The effect of automatic speech recognition on Iranian interpreters' cognitive load: An fNIRS study

In the dynamic landscape of language interpreting, the integration of technology has ushered in a new era, marked by the advent of computer-assisted interpreting (CAI) tools. However, while the promise of improved communication through these tools is evident, a crucial facet demanding scrutiny is the intricate relationship between computer-assisted interpreting and the cognitive load experienced by interpreters. The present study was an attempt to use functional Near-Infrared Spectroscopy (fNIRS) to compare the cognitive load experienced by interpreters when utilizing a CAI tool, in this case automatic speech recognition (ASR), versus interpreting without such a tool. To this end, 12 interpreters were asked to perform two tasks: 1) simultaneous interpreting with the help of ASR (ASRSI) and 2) interpreting without ASR (NoASR-SI). fNIRS records changes in the concentration of oxyhemoglobin [ΔHbO2] and deoxyhemoglobin [ΔHbR] as it is sensitive to hemodynamic changes in the blood. Therefore, the interpreters’ cognitive load in both tasks was measured using the analysis of the HbO2 signals, which are an indicator of brain activation, and a paired t test was used to determine if there was a significant difference between the means of concentration changes of the two tasks. The results showed that the left temporal cortex (LTC) was significantly activated (p<0.05) during simultaneous interpreting from English into Persian. Furthermore, the mean of changes in concentration of HbO2 revealed that more cognitive load was experienced in interpreting without ASR compared to interpreting with ASR, meaning cognitive load was reduced when using ASR. In addition, participants’ feedback regarding the integration of ASR into interpreting was investigated through a questionnaire. The findings showed that participants’ subjective perceptions of ASR did not fully correspond to the objective neural activity recorded during simultaneous interpreting with and without ASR support.