Investigating cognitive and interpersonal factors in hybrid human-AI practices

Abstract This article investigates interlingual respeaking (IRSP) as a hybrid practice involving direct interaction between a language professional and speech recognition software to produce live subtitles. This technique combines language transfer with the diamesic shift of subtitling, the immediacy factor of interpreting, and direct interaction with AI-driven technology. It is, therefore, a prime example of Multilectal Mediated Communication which requires language professionals to adapt the procedural skills acquired through training and professional experience in related fields. The article reports on an empirical study that explores cognitive abilities and interpersonal traits as underexamined but critical elements of competence in IRSP, guided by Hambrick et al.’s (2018) multifactorial model of expertise. A mixed-methods exploratory design was used to investigate factors influencing IRSP performance. Fifty-one language professionals participated in a twenty-five-hour upskilling course, completing pre- and post-experiment surveys as well as a battery of cognitive tasks and interpersonal skills scales. Multiple regression analyses were used to investigate predictors of IRSP accuracy and errors. Our study found that, among cognitive variables, complex working memory predicts accuracy. Additionally, different working memory resources were found to be involved in different error categories. In relation to interpersonal traits, integrated regulation (a measure of motivation) and conscientiousness were found to negatively predict accuracy. The study highlights the interplay between human cognitive and interpersonal variables in IRSP as a complex form of human-AI interaction, establishing a baseline for future research and advancing methodological approaches and training frameworks for language professionals working in AI-assisted environments.