Renewing intellectual history in the age of AI

In 2026, Helsinki Computational History Group is entering a new phase. Over the previous decade, COMHIS helped establish bibliographic data science, integrated workflows for historical newspapers, large-scale text-reuse research and computational studies of publishing and public discourse. That accumulated interdisciplinary work now supports a closer connection between intellectual-historical questions, AI development and production computing.

The change is practical as well as conceptual. New language and vision models can be tested against historical collections on CSC’s computing infrastructure, and results can arrive quickly enough to alter the live research design. A failed image crop, a misleading semantic match or a book-specific layout convention is no longer merely cleaned away at the end: it can become the reason to revise the data, model, validation rule or historical question.

Critical editions enter the workflow

Critical editions—one of the humanities’ core forms of cumulative scholarship—are central to this development. The work on Hume’s History of England shows how close editorial attention to sources, witnesses, variants and typography can become the basis for reusable tools. These tools allow editorial questions to scale across collections without surrendering the provenance, uncertainty and scholarly judgement that give a critical edition its authority.

Teaching an annotator to reason about books

One experiment now under development is an annotator agent trained through eighteenth-century book cases reviewed by eighteenth-century historians. It is learning to reason from page sequence, typography, ornaments, catchwords and other visible evidence; to distinguish related page roles; and to keep uncertainty explicit. Its decisions are frozen before comparison with human annotations, and accepted lessons enter a versioned casebook. The purpose is to make expert historical reasoning cumulative and auditable at scale, not to replace independent human judgement.

Following ideas beyond quotation

Recent articles extend the study of reception beyond direct quotation. identifies full, partial, composite and adapted translations across languages, while examines whether semantic retrieval can recover less explicit engagements with thinkers such as Locke. Combined with COMHIS work on text reuse, these approaches allow researchers to compare different ways in which an idea may be repeated, translated, transformed or indirectly taken up.

A wider programme

The group’s new projects give these developments a wider intellectual frame. IMPRINT examines early modern intellectual networks and texts; ReBeL investigates implicit meaning and the unsaid; CASCADE studies semantic change across environments; MECANO investigates canon formation; and FIN-CLARIAH develops sustainable research infrastructure.

Adaptive Evidence Construction is one method for organising this work. Humanities researchers design, govern and implement the process through which sources, models, computing and validation become historical evidence inside an integrated interdisciplinary environment. This is part of the wider COMHIS goal: to renew intellectual history while developing new ways for the humanities to work with large-scale data and AI.

See and .