A Critical Genealogy of Human Agency in Machine Translation

Description This repository contains the qualitative coding framework used in the analysis of human agency in machine translation (MT) across several historical periods. The data corresponds to a study that traces the historical evolution of discourses surrounding the translator’s role, from the first manifestations in mid-twentieth century to contemporary neural MT and AI-based systems. The study argues that the current tension between empowerment and control is not a recent phenomenon but the consequence of an early technocratic framing that initially established a paradigm in which human participation was regarded as a limitation to be overcome rather than a value to be cultivated. Research’s objective The goal is to identify recurring themes and metaphors that structure the conceptualization of the human agent in MT along time, paying attention to how notions of agency, responsibility, and expertise are framed. Method Methodologically, the research applies thematic analysis (Braun & Clarke, 2006) to a corpus of key documents that mark pivotal stages in the evolution of MT (see the list of primary source texts below). Dataset The dataset includes all materials relevant for the study: Filename Description 02_coding protocol.pdf Information on the analytical approach, the units of analysis, the coding structure and procedure, scope and limitations 03_codebook period 1.txt 03_codebook period 2.txt 03_codebook period 3.txt 03_codebook period 4.txt Complete codebooks for each period. Each coded entry includes the following information: period, theme, code, definition, inclusion criteria, exclusion criteria, example reference, analytical memo. 04_code to theme mapping Code-to-theme mappings Due to copyright restrictions, primary source texts are not included in this dataset (the list is provided below). Primary source texts To trace historical transformations in MT discourse, the corpus was organized into four genealogical phases corresponding to major technological and conceptual shifts, according to the table below. These phases are treated not as rigid technological periods but as discursive regimes in which particular configurations of human–machine relations become dominant. Period MT development phase Texts analyzed 1951-1966 Foundational period: early rule-based systems - ALPAC (1966) - Bar-Hillel (1951, 1960) - Reifler (1952) 1967 -1980s Post-ALPAC report - Kay (1980) 1990s - 2010s Statistical MT Era - European Commission (2008) - Foti (2012) - Hutchins (2001) 2016 - Present Neural MT and LLMs - CSA (2024) Läubli et al. (2018 - Wu et al. (2016) Selection criteria prioritized texts that explicitly articulate assumptions about the role of humans in MT, and reflect distinct socio-technical configurations of human–machine relations. Main findings The analysis identifies four dominant thematic configurations: 1. Automation and efficiency, where the human is cast as an error-prone or redundant element. 2. Evaluation and supervision, where the translator’s role is redefined as that of post-editor tasked with aligning human judgment to machine output. 3. Data extraction and invisibility, where human contributions are subsumed as training material for opaque systems. 4. Empowerment and responsibility, where users are rhetorically centered, while their experience is managed through interfaces that promise personalization but centralize decision-making. When examined diachronically, these themes reveal a persistent movement from the elimination of human intervention to its instrumentalization within increasingly centralized and profit-driven infrastructures. References ALPAC Automatic Language Processing Advisory Committee. 1966. Languages and machines: Computers in translation and linguistics. National Academy of Sciences, National Research Council. https://aclanthology.org/www.mt-archive.info/50/ALPAC-1966.pdf Bar-Hillel, Yehoshua. 1951. “The Present State of Research on Mechanical Translation.” American Documentation 2, no. 4: 229–237 Bar-Hillel, Yehoshua. 1960. “A Demonstration of the Nonfeasibility of Fully Automatic High Quality Translation.” In The Present Status of Automatic Translation of Languages, appendix III. Advances in Computers 1: 158–163. https://mt-archive.net/Bar-Hillel-1960-App3.pdf CSA Research. 2024. “The Global Enterprise Content Production Line—Our Analysts’ Insights.” Accessed 15 Decembre 2025. https://csa-research.com/Blogs-Events/Blog/The-Global-Enterprise-Content-Production-Line European Commission: Directorate-General for Translation. 2008. Translation tools and workflow. Foti, Marcus. 2012. “MT@EC: Working with translators.” Proceedings of Translating and the Computer 34. TC 2012. https://aclanthology.org/2012.tc-1.4/ Kay, Martin. 1980. “The Proper Place of Men and Machines in Language Translation.” Research Report CSL-80-11. Palo Alto, CA: Xerox Palo Alto Research Center. Reimpressed in Readings in Machine Translation, Sergei Nirenburg, Harold L. Somers and Yorick A. Wilks (eds.), 221–232. Cambridge, MA: MIT Press. https://doi.org/10.7551/mitpress/5779.003.0022 Läubli, Samuel, Rico Sennrich, and Martin Volk. 2018. “Has Machine Translation Achieved Human Parity? A Case for Document-level Evaluation.” Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, 4791-4796. https://doi.org/10.18653/v1/D18-1512 Reifler, Erwin. 1952. “Mechanical translation with a pre-editor, and writing for MT.” In Proceedings of the Conference on Mechanical Translation, Massachusetts Institute of Technology. https://aclanthology.org/1952.earlymt-1.10/ Wu, Y., Schuster, M., Chen, Z., Le, Q. V., Norouzi, M., Macherey, W., Krikun, M., Cao, Y., Gao, Q., Macherey, K., Klingner, J., Shah, A., Johnson, M., Liu, X., Kaiser, Ł., Gouws, S., Kato, Y., Kudo, T., Kazawa, H., … Dean, J. (2016). Google’s Neural Machine Translation System: Bridging the Gap between Human and Machine Translation (arXiv:1609.08144). arXiv. https://doi.org/10.48550/arXiv.1609.08144