An STS analysis of a digital humanities collaboration: trading zones, boundary objects, and interactional expertise in the DECRYPT project

The paper examines collaborations in digital humanities (DH), a field characterized by large datasets, tool development, and reflections on how digital methods reshape humanities research. Despite the prevalence of the “lone scholar” myth in the humanities, DH necessitates teamwork across disciplinary boundaries, requiring new vocabularies, practices, and methods. The authors adopt Science and Technology Studies (STS) as an analytical framework to understand such cross-disciplinary collaboration. They review concepts including trading zones, boundary objects, interactional expertise, referred expertise, and the ambassadorial model to explain how distinct epistemic cultures negotiate shared goals and communication. The research focuses on the DECRYPT project, a multi-party DH collaboration involving historians, linguists, computational linguists, cryptologists, computer scientists, and computer vision experts, among others. The central research question is: Do STS methodologies used to analyze cross-disciplinary collaborations work similarly in multi-party, multi-disciplinary contexts as in two-party collaborations? The paper situates DECRYPT within the broader history and practice of DH collaboration and outlines how the project’s structure and practices provide a testbed for applying STS concepts.

The authors synthesize key STS literature to frame cross-disciplinary collaboration. Galison’s trading zones describe local coordination between communities with partially incommensurable languages. Collins, Evans, and colleagues elaborate a trading zone typology distinguishing homogeneous/heterogeneous and collaborative/coercive cases; they emphasize collaborative heterogeneous zones facilitated by interactional expertise and boundary objects. Interactional expertise is the ability to speak the language of a field without contributory practice, contrasted with contributory expertise; related ideas include referred expertise (skills transferred indirectly from one domain to another) and the ambassadorial model (experts learning another community’s trade to broker understanding). Star and Griesemer’s boundary objects are plastic enough to satisfy local needs yet stable across sites, enabling cooperation without full consensus. Kemman extends the trading zone model for DH, adding a connected/disconnected diagonal, and argues that well-functioning DH collaborations are symmetric-heterogeneous (fractionated) trading zones where boundary objects and brokers with interactional expertise enable coordination. This literature provides the analytical toolkit for studying DECRYPT, which spans more than two epistemic cultures.

A mixed-methods approach combined: (1) participant observation by the two authors, both active DECRYPT members, offering close access but with potential bias; (2) documentary analysis of systematically recorded meeting notes (2–3 day general meetings held 3–4 times/year, including during COVID-19 via online formats), presentations, and wiki content focused on communication and organizational practices; (3) a structured LimeSurvey questionnaire completed by all 14 core members (not including short-term MA students or external users). The survey contained 30 open- and closed-ended questions across five categories: participation, objectives, disciplinary backgrounds, collaboration, and boundary objects. It was discussed in a meeting for coverage, refined, and followed by unstructured interviews and focus group–like discussions with the full team, including the PI. The closed questions informed figures (e.g., joining timelines, subgroup participation, publication preferences), while open-ended responses were analyzed qualitatively. The authors acknowledge limitations: small group size, potential identification of respondents despite anonymity, participant-observer bias, and the deliberate exclusion of analyses of personal power structures. The team did not conduct imitation game tests to validate interactional expertise; claims about such expertise are therefore based on self-assessment and observed participation.

