AI-assisted bibliographic revision
This project aims to test an AI-assisted copy preparation system. The proposed experiment focuses on a specific aspect: the revision of bibliographies, whose structuring and formatting are prone to numerous errors despite the existence of tools like Zotero. The goal is to assess the needs of editors, test different tools and methodologies, and document the experience to ensure its reproducibility.
This project is envisioned as the first step in a long-term experiment designed to evaluate the potential of AI in editorial preparation work, while proposing technical and methodological solutions that respect the work of authors and editors. The ecological impact of the proposed solutions will also need to be considered.
Issues
While any editor can appreciate the automation of certain tasks, the introduction of AI into the revision and copy preparation process must be carefully considered to assess its impact on editorial work. Mastery of language and, more broadly, knowledge and control of various editorial standards (orthographic, bibliographic) emerge as key competencies of the editor-reviewer, establishing themselves as a marker of the profession. The experience gained by editors, who often work within the same journal or group of journals for many years, represents significant added value. This, along with the quality of the author-editor relationship, could be disrupted by the widespread automation of editorial tasks.
On a larger scale, the epistemological impact of automating proofreading and revision must also be examined. At a time when large industrial companies (like Microsoft) are already deploying their own AI systems to automate certain tasks (e.g., automatic figure transcription), it is important to determine how to adapt our specific requirements—journal recommendations, language, disciplinary field, etc.—to an open model.
Therefore, we aim to explore the possibility of "reusability." Given that references may recur from one article to another—especially within journals in specific fields—the development and utilization of a thematically organized and correctly formatted bibliographic database could save time and improve quality.
Technical challenges
- Deploy an AI (to be determined)
- Write and test a series of prompts
Research activities
- Interviews with the editor of the HN journal: What difficulties are encountered, and what common errors occur?
- Overview of the various existing tools and selection of tools to be tested
- Implementation of the tools
- Documented experimentation on two articles from the HN journal
Deliverables
Interview framework and report with the editor
Documentation of the experiment
Editorial detailing the experience, to be published in issue 11 of the journal Humanités numériques (June 2025)
Development of a larger-scale project
