Welcome to PhilSoc

The Philological Society (PhilSoc) was established in 1830 and  is the oldest learned society in Great Britain devoted to the scholarly study of language and languages. As well as encouraging all aspects of the study of language, PhilSoc has a particular interest in historical and comparative linguistics, and in the structure, development, and varieties of Modern English.

Next Meeting

Oct
17
2025

October 2025

Agency in translation: some ideological battles and their implications for (machine-)translation
Erich Steiner (Universität des Saarlandes)

The lecture will be given at University College London.

Please note that all ordinary meetings commence at 4:15pm. Members are welcome to come for tea at 3:45 pm.

Abstract

Textual agency or voice has been variously assigned to authors, readers, editors, and reviewers in literary studies (Iser1978, Eagleton 1996, Frow 2022), with linguistics adding a focus on the text itself (Fairclough 1992, Matthiessen 2001, Halliday 2002, Steiner 2020). Both disciplines would also consider the situational and cultural context to various extents. Translation theorists (Venuti 2000, Munday 2016, Steiner 2021) have specifically been asking how translators’ agency and choice are influenced, constrained and possibly predicted by these complex variables. How does the situation change with the impact of machine translation (Bernardini et al 2020, Mikros and Boumparis 2024)?

Some debates under the influence of identity politics have insisted on a “match” between the  translator’s social or ethnic background and that of the source-text author. One such debate arose around translations of Amanda Gorman’s (2021) “The Hill We Climb” presidential address of 2020 into Catalan, Dutch and German. I shall argue on the basis of a brief translation-oriented text analysis of the Gorman source text with a view of its translation into German that the key requirements of such a translation are of a contrastive linguistic, translational and intercultural nature. There is no strictly shared social and ethnic background between source text author and translator in most cases, and if there were, it would not be a necessary, nor a sufficient condition for a good translation.

The intrusion of current machine translation technologies into translation activity presents us with a different scenario. As examples, we shall show Google, DeepL and ChatGPT machine translations of the Gorman text into German. Their quality is at this point in time not yet good enough as a serious competitor for human translation (cf. Hadley and Resende 2024), but it is significantly better than it was in former rule-based MT (e.g. Allegranza, Krauwer and Steiner 1991). (How) has agency changed in the process?

The primary agent i.e. the speaking voice remains the source text author. The “agent” of the translation process, however, is the MT-system, i.e. a tool managed by the natural language processing (NLP) engineer (van Genabith 2020). The decisive knowledge embedded in this system is engineering kowledge rather than knowledge about language or translation. Crucial properties of the target text such as adequacy, fluency, (possible) bias are consequences of the training materials used in the language models and in the MT-system, and these properties are outside of the control of any translator. In the case of translating with ChatGPT and current generative systems, part of control lies in the prompting mechanisms allowed. Such prompting ideally allows some degree of translational control arising out of pre-translational text analysis, yet automation of this relatively costly process will no doubt be attempted. Likewise, in pre-/post-editing and evaluation, language-based types of competence are involved, but even these are undergoing further automation.

The general tendency is something we currently observe in wider areas involving competencies regarded as typically human, such as doing research (Messeri and Crockett 2024), learning languages, and using language creatively. Our worries should start, when our creative abilities are being substituted and reduced by the technology, rather than being enhanced. And once the original author’s voice is made dispensable (Porter and Machery 2024), which is something ChatGPT and Agentive AI are moving towards, we are left with a rather lonely and dystopic reader.

 

References

Allegranza Valerio. & Krauwer. Stephen. & Steiner, Erich eds. 1991. Special Issue of MACHINE TRANSLATION on EUROTRA. Vol. 6 No. 2 and 3. Dordrecht : Reidel

Bernardini, Silvia,  Bouillon, P.,  Ciobanu, D.  Genabith, J. Hansen-Schirra, S.,  O’Brien, Sh.  Steiner, E.,  Teich. E. 2020. Language service provision in the 21st century: challenges, opportunities and educational perspectives for translation studies. In: Noorda, Sijbold, Scott Peter and Vocasovic, Martina Proceedings of Bologna Process Beyond 2020. Fundamental Values of the EHEA. Bologna University Press: pp. 297-303

Eagleton, Terry. 1996. Literary Theory. Second edition. London: Blackwell

Fairclough, Norman. 1992. Discourse and Social Change. Cambridge: Polity Press

Frow, John (ed.) 2022. The Oxford Encyclopedia of Literary Theory. Oxford: Oxford University Press

Gorman, Amanda 2021. The Hill We Climb - an Inaugural Poem for the Country. Viking; Penguin Random House. German bilingual edition and translation and commentary by Uda Strätling, Hadija Haruna-Oelker and Kübra Gümüsay. 2021

Halliday, M.A.K. (2002). Linguistic studies of text and discourse. Edited by. Jonathan Webster. London and New York: Continuum

Hadley, James and Resende, Natalia. 2024. “The Translator’s Canvas: Using LLMs to Enhance Poetry Translation”. Proceedings of the 16th Conference of the Association for Machine Translation in the Americas. Vol.1. pp. 178-189.

Iser, Wolfgang 1978. The act of reading: a theory of aesthetic response. London: Routledge

Matthiessen, Christian M.I.M. (2001). "The environments of translation" in: Steiner and Yallop. eds. 2001 Exploring Translation and multilingual textproduction. Beyond Content. Berlin: New York De Gruyter: 41-126.

Messeri, Lisa & M. J. Crockett 2024 “Artificial intelligence and illusions of understanding in scientific research”. In: Nature  Vol 627 . 7 March 2024 . 49-58

Mikros, George, and Boumparis, Dimitris. 2024. “Cross-linguistic authorship attribution and gender profiling. Machine translation as a method for bridging the language gap”. In: Digital Scholarship in the Humanities, Volume 39, Issue 3, September 2024, Pages 954–967, https://doi.org/10.1093/llc/fqae028. Oxford University Press.

Munday Jeremy 2016 Introducing translation studies. Theories and applications.  (2016 4th edition.) London: Routledge

Porter, B., Machery, E. 2024. “AI-generated poetry is indistinguishable from human-written poetry and is rated more favorably”. Sci Rep 14, 26133 (2024). https://doi.org/10.1038/s41598-024-76900-1

Steiner, Erich. 2020. “Translation, equivalence and cognition”, in: Alves, Fabio and Jakobsen, Arnt Lykke. Eds. 2020. The Routledge Handbook of Translation and Cognition. London, New York: Routledge Taylor and Frances pp. 344-359

Steiner, Erich. 2021. Textual instantiation, the notion of “readings of texts“, and translational agency” In: Kim, M., Munday, J. , Wang, P and Wang. Z. 2021. Systemic Functional Linguistics and Translation Studies. London: Bloomsbury 35-64

van Genabith, Josef. 2020. Neural Machine Translation. In (ed.) Jörg Porsiel. Maschinelle Übersetzung für Übersetzungsprofis. BDÜ Fachverlag. ISBN: 978-3-946702-09-2. pp. 59-115.

Venuti, Lawrence. 2000. “Translation, Community, Utopia.” In: Venuti 2000. ed.: The translation studies reader. London: Routledge 468-488

View all meetings

Latest Recordings

From the PhilSoc Blog


Read more on the PhilSoc blog

News & Announcements

Funding & Bursaries

Information about funding support available from PhilSoc may be found here