Assessing subjective workload for live captioners

Live captioning is a challenging task that requires intense concentration to convert audio to text in real-time. Despite the importance of live captioning for accessibility, little is known about the subjective workload of captioners in this context. This study aimed to measure the subjective workload of live captioners using the NASA-TLX and to explore the factors that contribute to their mental workload. Thirty live captioners participated in the study, completing questionnaires and interviews. Results showed that the subjective mental workload of live captioners is high, similar to that of neurosurgeons. The mental workload was found to be associated with caption paraphrasing and employment status. The challenges of the job, such as the speed of speaking in live content, cognitive tasks involved in paraphrasing, the concern about poor performance, the impact on audiences, and the lack of control over job scheduling contribute to this high workload. These findings suggest the need for modulating the scheduling of the workers, having longer breaks, and working in teams rather than independently. Introducing paradigm changes for live captioning workflow, such as reducing the human effort of typing by adopting auto-generated captions, so that captioners become decision-makers or managers of generational AI systems should also be considered. By addressing these issues, we can help improve the well-being of live captioners and the quality of captions, ultimately enhancing accessibility for all.