Machine Versus Human Explicitation in Consecutive Interpreting: Evidence From ChatGPT

Despite growing research interest in machine interpreting, limited studies have systematically examined machine versus human explicitation, which represents an important contextualised decision-making process in interpreting. This exploratory study seeks to address this gap by comparing machine and human explicitation in the consecutive interpreting of political speeches. Methodologically, this study is distinguished from prior research that employed cascaded systems, by using ChatGPT-4o Advanced Audio Mode as an end-to-end, audio-mode large-language-model-powered machine interpreting system. Results show that while ChatGPT-4o produced a human-like explicitation pattern, with experiential explicitation being predominant, followed by interpersonal and textual explicitation, it generated much fewer explicitation shifts. This conservatism is particularly evident in engagement-based interpersonal explicitation, process- and circumstance-based experiential explicitation, and elliptical textual explicitation. Qualitative analysis reveals that ChatGPT-4o’s explicitation is primarily driven by the co-text, whereas the more active human explicitation is underpinned by a multi-layer awareness of the co-text, the situation and the sociopolitical context, leading to more precise outputs and more effective cross-lingual communication. This study illuminates the added value of human explicitation for high-stakes communicative events, which warrants attention in pedagogy, practice and reception.