Towards the Integration of Generative AI Conversation Tools into the Creation of Wikipedia Articles within a Postdigital Literacies Approach

Wikipedia as an Open Educational Resource (OER)

Wikipedia is an example of an open educational resource (OER) sustained by the voluntary collaboration of the community. It is a collective effort whose motto is: “Imagine a world in which every single person on the planet is given free access to the sum of all human knowledge. That's what we're doing.”

Both UNESCO's Recommendation on Open Educational Resources (OER) (2019): “[…] educators are creating a world where each and every person on earth can access and contribute to the sum of all human knowledge,” and the Cape Town Open Education Declaration (2007): “[ICT] can open possibilities for OER to be accessible anytime and anywhere for everyone”, align perfectly with Wikipedia's mission to expand access to knowledge generated by humanity. Therefore, participating in editing this encyclopedia is an act of collaboration to maintain this participatory space, which Petrucco & Ferranti (2020) also describe as a “true learning environment, well-organized with precise rules that can stimulate the adoption of Open Educational Practices,” a view echoed by McDowell & Vetter (2022).

Beyond being an encyclopedia, Wikipedia is a community that keeps this platform, housing humanity's knowledge, updated and accessible. Wikipedia is an open resource that allows free access, reuse, and adaptation of its content. Throughout its existence, Wikipedia has incorporated various forms of artificial intelligence into several projects. However, as of 2024, the Wikipedian community still lacks a clear consensus on how Generative AI (GenAI) can contribute to the encyclopedia's growth, although various possibilities have been discussed at the Wikimania events in 2023 and 2024.

Generative Artificial Intelligence (GenAI) is evolving at such a rapid pace that it is difficult to fully grasp the implications of one tool before another, with enhanced capabilities, emerges. For education in general, and particularly for open and online education, the implications of GenAI are just beginning to be addressed—at best through general guidelines developed by universities. However, in daily practice, educators are now faced with the need to rethink both the design of classroom assignments and traditional evaluation methods.

Editing content on Wikipedia pages offers an alternative approach in this context. Ideally, this would involve a semester-long process in which a group collaboratively develops specific improvements or creates articles from scratch, contributing to the encyclopedia's enhancement. Wikipedia is an open environment, both as a resource for consultation and as a fertile field for work within a real-world interaction context. It is thus doubly open: as a learning resource and as an educational space.

Wikipedia meets Generative Artificial Intelligence (GenAI)

The emergence of tools like ChatGPT initially gave the impression that Wikipedia articles could be quickly written with nothing more than a prompt. This led to a legitimate concern within the Wikipedian community that the arrival of GenAI might result in the encyclopedia being filled with automatically generated content without proper human review, as well as a significant reduction in the use of Wikipedia for consultations. However, the latter has not occurred (Reeves et al., 2024). Additionally, platform updates have been implemented to help identify generative AI-created content, such as the introduction of an option on Wikimedia Commons to label such material.

Tools like STORM have proven highly effective, especially for drafting initial versions of explanatory texts. However, it is acknowledged that improvements are still needed, particularly in terms of neutrality (Shao et al., 2024). For now, these tools are not expected to replace the volunteer community, and their effective use remains reliant on community contributions (Ackerman, February 26).

There are also initiatives exploring collaborative work where generative AI tools are used to support and enhance activities like summarizing or translating (Brooks et al., 2024, p. 5). Additionally, some studies view AI as a tool that can act as an equalizer (Choi et al., 2023).

GenAI with a retrieval-augmented generation (RAG) approach

Beyond ChatGPT, other tools are emerging that facilitate dialogue with academic research articles—through interfaces like SciSpace, Epsilon, Elicit, or NotebookLM. Retrieval-Augmented Generation it is a methodology where a generative AI model is combined with a retrieval mechanism to enhance the accuracy, relevance, and depth of its responses. These GenAI interfaces provide a novel way to engage with more precision and nuance when interacting with academic content, which can be significantly enhanced with the guidance of a mentor to ensure their critical and effective use in creating Wikipedia articles, particularly in Spanish.

