Wikipedia: A Foundational Source in the Age of AI?

Wikipedia is arguably a staple reference source in the information diet of Canadians and in countries around the globe, with roughly 4.4 billion unique global visitors to the site in March 2024 alone (Bianchi 2024). This reach is impressive, highlighting Wikipedia’s role as a key source that people turn to for information. Routinely, this source is integrated into Large Language Models (LLMs) for use in generative AI platforms like Chat GPT. Through this, Wikipedia’s significance and reach are further extended.

Chief product and technology officer at the Wikimedia Foundation, Selena Deckelmann (2023), stated that “Wikipedia is one of the largest open corpuses of information on the internet, with versions in over 300 languages. To date, every LLM is trained on Wikipedia content, and it is almost always the largest source of training data in their data sets” (Par. 3). As an encyclopedia, Wikipedia influences the information landscape significantly, shaping the foundational knowledge that both humans and artificial intelligence systems rely on.

Thomas (2024) notes, “Any biases present in Wikipedia content are at risk of being absorbed into the foundational parameters of contemporary AI systems” (Par. 1). It is well documented that Wikipedia content contains biases along gender and racial intersections (de la Peña McCook 2024; Jemielniak and Wilamowski 2017; Smith 2021), and studies show there may also be political bias in Wikipedia content (L. Thomas 2024). Critics have pointed out that as an open and crowdsourced encyclopedia, Wikipedia has inequities that influence the way information is presented and managed on the platform. “Facts” are produced and distributed, from a primarily male, North American/European perspective (Ford 2022). Wikipedia’s content “draws liberally on eighty odd years of neoliberal techniques and epistemologies, albeit in a piecemeal and often less-than-conscious manner” (Tkacz 2015, 179).

Though less-than-conscious, biases in content and operations can perpetuate systemic inequalities and influence the perspectives presented on Wikipedia. The lack of inclusivity on Wikipedia is noted as a limitation on its progressive potential (Luyt 2012). And this is the crux of the matter. As Wikipedia editors, Library and information science (LIS) practitioners can engage in an ongoing and conscious way in critical editing practices in pursuit of equity in knowledge representation and a public good in Wikipedia content. These efforts extend beyond the platform and into other sources that rely on Wikipedia, such as AI-generated content.

Librarians as Critical Editors and Stewards of Knowledge?

As librarians and library workers, we are encouraged to edit Wikipedia. Thomas (2021) suggests that we embrace “critical editing”. This means being aware of power dynamics, focusing on critical literacy, challenging social norms, and continuous self-education. Critical editing praxis among information professionals can help to address Wikipedia’s content gaps and biases while referencing from a variety of resources. Michael David Miller (2021) frames librarian contributions within a lens of our privileged positions as information professionals. He writes, “We can and should be leveraging our own library’s collections and working with colleagues in other institutions to help make the information on Wikipedia more robust and accurate because it is the most used source of information that our patrons, colleagues, friends, family, and acquaintances use” (2021, 115–16).

Writing about encyclopedic development, de la Peña McCook (2024) notes that Wikipedia has superseded all prior encyclopedic initiatives, and she reflects on how librarians, with bibliographic understanding, can add to the depth and reliability of this widely used resource. She writes, “By encouraging librarians to go behind paywalls to enhance the citations on Wikipedia, the never complete project of a universal encyclopedia improves” (2024, 574).

Surrounding concepts of “critical editing” and the “never complete project of a universal encyclopedia” expand the list of arguments for Wikipedia editorship among librarians. They highlight how librarians with expertise in curating, validating, and organizing information are inimitably suited to contribute to Wikipedia. It is a dynamic and ever-evolving repository of human knowledge, which needs constant input and refinement, in alignment with the skills librarians embody. Critical editing, with its focus on reducing biases and gaps in information, aligns well with the American Library Association’s “Core Values of Librarianship” in principles of information access, information equity, and information as a public good. Wikipedia, as one of the most visited knowledge platforms globally, provides an ideal avenue for librarians aiming to foster equity in knowledge systems.

Wikipedia and the Library Profession

During the early years of Wikipedia, and to an extent still, there has been concern with its use in library circles. Noting that “Wikipedia never closes”, one of the earlier concerns with the platform is one of competition in the information market (Batt 2009, 60). In relation to the platform’s open editing policies, among other aspects, it was found that “librarians are cautious in advocating for Wikipedia in their work life” (Snyder 2013, 161).

