Scaling Microhistory: AI, Data Visualization, and the Structural Analysis of 19th-Century Choson Diaries

Hyoji Ha, Ajou University
Donggue Lee, Ajou University

Microhistory emphasizes that the detailed daily records of ordinary individuals can illuminate the macro-level social structures. Nineteenth-century Choson diaries—such as Unsong Ilgi and Chongsa Ilgi—are microhistorical treasure troves, capturing the daily routines, material exchanges, and intimate interactions across various social strata. However, while traditional qualitative reading excels at uncovering isolated micro-narratives, it faces limitations in connecting vast, unstructured corpora to reveal macro-level social networks. To overcome this limitation, this study proposes a scalable methodology utilizing artificial intelligence (AI) and data visualization. First, we employ Large Language Models (LLMs) combined with an expert validation system to extract micro-level historical elements—actors, locations, and actions—from unstructured diary texts. This process transforms historical records into a structured, computable knowledge base that AI can fully comprehend. Crucially, the continuous expert validation ensures historiographical accuracy and refines the AI's inference capabilities. This human-AI collaborative framework will significantly contribute to easily deciphering and translating vast amounts of previously untranslated historical diaries in the future. Second, we deploy interactive visual analytics to bridge the micro and macro scales by integrating data from multiple diaries. Through multidimensional network and spatiotemporal visualizations, historians can trace the dynamic interactions between elites and non-elites across different sources, comprehensively mapping the seasonal and geographic patterns of local life. By enabling Exploratory Data Analysis (EDA) on historical texts, this interdisciplinary pipeline allows researchers to zoom in on individual lived experiences while simultaneously observing systemic social patterns. Ultimately, this AI-based approach contributes to grasping broader socio-historical structures in a semantic and intuitive manner by reconstructing and linking micro-level patterns from everyday historical records.

No extended abstract or paper available

 Presented in Session 144. Beyond the Census II