English Wikipedia contains nearly 7 million articles, while Spanish Wikipedia is just approaching 2 million. Wikipedia is not new to higher education settings (Evenstein Sigalov, 2023; Remmik et al., 2024) and has been envisioned as a pathway to fostering “epistemic agency” (Nieminen et al., 2024). Therefore, training students to create Wikipedia articles could help bridge this knowledge gap while benefiting the entire Spanish-speaking community.

Wikipedia as OER + GenAI: a specific case

Just as the Midjourney platform prompts users to begin their commands with the word "Imagine" to generate AI-created images, a possible interaction using generative AI tools to create Wikipedia articles is outlined below:

/Imagine a group of students from open or distance learning programs who need to fulfill the community service requirement (480 hours of community work) by creating missing content on Spanish Wikipedia related to their fields of study. These students would work alongside an academic mentor and a small community of peers, engaging online to share experiences and learnings from this activity

For students in open and distance modalities, finding opportunities to fulfill this graduation requirement can be particularly challenging, often delaying the completion of their studies. This activity would be carried out over six months, guided by a tutor who oversees the editing process, with additional support from members of the Education Program of Wikimedia Mexico's chapter.

Why integrate GenAI?

In the proposed scenario, the goal is to create or improve Wikipedia articles as an academic activity that fulfills a community service requirement while enriching the content of the encyclopedia. However, one of the major barriers to creating Wikipedia articles is not access to the academic articles that support editing, as universities typically have extensive access to databases. The challenge lies in understanding these sources, which can be intimidating, as well as the responsibility of editing a Wikipedia entry that will be viewed by a wide audience.

Experience has shown that successful Wikipedia editing requires careful reading of academic articles, which can be difficult for students. As a pilot activity, the use of GenAI interfaces employing a Retrieval-Augmented Generation (RAG) approach is proposed to interact with the provided information. This allows students to engage initially with academic articles, giving them the opportunity to form an opinion before attending online sessions where they receive guidance for the editing process.

For the proposed experience, the tool NotebookLM (Google) was selected. In addition to enabling written conversation via chat with sources, it offers the option to generate an audio summary in podcast format, where two individuals engage in an animated discussion about the provided information. The tool allows the integration of up to 50 multimedia sources (audio, video, text, PDFs, presentations), enabling the generation of a conversation focused exclusively on this material.

Openness coexisting with proprietary resources

In an ideal scenario, one might envision a proposal that exclusively uses open resources, but reality shows that this is not easily achievable. The academic articles to be used in this project are not all open access; some are, while others are accessible through the databases provided by the university. NotebookLM is not an open-source tool; however, it is free to use and currently claims that user data is not used to train its language model. A Google account is required to access the tool.

First stage: conversations to establish a baseline

As an initial interaction, this group of students is expected to engage with both the chat-based conversation function and the active listening of the generated podcasts. This last proposal aims to explore the effect of hearing opinions from a “third place,” allowing students to approach the basic concepts of a topic through an informal conversation. This format facilitates the identification of key elements at play, helping students gain perspective.

This tool is not intended to create complete articles; rather, its purpose is to establish a foundational understanding that enables the group to engage in deeper human conversations about the topics to be edited. Among its advantages are the opportunity to hear the pronunciation of specialized terminology, and the relationships automatically established between the provided sources.

One noted drawback is that the podcast hosts tend to maintain an overly optimistic tone, which can, however, be adjusted using a prompt to guide the audio’s intent. The fundamental idea is that interaction with this tool serves to foster better human conversations.

