Human Agency and Machine Agency in Digitalised and Datafied Translation Production Networks

Abstract

This essay draws on concepts and findings from translation so­ci­ol­ogy, translation technology and translation process research and in­ves­ti­gates translators’ agency as it manifests itself in modern digitalised, datafied and distributed working environments (translation production networks) as well as in the interplay with non-human ‘agents’ such as neural machine trans­lation (NMT) systems or recent large language models (LLMs). My dis­cus­sion of translators’ agency in translation production networks adopts a macroscopic perspective. It conceptualises such networks as sociotechnical systems with asymmetric power relations between the individual network actors and analyses how translators can exercise their agency vis-à-vis these actors and how translators’ agency may be curtailed by the other actors’ agen­cy. Then the perspective is narrowed down to the interplay of human and machine agency with a focus on translator-NMT interaction in pro­duc­tion networks. In this interaction, translators and NMT systems form a hy­brid system in which both system elements contribute to the successful com­pletion of a task. In such a hybrid human-machine system, human and machine agency become intertwined in a “double dance of agency” (Rose/ Jones 2005), where both forms of agency can either converge or diverge. The chapter concludes with some reflections on human vs. machine agency in the context of recent LLMs such as GPT-4. These language models, I argue, bring about a further redistribution of cognitive labour between hu­mans and machines – not just in the translation sector but also in a wide range of other professional fields – and therefore lead to a reconfiguration of the interplay between human and machine agency.