2026-06-27 –, Room 403 PC Desk (Seats 30)
As historians increasingly rely on digital tools to document lived experience at scale, artificial intelligence (AI) transcription has emerged as a practical response to the labor-intensive work of oral history. Yet the global turn toward automated transcription raises fundamental questions about how historical knowledge is shaped when spoken testimony is rendered into text by algorithmic systems. This paper examines these issues through a case study of the Korean Memories Project, a crowdsourced oral history initiative documenting the experiences of elderly South Koreans aged eighty and above who lived through Japanese colonial rule, the Korean War, and rapid postwar modernization—processes deeply embedded in global and transnational histories.
Using fifty interviews recorded in diverse community settings, this study compares two AI transcription systems: ClovaNote, developed specifically for Korean-language speech, and TurboScribe, based on OpenAI’s Whisper model. A mixed-methods analysis of representative transcripts evaluates transcription accuracy alongside qualitative patterns that affect historical interpretation. While both systems generate text suitable for indexing and large-scale archival use, they differ in omission patterns and error types. More significantly, both systems consistently normalize dialect, remove repetitions and hesitations, and restructure speech to enhance narrative coherence.
We argue that this process of textual standardization reflects a broader epistemological tension between AI systems optimized for readability and information extraction and oral history methodologies that treat pauses, affect, and speech patterns as historically meaningful. By centering Korean-language testimony from elderly speakers whose linguistic practices often fall outside dominant AI training data, this paper highlights how AI-mediated transcription reshapes what becomes legible in global historical archives. The paper concludes by proposing a hybrid, human-centered framework for AI-assisted transcription applicable to large-scale oral history projects worldwide.
Oral history; Digital history; AI transcription; Memory and testimony; Korean history
Alice Wrigglesworth is an Associate Professor of English at George Mason University Korea. A writing instructor with more than twenty years of international teaching experience, her research focuses on the scholarship of teaching and learning, including peer feedback, intercultural communication in the classroom, and L2 writing pedagogy. She is the principal investigator and project manager of the Korean Memories Project, an interdisciplinary oral history initiative examining AI-assisted transcription and translation.
Mingyo Chu is an undergraduate student at George Mason University Korea majoring in Global Affairs. She is a research assistant on the Korean Memories Project, where she contributes to Korean-language transcription and transcription analysis for oral history interviews. Her academic interests include human rights, inequalities, media, cultural, and peace studies.
Lynnette G. Leonard is an Associate Professor at George Mason University Korea. She has experience with interview-based journalism and oral history projects. She led the creation of the North Omaha Media Alliance (a student-based service-learning journalism project) and advised several student projects in Bulgaria. Her research explores the role of media and new technology in human communication as well as rhetorical critique of historical primary source documents.
H. Deborah Kwak is an Assistant Professor at George Mason University Korea. has extensive experience conducting research in cross-cultural settings. She has published on peacebuilding, conflict transformation, emotions and social change (Kwak 2019; Leguro and Kwak 2016; Kwak and Rais 2016; Summers-Effler and Kwak 2015). Her scholarship has utilized empirical materials from multiple data sources using a feminist theoretical lens.