Second stage: group conversations

In open and online education, interactions often occur through platforms where instructions are provided in writing for students to complete autonomously and submit later. While forums are available, they do not always foster meaningful interactions, making complex projects a challenging management task for both educators and students. Therefore, the next step is to create an interactive space that does not rely on traditional videoconferencing. The tutor accompanying this online project will not simply open a videoconferencing session as a help desk. Instead, it is suggested to use a virtual classroom like BigBlueButton (BBB), which, in its 3.0 version, integrates via a sub-plugin with Moodle 4.5, and both tools are open-source. BBB prides itself on being a virtual classroom designed under active learning principles to ensure that each student genuinely learns and does not merely attend a session.

The project will require, for its launch, a mentor responsible for 10 students, each of whom will create six Wikipedia entries, one per month. The selection of articles will be made by consensus, considering student interests, fields of study, and content gaps identified on Wikipedia. Students already have access to a computer, academic databases, and an internet connection. Additionally, guidance and support could be provided by the staff of the Education Program at the local Wikimedia chapter, with whom there is already close collaboration. Academic support may also come from the BigBlueButton community, whose members are eager to share and receive feedback from BBB use cases.

Online sessions will allow the tutor to interact with the group participants, who will share their insights from the chat-based conversations and podcast listening activities. These sessions will focus on how best to articulate and detail the elements of both interactions to guide meaningful learning experiences. Here, the role of the educator is crucial in mediating and shaping ideas.

At the end of this project, 60 new articles on Spanish Wikipedia are expected. A project of this nature can contribute to the graduation of students in open and distance learning modalities. Additionally, students who complete this project could become mentors for new generations (with future funding required). During this initiative, interactions with GenAI research tools, the challenges of editing Wikipedia articles, and community interaction through BigBlueButton analytics will be observed and subsequently analyzed.

Third stage: reflective conversations on postdigital literacies

Finally, a closing conversation is proposed as a way to reflect on the work accomplished. Community service does not require a grade, only accreditation, making this an ideal environment at the intersection of formal and informal education. While students often evaluate their classes and teachers, they rarely evaluate or reflect on their own learning through a critical lens, questioning the effects of their interactions with technology.

What does it mean to have contributed knowledge to Wikipedia with the support of GenAI interfaces and human conversations? What does that signify, and how was it achieved? These complex learning environments are described by Lacković, Olteanu, and Campbell (2024) in their discussions of entanglement, digital materiality, and spaciotemporality as characteristics of postdigital literacy events. They also explore the idea of interfacing or transmedial meaning-making to refer to the phenomenon of transitioning across various platforms and still achieving moments of learning. Similarly, Fawns (2022) refers to this kind of complex interaction as entangled pedagogies, while Fawns and Nieminen (2023) suggest collective and communal ways of knowing in discussions about assessments.

The purpose of the closing session is to delve into the experience and the relationships that emerge from this interplay and transitions between different digital interfaces and human interaction. It is a moment to reflect on the postdigital literacy that has developed and to explore the role of social generative AI for education, as suggested by Sharples (2023).

Students will be encouraged to attend this last season with anecdotes recorded in writing, preferably handwritten, which can later be analyzed as part of a (post) qualitative methodology, as recommended by Lacković, Olteanu, and Campbell (2024). By documenting their experiences in a tangible, personal format, the reflections aim to serve as a form of collaborative research that is collectively analyzed and published in a format chosen by the community itself.

“Be bold, but not reckless”

One of the pillars of Wikipedia is that there are no rules set in stone, and volunteers are encouraged to “Be bold, but not reckless” when editing on the platform. This same context surrounds this proposal, as it unfolds at a time when there is no complete clarity on how GenAI tools function, while the turbulence of emerging innovations obscures the bigger picture. GenAI arrives like a guest who shows up hours early to a party, forcing us to improvise a welcome.

The rapid pace at which these tools evolve generates a sense of instability and unease, contrasted by the deliberate choice to embrace the hope that GenAI can foster and enrich learning. This contradictory feeling finds resonance in the words of Czerniewicz and Cronin (2023):

“Hope is practical. It means taking action, being disciplined, making plans. Hope is impractical. It means dreaming, being undisciplined, being open-ended.”

