a means of optimizing performance is intertwined with the appeal of platform-based models. It involves integrating data from various departments or organizations into a centralized system, streamlining analysis, and facilitating reuse by different stakeholders (Cristofari, 2023; Helmond, 2015). In this framing, data culture is associated with improving the connectivity and fluidity of data flows, both internally and in its interactions with external entities like the government.

A central goal of the BAnQ's data initiative is to streamline interactions between “principal data holders”—employees and managers who generate data on a daily basis and control its access (BAnQ, 2022b: 4). The term “holder” not only implies possession but also suggests the potential to “retain” data. Streamlining these efforts requires removing barriers to data circulation. As the CEO puts it, the goal is to “knock down the walls between the organization's units and services” to foster interdepartmental connectivity (Grégoire, 2023a: 14). This approach reflects the OECD view that organizational fragmentation hampers data flow, framing these barriers as legacy “fragmentation” challenges that result in “siloed policies” and obstruct the creation of an integrated, connected organization (OECD, 2019: 14, 24, 26). While it is true that each department at the BAnQ tends to develop its own data production and management tools, leading to heterogeneous data practices, the injunction to dissolve these distinct identities reflects a neoliberal management style that carries implications. Eroding the social boundaries and practices that foster a sense of belonging within organizational units can profoundly alter shared work experiences and the fabric of workers’ lives, as observed by Thomas and Davies (2005).

The push to streamline data circulation at the BAnQ extends beyond internal processes, significantly affecting its external relations with the government. As a strategic provider of government digital data, the BAnQ is obligated to ensure the seamless mobility of its data, requiring it to facilitate their transfer to government bodies and other public entities. This mandate highlights the pivotal role of data as a critical resource in governmental operations, aligning the BAnQ's data management practices with broader public sector strategies for data utilization and governance.

Data gathering systems have long been integral to the exercise of governmental power, as Desrosières (1998) demonstrated, with national bureaus using statistical measures to aid administration and inform policy. However, the rise of platform-based data models has shifted this approach toward a big data paradigm, promoting a “data-driven public sector.” In the OECD's (2019') advocacy for “government as a platform,” fostering a data culture is seen as a key strategy for creating public value and strengthening governmental capabilities in policy making, service delivery, and performance management.

Yet organizational models inspired by the functioning of digital platforms tend to overestimate the potential for automating data flows, often colliding with heterogeneous data practices and infrastructure limitations. Moreover, the classification of such government model as “platform” is questionable, particularly when they primarily involve data centralization within a governmental apparatus, as seen in Quebec, without granting all stakeholders access to develop their own analytical applications. This approach may be better described as a platformization fantasy superimposed on a system of governmental data centralization.

Ultimately, the two major shifts in framing data culture as a performance optimization strategy at the BAnQ—intensified activity measurement and a move toward platform-based models—carry significant ideological weight. These changes promote a neoliberal management approach and idealize the automation of data flows, mirroring broader trends in technological solutions that often overlook the complexity and diversity of organizational data practices.

Aligning data innovation with public service mandates

The analysis of the BAnQ shows that the challenge of cultivating a unified data culture extends well beyond the simple tasks of data production and management. It involves reconciling diverse data practices and conflicting logics rooted in organizational structures, professional values, institutional goals, and governmental requirements. More critically, it highlights the challenge public institutions face in navigating competing priorities while aligning data innovation with their public service mandate. Especially since the BAnQ's mandate spans several key domains: the stewardship of cultural heritage, the democratization of knowledge, the provision of data to the government for public policy planning, and the implementation of best practices in data governance. Data governance, in particular, serves as the guiding thread, striving to ensure that innovation in data remains consistent with BAnQ's broader public service ethos.

In an era of growing skepticism toward data practices, institutions like public libraries act as ethical compasses. As Ruttan et al. (2019) emphasize, public libraries are seen as institutions with “the necessary resources, expertise, and public trust” to manage citizens’ data responsibly (Ruttan et al., 2019: 1). This position was illustrated in the Quayside smart city project in Toronto, where public libraries were envisioned as trusted civic data hubs, counterbalancing Google's control over urban data collection. As public libraries are trusted stewards of civic data, BAnQ is similarly tasked with upholding its ethical obligations while integrating innovative practices. Its role in setting data governance standards is crucial, particularly as it serves as a model for other cultural institutions and governmental agencies in Quebec and beyond in the francophone world.

