AI and Contemporary Historiography: From Instrumental Use to Methodological Transformation (Preprint v2, November 2025)

Status: Preprint (non-peer-reviewed working paper).

This preprint explores how artificial intelligence (AI) and machine learning (ML) are reshaping the methodology of contemporary historiography. It distinguishes three levels of AI integration into historical research: (1) technical reconstruction of sources, (2) factual extraction and verification through NLP and machine learning, and (3) interpretative modeling and hypothesis generation via probabilistic reasoning.

By referencing both classic and current works—from Alan Turing and Marvin Minsky to recent neural-network applications such as the Ithaca project—the paper situates AI within a longer history of epistemological transformation. It argues that AI-assisted historiography represents a gradual transition from narrative hermeneutics to data-driven empiricism, analogous to the earlier methodological shifts seen in archaeology and cliometrics.

A brief empirical illustration (Werner & Wilczyński 2024) demonstrates the application of classification algorithms to prosopographic data from the PRL Security Service archives. The paper also discusses ethical risks, including the misuse of generative models, and references Marnie Hughes-Warrington’s Artificial Historians (2025) for context.

Disclaimer:
– This version (v1, November 2025) is a pre-submission working draft shared for scholarly feedback.
– It has not undergone peer review and should not be cited as a published article.
– A revised version will be submitted to a peer-reviewed journal