Daejeon, July 27–31
This paper is based on a research project of a dissertation and explores how text mining methods can reveal processes of intellectual "translation" between different thought collectives within a single disciplinary field. Drawing on Ludwik Fleck's concepts of Denkstile (thought styles) and Denkkollektive (thought collectives), The project examines how distinct theoretical orientations in German educational science (Erziehungswissenschaft) developed characteristic linguistic patterns from 1945 to 2020. Using a corpus of eight German-language educational science journals spanning different publication periods, this paper demonstrates how computational methods can identify linguistic boundaries between scholarly communities and trace moments when these communities engage in acts of conceptual translation.
Educational science in German-speaking countries has long been characterized by its theoretical pluralism and lack of unified disciplinary identity (Heid 1987; Horn 2008). Rather than viewing this fragmentation as merely organizational, this paper approaches it as a linguistic-epistemic phenomenon. Following Fleck (1980), scientific disciplines are understood as comprising multiple thought collectives, each characterized by distinct thought styles that manifest in language (Fix 2021). As Terhart (1992: 202) notes, educational science encompasses "very high variability and plurality of approaches or 'paradigms' – and thus also: languages." (translation by the author)
These linguistic differences are not merely stylistic. Lenzen and Rost (1998; translations by the author) observe that while Geisteswissenschaftliche Pädagogik (hermeneutic-humanistic pedagogy) maintained a "national-linguistic orientation of terminology" (p. 1316), more empirically-oriented approaches adopted international terminology, and Kritische Erziehungswissenschaft (critical educational science) gave "imported concepts like 'emancipation' a specifically educational-scientific meaning" (p. 1317). These distinct linguistic practices represent different epistemic cultures that must "translate" concepts when engaging with each other.
Established computational methods as text mining combined with recent developments in AI-driven translation technologies raise new questions for disciplinary historiography. While machine translation facilitates cross-linguistic exchange, the intra-disciplinary translation between thought collectives represents a fundamentally different challenge – one requiring not just lexical transfer but conceptual negotiation between incommensurable epistemic frameworks.
This study asks: How do different thought collectives within educational science develop distinctive linguistic patterns and how can they be characterized? What linguistic markers indicate moments of inter-collective "translation"? How can computational methods help identify these boundaries and translation acts?
This paper employs text mining methods (Biemann et al. 2022; Blaxill 2020; Guldi 2023) in a "scalable reading" approach (Weitin et al. 2023), combining distant and close reading perspectives. The corpus consists of full texts from eight educational science journals published between 1945 and 2020, encompassing publications from both the Federal Republic of Germany (BRD) and the German Democratic Republic (GDR). This temporal scope captures major theoretical transformations in the discipline. The corpus comprises approximately 105 million tokens (excluding punctuation) across eight journals. Journal selection was guided by previous investigations in German-language science studies and the practical criterion of digitized full-text availability.
The analytical approach combines:
All analyses are implemented using R, drawing on established packages for text mining and natural language processing (R Core Team 2021).
This research contributes to digital humanities in three ways. First, it demonstrates how computational methods can operationalize theoretical concepts from science studies – specifically Fleck's thought collectives – making them tractable for empirical investigation. Second, it extends "distant reading" from literary studies to disciplinary historiography, showing how macro-analytical methods reveal structures invisible at the micro-level. Third, it offers a model for studying intra-disciplinary translation that complements current interest in cross-linguistic translation.
This approach also raises methodological questions relevant to AI-driven translation. Current machine translation systems optimize for lexical equivalence, but disciplinary translation requires understanding conceptual incommensurability. By mapping how human scholars negotiate these differences historically, we may inform more sophisticated approaches to computational translation in specialized domains.
This work challenges narratives of disciplinary coherence by empirically documenting linguistic fragmentation. Rather than treating theoretical pluralism as a problem to be solved, this paper reveals it as a productive space of ongoing translation. The computational methods employed are scalable to other disciplines and national contexts, offering a framework for comparative disciplinary historiography.
Future work will extend the analysis to include non-journal publications (textbooks, dissertations) and explore machine learning approaches to automatically identifying translation moments. A comparative analysis with educational science literature in other languages, examining how linguistic boundaries intersect with national-linguistic traditions, is also conceivable.
By treating disciplinary history as a history of linguistic practices and translation acts, this research opens new perspectives on how scientific knowledge circulates within fragmented fields. Text mining methods make visible the boundaries and bridges between thought collectives, revealing the ongoing work of conceptual translation that enables interdisciplinary communication. In an era of increasingly sophisticated AI translation, understanding these human processes of epistemic negotiation remains crucial.