The AI-Augmented Research Process: A Historian's Perspective

TL;DR

This paper explores how large language models can be integrated into historical research workflows, emphasizing responsible use, quality control, and a structured approach based on a case study of Shanghai merchants.

Contribution

It introduces a detailed, reproducible workflow for AI-augmented historical research, combining computational tools and interpretive analysis, with a focus on transparency and scholarly standards.

Findings

LLMs support historians in processing large text corpora

Rigorous quality control is essential for responsible AI use

A structured workflow enhances reproducibility in AI-augmented history research

Abstract

This paper presents a detailed case study of how artificial intelligence, especially large language models, can be integrated into historical research workflows. The workflow is divided into nine steps, covering the full research cycle from question formulation to dissemination and reproducibility, and includes two framing phases that address setup and documentation. Each research step is mapped across three operational domains: 1. LLM, referring to tasks delegated to language models; 2. Mind, referring to conceptual and interpretive contributions by the historian; and 3. Computational, referring to conventional programming-based methods like Python, R, Cytoscape, etc. The study emphasizes that LLMs are not replacements for domain expertise but can support and expand capacity of historians to process, verify, and interpret large corpora of texts. At the same time, it highlights the necessity of rigorous quality control, cross-checking outputs, and maintaining scholarly standards. Drawing from an in-depth study of three Shanghai merchants, the paper also proposes a structured workflow based on a real case study hat articulates the cognitive labor of the historian with both computational tools and generative AI. This paper makes both a methodological and epistemological contribution by showing how AI can be responsibly incorporated into historical research through transparent and reproducible workflows. It is intended as a practical guide and critical reflection for historians facing the increasingly complex landscape of AI-enhanced scholarship.