Conceptualisations of ‘translation’ in machine translation research and development

Abstract What constitutes translation when performed by a machine? This article explores this question by examining how machine translation (MT) researchers and developers conceptualise translation within their professional practice. Based on a series of expert interviews, the study uncovers divergent notions ranging from translation as systematic information transfer to translation as cultural mediation. These perspectives are then situated against linguistic and functional theories in Translation Studies, which provide contrasting reference points: the former treating translation as a primarily mechanical process, the latter as a purposive, human-centred act of mediation. Drawing on Social Construction of Technology (SCOT), the analysis highlights how conceptualisations of translation in MT R&D are shaped both by technological affordances — such as neural architectures and machine learning logics — and by broader social and economic contexts. The findings show that whether MT qualifies as translation depends less on a definitive resolution than on the lens through which the concept is framed.