In BAnQ's efforts to align data innovation with best practices in governance, one major tension arises from restricting access to heritage data for commercial purposes through partnerships for AI training, which can conflict with the institution's commitment to open and transparent access. Similarly, leveraging user data to improve discoverability introduces challenges in preserving the privacy and trust of library users, highlighting the difficulty of maintaining ethical standards while pursuing innovation. While Kennedy (2016) duly notes that mining user data can yield valuable insights and foster civic engagement, the risk of violating user privacy remains a significant concern. The BAnQ must also carefully navigate between fulfilling governmental requests for strategic data and its ethical responsibility to protect user privacy. The increased transfer of various types of data to governmental bodies has raised concerns among librarians about state surveillance. Across the border, the USA Patriot Act (2001), which permitted the U.S. government to monitor library users’ reading habits in the name of public security (Shaffer, 2014), led many libraries to strengthen their privacy protections (Thielman, 2016).

Moving forward, the BAnQ must continue to develop a model that aligns with its public service mission while embracing opportunities for data innovation. Achieving this balance demands careful attention to data quality, privacy, and transparency to ensure that the institution's core goals—democratizing knowledge and preserving cultural heritage—remain intact. Success in this endeavor will require not only internal organizational focus but also active engagement with users and government bodies. However, as Steedman et al. (2020) caution, institutional-level solutions alone may not be sufficient; broader collective or macro-level approaches are required to further address public concerns over data governance. Indeed, while the organizational level is key to understanding negotiations around the practices and meanings of data culture, it must be considered in relation to national power structures and global market forces.

Conclusion

In this article, I have analyzed the specific configuration of meanings and practices related to data culture at the BAnQ, with the objective of showing how data culture changes in the context of new means of production and automated data analysis. First, this analysis highlighted how the datafication dynamic extends to ever-broader aspects of library and archive activity, such as the datafication of publics through exploitation of usage data, and the datafication of collections and by fonds for generative AI training. The expanding scope of datafication at the BAnQ multiplies the resources around which the BAnQ aims to build a data culture.

Second, the analysis elucidated how the BAnQ's institutional strategy to foster a unified vision of data culture confronts significant challenges due to the diverse functions across departments, profession-specific frames of reference, and varying data-usage profiles within those professions. It would therefore be reductive to assume that librarians or archivists operate within internally homogeneous data cultures that sharply distinguish them from other professional sectors. These fields are themselves shaped by internal tensions and competing interpretations. As Monjardet (1994) argues, the so-called “culturalist” approach to professional culture often compresses a plurality of attitudes, values, and practices into a fixed identity attributed to a group. This view reinforces the idea that data cultures—even within the same professional domain—are never settled, but continually negotiated, fragmented, and plural. At the BAnQ, the various perspectives and practices regarding data, though aimed at fulfilling the same organizational goals of preserving, documenting, and providing access to collections, often manifest distinct and sometimes contradictory sets of meanings and principles of action.

Moreover, it is important to emphasize that changes in data culture are not linear or progressive, as might be implied by the notion of acculturation to data. This term presumes a unidirectional evolution toward a coherent and more efficient set of practices, often framed as a vector of organizational progress. Such a teleological perspective neglects the frictions, ambiguities, and iterative negotiations that characterize data practices in everyday institutional life.

Instead of being resolved, the tensions at the BAnQ appear to be constitutive of data culture itself, reflecting a dynamic and continuous state of negotiation rather than a stable consensus. Such contradictory sense is not anomalous but inherent to meaning-making processes in complex organizational settings. From this perspective, institutions like BAnQ do not simply “build” a data culture—they continuously assemble, translate, and sometimes mistranslate between partial and situated understandings. Borrowing from Lin and Law (2019: 126), translation refers to “locally specific attempts to assemble fragile but workable practices,” while mistranslation invites attention to how divergences can be rendered explicit and generative. This implies a shift in focus—from seeking unification to foregrounding the relational processes through which meaning is negotiated.

Beyond the specific case of BAnQ and the challenges heritage institutions face in the digital age, the paper's findings shed light on key dynamics of change in data culture that are likely relevant to other cultural organizations and public institutions. The study reveals how the professional adaptation to data-driven practices creates tensions between embracing new technologies—such as AI and automation—and upholding established values. It also underscores the need for public institutions to develop data governance models that balance data innovation with their public service missions. These challenges extend beyond individual institutions, intersecting with broader power dynamics involving users, governmental bodies, and the data industry.