Examining anxiety in machine translation post-editing: A multimethod approach using eye-tracking and psychometric...

This paper presents a multimethod research design to investigate anxiety in machine translation post-editing (MTPE) and its potential relationship with pupil dilation and visual attention distribution. It compares data from 20 translation students and 18 professional translators performing MTPE with Statistical Machine Translation (SMT) and Neural Machine Translation (NMT) output. Although pupil dilation has been used in Cognitive Translation and Interpreting Studies (CTIS), primarily as an index of cognitive effort, its use as a potential indicator of anxiety-related arousal during MTPE remains limited. This study integrates psychometric data from the State-Trait Anxiety Inventory (STAI), physiological data from pupil dilation, and behavioral data from eye fixations to examine anxiety-related responses during MTPE. In this exploratory sample, no statistically significant differences in self-reported anxiety were detected between participant groups or MT technologies. The integrated pupil-dilation model showed no significant group effect or Group × MT technology interaction, although professionals showed greater pupil dilation when post-editing SMT than NMT. Participant-level correlation analyses did not reveal statistically significant associations between self-reported anxiety and pupil dilation. Although some correlations among students were positive, they did not reach statistical significance. Among students in the NMT condition, higher trait anxiety and pre-task state anxiety were associated with increased fixation counts in the Internet browser AOI, suggesting greater consultation activity during MTPE. The findings support a cautious multimethod approach to studying anxiety in MTPE, where pupil dilation and fixation behavior are interpreted as complementary rather than direct indicators of anxiety.