Rencontre SHS Computationnelles

The Rencontres SHS computationnelles, held at Campus Condorcet in September 2025, marked an important step in consolidating institutional attention to computational approaches across the humanities and social sciences in France. The program offered a balanced mix of conceptual framing, infrastructural presentations, and institutional dialogue. Yet, when viewed from the perspective of ongoing work in computational history, significant gaps became visible. These gaps not only delimit the scope of the meeting but also suggest avenues for future enrichment of the computational humanities agenda.

The workshop rightly emphasized the necessity of moving “beyond digitization,” situating computation as a methodological leap rather than a continuation of digitization programs. The roundtable on this theme (“Au-delà de la numérisation”) gave a platform to a number of concrete initiatives: the SocFace project at INED (digitized French census records, 1836–1936, with data extraction and linking), Lea Saint-Raymond’s work on art sales catalogues, and Éric Mermé’s GIS platform (EHESS) for historical georeferencing (GalliGeo). These examples illustrate the potential of computational workflows, but they remained isolated showcases rather than being embedded in a systematic discussion of methods. What was absent were demonstrations of how techniques such as topic modeling, network analysis, sequence mining, or named-entity recognition can be integrated into research design, transforming not only the sources we use but also the questions we ask. Without this methodological engagement, computation risks being framed as a horizon of potential rather than a body of demonstrable practices.

The program foregrounded philology, art history, linguistics, and demography, fields with strong digital traditions. Social sciences were less visible, including history, sociology, law, anthropology, etc.. This imbalance was even noted in Julianne Nyhan’s keynote, where she observed that many projects remain heritage-driven, raising the question: what historical questions can actually be asked?

One of the most striking omissions in the program was the question of training. Computational humanities will not spread by infrastructures alone; they depend on cultivating a new generation of scholars who can combine disciplinary expertise with computational competence. The presentation pointed to promising but still fragmented initiatives: the new ENS master’s program in “IA et Société” or the well-establish master’s program in Digital Humanities at Ecole des Chartes, and the discussion of recruiting double-trained scholars with backgrounds in both computer science and the humanities. These examples reveal awareness of the problem but also highlight its limits. Without structured curricula and broader pedagogical strategies, computational SSH risks remaining confined to isolated groups rather than becoming a shared competence.

The ethical and epistemological dimensions of computational research remain underexplored. Large-scale digitization and algorithmic processing are never neutral: they embed biases, reproduce power asymmetries, and pose questions of governance. In the PSL CultureLab presentation, I raised the issue of the “heart of the documentary project”: was the initiative to be driven by sources or abstract models? This is an epistemological fault line. Elsewhere, references to the Observatoire des humanités numériques hinted at first steps towards observing digital practices. Yet these remained peripheral. A deeper engagement with epistemic stakes—bias in corpora, opacity of AI tools, politics of access—was largely absent from the workshop agenda.

Sustainability was candidly addressed in the keynote: Julianne Nyhan acknowledged the fragility of many digital humanities projects, which are often project-based and underfunded. But beyond recognition, few concrete solutions were proposed. For historians constructing large-scale corpora such as the Modern China Gazetteer Database (MCGD) or the Modern China Text Base (MCTB), sustainability is central. Software obsolescence, version control, and reproducibility are not marginal concerns but conditions of scholarly credibility. The absence of structured discussion on reproducibility standards or long-term funding models highlights a gap in the current agenda.

The program highlighted infrastructures and collections, but left little room for the discoveries they enable. Younger researchers presented innovative experiments like Creepypastas (ongoing Ph.D. dissertation) as canon formation, automatic metaphor extraction from dictionaries, clustering of recipe datasets. These showed creativity, but they remained isolated case studies, not linked to broader historiographical debates. For historians, the legitimacy of computational methods rests on their epistemic payoff: new insights into social mobility, elite networks, or institutional transformation. Without highlighting such outcomes, computational humanities risk appearing as an infrastructural project without scholarly returns.

The Rencontres SHS computationnelles successfully foregrounded the institutional and infrastructural dimensions of computational humanities. Yet they also revealed the gaps that need to be filled in: insufficient engagement with methods, limited disciplinary balance, lack of structured training, weak attention to epistemology, and little discussion of reproducibility or global integration. Yet these gaps are not blind spots but live tensions within the field: questions were raised about heritage vs. research-driven projects, about sustainability, about the need for double training, about the heart of documentary sources. The challenge now is to turn these questions into a collective agenda. Bridging these gaps will be essential if computational humanities are to move from promise to practice, and from infrastructure to discovery.

---

---