Des zones d'indécidabilité dans la traduction automatique et dans la traduction humaine

Abstract

This article examines Jacques Derrida's concept of undecidability in machine and human translation. Defined as the very condition of possibility of acting and deciding (Caputo 1997: 137), undecidability is an essential component of decision-making. Using Derrida's concept as a starting point for thinking the intricacies of translation decision processes, this article explores manifestations of undecidability in the translations of an extract from Henri Meschonnic's Traduire au XXIe siècle (2008). Deliberately selected because of its direct resonance with the notion of undecidability, Meschonnic's comment claims that, except for automated translations, no practice of translation is possible without reflexive decision-making (Meschonnic 2008: 60). The observations discussed in this article challenge clear-cut distinctions between human and automated translations by showing that programmability and undecidability are features shared by both humans and machines in translation. As such, this study suggests that uncertainty, as a pre-condition of ethical decision-making, must be fostered within mechanical translations themselves, and that in-depth training in translation technology is necessary for a responsible use of machine translation.