We can note that the growth of collaborative efforts between the library profession and Wikimedia flourished after the first, GLAM-Wiki Conference of 2009, held in Canberra, Australia. At this conference, recommendations were made to “encourage sustainable collaboration between institutions in the cultural sector and the Wikimedia community” and to “create a rotating position of "Wikimedian in residence" within the GLAM sector (Galleries, Libraries, Archives, Museums) to allow tertiary students to gain experience in the professional sector and to have their Wikipedia efforts recognized” (“GLAM-WIKI Recommendations - Meta” n.d.). This spurred a wave of Wikimedian in Residence positions, beginning with Liam Wyatt’s 2010 volunteer position at the British Libraries.

Within the following year, similar positions were formed at cultural institutions around the globe. Within the Canadian context, a decade after the first Wikimedian in Residence (WiR) position was formed, there have been several such formal positions in Canada (see Table 1). In writing about their experiences, three WiR from institutions across Canada concluded that WiR positions “are critical to building meaningful connections on campuses for expanding Wikipedia and engaging with its complex issues of bias, representation, notability, and editor alienation” (O’Neil and Severson 2021, 228).

Figure 1 - Representative Timeline of Library and Wikimedia collaboration, predominantly in Canada!

This timeline is non-exhaustive.

WiR = Wikimedian in Residence

Source

XXXX-XX-XX en

These formal WiR positions in libraries are coupled with more informal positions where LIS practitioners are practicing Wikipedians, or contributors to Wikimedia’s sister projects, Wikidata and Wikimedia Commons. For example, Anne-Marie Boisvert at Bibliothèque et Archives nationales du Québec (BAnQ) has led ‘Mardi, c’est Wiki’, and I have led ‘Wiki Tuesdays’ at Western University. Both these events provided an ongoing drop-in space at our respective institutions for librarians and other academics to have support in Wikipedia editing. These spaces are used to advocate for Wikipedia in academic settings by organizing edit-a-thons and other editing activities. Similarly, Jean-Michel Lapointe, a librarian at l’Université du Québec à Montréal (UQAM), works with faculty advocating for the integration of Wikipedia editing into course assignments (Aschaiek 2019). Wikipedia complements traditional approaches to teaching with a learning activity in collaborative knowledge building (Petrucco 2018). Wikipedia assignments can also help students to see the value of what they are learning beyond the classroom (Finkel 2024).

Today, many library professionals have embraced the collaborative spirit of GLAM-Wiki, getting involved with Wikimedia and organizing events like edit-a-thons that seek to improve the accuracy and breadth of content on Wikipedia. As an outreach project, libraries have used the #1Lib1Ref campaign to organize events (Lubbock 2018, 57). Another example of a library-led Wikipedia initiative is the #CiteNLM Wikipedia edit-a-thons hosted across academic libraries (Cowles, K. et al. 2021). These types of events have been spaces for conversation among library staff about Wikipedia and their collections (Sengul-Jones 2018, 228–29).

Librarians in Canada are now also embedded within the organizational structure of Wikimedia Canada. From 2019-2022, librarian Lëa-Kim Châteauneuf was President of Wikimedia Canada, and librarian Michael David Miller has been Vice-President of Wikimedia Canada since 2019. Additionally, Jean-Michel Lapointe, has been actively involved as a member of the board of directors for Wikimedia Canada. It is noteworthy that this is a strong representation of LIS practitioners as Wikipedians from Québec, presumably editing both French and English Wikipedia. A motivating factor in this involvement might be a desire to “contribute to French-language content on Wikipedia in order to better represent the perspectives of Francophone Canadians on the platform” (Aschaiek 2019, par. 17).

1Lib1Ref: A Call to Action for Librarians

Thinking about Wikipedia as a never-ending open information project to which librarian skillsets and privileged access to information can add depth and reliability, One Librarian, One Reference (1Lib1Ref) is a campaign aimed at getting librarians and library workers to add references to Wikipedia articles and a way that we, as library workers, can engage with Wikipedia as novice editors. Reflecting on 1Lib1Ref, Jake Orlowitz (2018a) frames it as an online micro contribution campaign. His viewpoint is that the campaign serves to rewrite the story of the relationship between Wikipedia and libraries. In 2016, 1Lib1Ref received formal support and encouragement from the International Federation of Library Associations and Institutions (Scheeder 2016). The Canadian Federation of Library Associations followed in 2019 encouraging librarians and library workers to participate in the campaign (FOPLED n.d.).