References

Ackerman, R. (2024). Wikimedia’s CTO: In the age of AI, human contributors still matter. MIT Technology Review. https://www.technologyreview.com/2024/02/26/1088137/wikimedia-wikipedia-cto-selena-deckelmann-ai-human-contributions/

Brooks, C., Eggert, S., & Peskoff, D. (2024). The Rise of AI-Generated Content in Wikipedia (arXiv:2410.08044). arXiv. http://arxiv.org/abs/2410.08044

Choi, J. H., Monahan, A., & Schwarcz, D. (2023). Lawyering in the Age of Artificial Intelligence (SSRN Scholarly Paper 4626276). Social Science Research Network. https://doi.org/10.2139/ssrn.4626276

Czerniewicz, L., & Cronin, C. (Eds.). (2023). Higher Education for Good: Teaching and Learning Futures (1a ed.). Open Book Publishers. https://doi.org/10.11647/OBP.0363

Evenstein Sigalov, S., Cohen, A., & Nachmias, R. (2024). Transforming higher education: A decade of integrating wikipedia and wikidata for literacy enhancement and social impact. Journal of Computers in Education. https://doi.org/10.1007/s40692-024-00334-x 

Fawns, T. (2022). An entangled pedagogy: Looking beyond the pedagogy technology dichotomy. Postdigital Science and Education, 4, 711–28. https://doi.org/10.1007/s42438-022-00302-7

Fawns, T. & Nieminen, J. (2023). The only way is ethics: A dialogue of assessment and social good. In Higher Education for Good: Teaching and Learning Futures (1a ed.). Open Book Publishers. https://doi.org/10.11647/OBP.0363 

Lacković, N., Olteanu, A., & Campbell, C. (2024). Postdigital Literacies in Everyday Life and Pedagogic Practices. Postdigital Science and Education. https://doi.org/10.1007/s42438-024-00500-5

McDowell, Z. J., & Vetter, M. A. (2022). Wikipedia as Open Educational Practice: Experiential Learning, Critical Information Literacy, and Social Justice. Social Media + Society, 8(1), 20563051221078224. https://doi.org/10.1177/20563051221078224

Moodle (2024, noviembre 8). BigBlueButton: The Future of Virtual Classrooms | MoodleMoot Global 2024. Fred Dixon presentation [Video recording]. https://www.youtube.com/watch?v=0mBmO-ZPcnw

Nieminen, J. H., Haataja, E., & Cobb, P. J. (2024). From active learners to knowledge contributors: Authentic assessment as a catalyst for students’ epistemic agency. Teaching in Higher Education, 1–21. https://doi.org/10.1080/13562517.2024.2332252

Petrucco, C., & Ferranti, C. (2020). Wikipedia as OER: The “Learning with Wikipedia” project. Journal of E-Learning and Knowledge Society, 16(4), 38–45. https://doi.org/10.20368/1971-8829/1135322

Reeves, N., Yin, W., & Simperl, E. (2024). Exploring the Impact of ChatGPT on Wikipedia Engagement (arXiv:2405.10205). arXiv. http://arxiv.org/abs/2405.10205

Remmik, M., Siiman, A., Reinsalu, R., Vija, M., & Org, A. (2024). Using Wikipedia to Develop 21st Century Skills: Perspectives from General Education Students. Education Sciences, 14(1), 101. https://doi.org/10.3390/educsci14010101

Shao, Y., Jiang, Y., Kanell, T. A., Xu, P., Khattab, O., & Lam, M. S. (2024). Assisting in Writing Wikipedia-like Articles From Scratch with Large Language Models (arXiv:2402.14207). arXiv. http://arxiv.org/abs/2402.14207

Sharples, M. (2023). Towards social generative AI for education: Theory, practices and ethics. Learning: Research and Practice, 9(2), 159–167.  https://doi.org/10.1080/23735082.2023.2261131