- Project scope and outputs: DECRYPT (2018–2024; Swedish Research Council; 29.5 million SEK ≈ €3M) aims to establish historical cryptology by building infrastructure to collect, digitize, process, and decrypt historical encrypted sources and release data/tools via a web service. The team has collected over 7000 encrypted sources from 13 European countries, released a publicly available DECODE database, compiled historical corpora for 17 European languages (HistCorp) with language models and normalization tools, developed interactive transcription (handwritten text recognition) models, and built/semi-automated decryption tools (e.g., CrypTool2). Over 60 scientific publications were produced in the first four years across cryptology, computational linguistics, image processing, history, and linguistics. - Collaboration structure: Cross-disciplinary subgroups focus on data collection, historical language corpus, CrypTool2, transcription tool, evolution of keys, and historical language models’ impact on decryption. Regular general meetings and subgroup meetings, shared wiki, and email lists support coordination. Onboarding includes mentorship and presentation opportunities for juniors. - Roles and expertise: Survey of 14 core members showed diverse contributory roles (source collection, metadata, transcription, historical analysis, decryption, tool development, image processing models, database building, corpus building, transcription tool). Many participants reported additional fields where they could follow discussions and contribute to debates, indicating development toward interactional expertise. The PI, with a computational linguistics background spanning humanities and computer science, functions as a broker, exhibiting interactional and referred expertise and, at times, ambassadorial roles into areas like image processing. - Epistemic cultures and publication patterns: Participants reported differing publication norms (e.g., historians value single-authored monographs; natural sciences emphasize multi-authored journal/conference papers). Some tension was noted between academic aims (novel research, papers) and creating robust, user-friendly tools. Terminology required negotiation (e.g., metadata; plaintext vs. cleartext; nomenclator; clustering; periodization labels), and methodological learning occurred across domains (e.g., cryptanalysis heuristics, quantitative methods, archival practices). - Trading zones: Many perceived the collaboration as symmetric, with all disciplines’ expertise needed; others described asymmetry where humanities problem-setting guides toolmakers, consistent with a fractionated trading zone that can shift across the typology. Overall, DECRYPT operates largely in the collaborative heterogeneous (fractionated) region, with episodes reflecting asymmetric commissioning. - Boundary objects identified: Grant application; cipher keys; encrypted messages; unsolved ciphers; DECODE database; transcription guidelines; historical language corpora; TranscripTool; CrypTool2. These artifacts serve different purposes across communities (e.g., sources to transcribe vs. testbeds for algorithms; instructional guidelines vs. implementation specifications), yet maintain shared identity and enable coordination. - Online vs. in-person: In-person meetings were deemed crucial for socialization, trust, and deep brainstorming; online meetings were efficient and inclusive but lacked informal exchanges and sustained focus. This mirrors wider findings that physical gatherings are key to community building. - Risks/challenges noted: Achieving high-quality automatic transcription of historical cipher manuscripts remains challenging; balancing research publications with production-grade tools; integrating disciplines with distinct workflows and expectations.

Findings support the applicability of STS concepts to a multi-party DH collaboration. DECRYPT largely exemplifies a collaborative heterogeneous (fractionated) trading zone, enabled by boundary objects and by participants’ growth of interactional expertise—especially the PI’s broker role incorporating referred and ambassadorial elements. Negotiations around terminology, methods, workflows, and publication strategies demonstrate the practical mechanisms by which epistemic cultures coordinate without merging. The project’s shared artifacts (databases, corpora, tools, cipher materials, and documentation) stabilize collaboration while allowing discipline-specific interpretations and uses. Perceptions of both symmetry and asymmetry reveal dynamic movement within the trading zone typology: humanities often set research agendas (problem-setting), while technologists develop methods and tools—yet mutual dependence and feedback were evident. In-person meetings proved essential to building the socialization needed for interactional expertise development, aligning with broader STS findings on conference-based socialization. Overall, the STS framework helps explain how DECRYPT navigates cross-disciplinary tensions to produce both scientific insights and infrastructure.

The study demonstrates that STS concepts—trading zones, boundary objects, interactional and referred expertise, and ambassadorial roles—provide a robust framework for analyzing and guiding a complex, multi-disciplinary DH collaboration. In DECRYPT, diverse epistemic cultures coordinate around shared artifacts and through brokers with interactional expertise, enabling progress on both research and infrastructural goals. The project has delivered substantial datasets, tools, and publications while revealing challenges in terminology harmonization, publication strategies, and tooling maturity. The authors offer practical insights and best practices useful to other DH projects, emphasizing structured meetings, transparent documentation, subgroup organization, mentorship, and deliberate cultivation of shared terminology and boundary objects. Future directions include tighter integration of transcription and decryption into a unified, user-friendly web service and further strengthening ties between image processing and cryptanalysis. The work encourages meta-level exchanges among DH projects to share strategies and address common pitfalls, enhancing the success of future cross-disciplinary initiatives.

- Participant-observer bias: Both authors are project members; colleagues knew they were being analyzed, potentially influencing responses. - Small sample and identifiability: Only 14 core members responded; despite anonymity, disciplinary backgrounds could render responses identifiable. - Scope of respondents: Excludes short-term MA students and external users; findings reflect core team perspectives. - Power relations: Personal power dynamics among individuals were out of scope; alternative methods would be required to analyze them. - Interactional expertise validation: No imitation game or equivalent tests were conducted; claims rely on self-assessment and observation. - Qualitative emphasis: Closed-ended items informed figures, but primary analysis was qualitative; generalizability may be limited.