Globally, libraries are now collaborating to improve Wikipedia pages in multiple languages through the #1Lib1Ref campaign (Gutiérrez 2023). An example of this is #1Bib1Ref in Latin America, which began in May 2018 “to commemorate the birthday of Spanish Wikipedia”. The May campaign includes libraries across the southern hemisphere in countries such as Ghana, Uruguay, Argentina, South Africa, Brazil, Australia, and India (Nartey 2018). “While the January version of the #1Lib1Ref campaign works well mostly for the northern hemisphere, it doesn’t for the countries below the equator” (Hinojo 2019). Within the northern hemisphere, four libraries in Québec led Canadian libraries’ participation in the #1Bib1Ref/1Lib1Ref campaign by prompting a friendly competition through Twitter to see who could add the most citations to Wikipedia (Orlowitz 2018a). The friendly competition came together virtually through social media and joint Wikimedia Dashboards. Competitions, like that in Canada, have been one of the “great drivers of engagement” in the #1Lib1Ref campaign (West 2018).

Library Professionals as Wikipedians

I published a study in CAIS conference proceedings that examines the editing patterns of the #1Lib1Ref participants within a Canadian context to determine if its newly recruited editors continue to engage in sustained Wikipedia editorship (Bourgard 2024). This research asks the question: Do #1Lib1Ref recruit editors go on to edit Wikipedia outside of the #1Lib1Ref campaign period. It examines data across six instances of the annual campaign in January.

For this study, sustained Wikipedia editorship is achieved when an editor has contributed to Wikipedia in 2022 or later, has edited during at least three months of the observed period, and has contributed to Wikipedia outside of the campaign period. Editors who first edited Wikipedia during the campaign period are defined as recruit editors because they signed up as #1Lib1Ref participants and first edited Wikipedia during the campaign. They are compared to other editors,

those participants who first edited Wikipedia outside of the January #1Lib1Ref campaign period. Sustained Wikipedia editorship has been reached for almost 22% of #1Lib1Ref recruited editors, however, most do not go on to edit outside the campaign period.

Interestingly, there is a small contingent of inactive Wikipedians (18% of Other Editors) who have added their username to #1Lib1Ref dashboards without adding any edits within the data’s timeline, January 2018 to June 2023. This small group of inactive #1Lib1Ref participants signals that engagement in the campaign is motivated by more than a desire to contribute.

The impact of #1Lib1Ref extends beyond counts of citations added to Wikipedia and into community building and outreach. “This kind of campaign builds awareness and interest among the library community and allows for a dialogue that further shifts the Wikimedia community closer to the needs of libraries, and libraries closer to the goals of the Wikimedia community” (Orlowitz 2018b, 81). The #1Lib1Ref campaign is a form of virtual community building for libraries that is global in scale.

While there is demonstrated motivation in Canada to engage in the #1Lib1Ref event, considering the editing patterns of participants, there are limitations for recruit editors and those inactive editors to becoming active and sustained Wikipedians. Professionally, we may continue to carry caution in our alignment with the platform. Perhaps, however, there are other factors such as workload or lack of formal recognition within librarianship as limitations to participation.

Final Thoughts

Recently, I’ve noticed that in library circles generative AI is being discussed with frequency. High-level examples of the recent focus of libraries on AI include the Canadian Academic Research Libraries’ (CARL) document Generative Artificial Intelligence: A Brief Primer for CARL Institutions, and the Ontario Council of University Libraries’ (OCUL) Artificial Intelligence and Machine Learning Program (AIML). Moreover, CAUT librarians have expressed unease with generative AI. “It seems like there is an entrant in every area of the knowledge economy trying to sell an AI-based tool” (Ribaric and Farnum 2024).

Thinking about Wikipedia as a critical foundation for Large Language Models in artificial intelligence, librarians can play a key role in ensuring equitable knowledge representation by becoming active Wikipedia editors. Such actions can improve Wikipedia’s coverage and, in turn, enhance the reliability of AI systems. With the buzz of generative AI stirring in library and information spaces, the critical act of Wikipedia editorship should be considered as a core practice in information management. It is the content creation and maintenance of the most widely used encyclopedia by humans and machines. As LIS practitioners, we can start with smaller edits and evolve our contributions into full article creation. One Librarian, One Reference (1Lib1Ref) is an impactful way that novice editors within the library profession can begin to engage with Wikipedia and critical editing praxis.