Daejeon, July 27–31
The Itaewon tragedy of October 29, 2022, revealed a significant tension between state-imposed, government-mandated mourning and civil society’s establishment of a counter-archive. Following the tragedy, the government declared an official mourning period without identifying the accident’s causes, obstructing mourning initiatives led by bereaved families and citizens. (Bahc 2025) Civil society, however, formed the 10.29 Itaewon Disaster Bereaved Families Council and the Itaewon Disaster Citizens' Countermeasures Committee to organize on-site memorial services and documentation efforts. Citizens erected memorial altars at Exit 1 of Itaewon Station and in front of Seoul City Hall, collected and preserved floral tributes and condolence messages (Figure 1), and initiated the Archiving 10.29 Itaewon’s Memory Project. This endeavour transcended mere remembrance, embodying a politics of mourning by documenting the social disaster’s truth and demanding preventive measures. It underscores the importance of citizen-led archival movements as counter-narratives to state-directed memory politics.
This study examines the Archiving 10.29 Itaewon’s Memory Project, a civil society initiative that archives tribute messages from the October 29 Itaewon tragedy and converts them into text data via crowdsourcing. See: Archiving 10.29 Itaewon’s Memory Project, https://campaigns.do/campaigns/1686; and Archiving 10.29 Itaewon’s Memory Project (English messages), https://campaigns.do/campaigns/1703.
Record preservation efforts began in March 2023 and encompass the collection, classification, and safeguarding of memorial messages, site maintenance, and public engagement. Participants included researchers, artists, educators, cultural organizers, and students. Approximately 200 cumulative volunteers negotiated protocols to establish a democratic and flexible records management system. (Kim 2025) Categorizing sticky notes into four types, namely "name-mentioning messages," "survivor messages," "foreign language messages," and "general messages," participants aimed to restore victims’ identities and contextualize the disaster’s multilayered impact. Throughout the archiving process, dialogue and communication among participants took precedence over strict adherence to record integrity, with simple guidelines fostering consensus on exceptions.
The classification process extended beyond physical organization, functioning as a practice of interpretation and remembrance of the writers’ emotions and contexts. During disposal deliberations, participants addressed problematic records, such as hate speech targeted at victims, survivors, families, and more, reflecting on the disaster’s complex implications. This process transcended simple record classification, constituting an ethical practice that collectively shaped social mourning standards.
By 2025, the third anniversary of the disaster, preservation efforts expanded into digitization. Collaborating with digital technologists Parti Co-op, the initiative launched the Archiving 10.29 Itaewon’s Memory online campaign in which data annotations were crowdsourced. Preprocessing on digitized scanned images was streamlined using AI technologies including automatic image segmentation and OCR. A crowdsourcing platform enabled participants to review condolence message images, edit initial machine-recognized texts, and submit final annotations. The combination of AI tools and manual crowdsourcing, instead of full automation, emphasizes ethical engagement over efficiency, demonstrating the possibility of data work as solidarity. Submissions were displayed in a gallery format, emphasizing a collaborative user experience (Figure 2). Workshops, datathons, and digital reward systems further supported the campaign.
As of May 2026, approximately 30,000 of 33,000 messages had been digitized, with over 700 citizens participating, fostering social solidarity. This project innovatively applies data crowdsourcing strategies—traditionally used in AI training, citizen science (Ara et al. 2024; Cartwright et al. 2019; Karanikola 2023; Leonard et al. 2022; Matabos et al. 2025), and cultural metadata (BrodeFrank 2022)—to memorial records. The resulting 1.5 million-character text surpasses prior examples, such as analyses of a 90,000-character memorial corpus (Lee / Choi 2021; Lee / Choi 2024) and 163 songs (Lee 2024) relating to the Sewol Sinking and 1,000 sticky notes from the Gangnam Station Post-it Note protest (The Kyunghyang Shinmun 2016), offering potential for future applications. Digitization of ~24k Korean messages was concluded in December 2025, with the dataset planned for public release in 2026. Foreign language messages remain under ongoing digitization, and will be released as a separate dataset.
Analysis of the messages uncovered affects such as empathy and consolation towards victims and families, remorse and apologies, helplessness, and anger; as well as multilayered discourses that include demands for state accountability. Applying multilingual embedding models and clustering techniques, we identify keyword patterns reflecting social memory and policy demands, such as "rest in peace," "we will not forget," and "enactment of a special law." (Figure 3) The high percentage of non-Korean messages (20-30% of the total) highlights Itaewon’s multicultural identity, illustrating that disaster memory transcends national boundaries. Survivor and rescuer testimonies exposed systemic failures in the disaster response, suggesting that collective memory can foster social solidarity beyond individual trauma. As public interest in the October 29 Itaewon disaster declines and activism narrows to bereaved families and activists, archiving activities provide a renewed avenue for civic engagement and awareness. Ultimately, Archiving 10.29 Itaewon’s Memory Project provides a replicable model for participatory archiving in post-disaster communities.
Figure 1. Example of preserved memorial message. https://campaigns.do/campaigns/1686/memories/221195.
Figure 2. Gallery display of submitted message annotations.
Figure 3. Text embeddings visualization of condolence messages.