Logo image
Disrupter or Enabler? Assessing the impact of using automatic speech recognition technology in interpreter-mediated legal proceedings
Doctoral Thesis   Open access

Disrupter or Enabler? Assessing the impact of using automatic speech recognition technology in interpreter-mediated legal proceedings

Wangyi Tang
University of Surrey
Doctor of Philosophy (PhD), University of Surrey
30/10/2026
DOI:
https://doi.org/10.15126/thesis.902322

Abstract

Remote Legal Interpreting Simultaneous Interpreting Consecutive Interpreting Artificial Intelligence or Cybernetics Human-Computer Interaction

In interpreter-mediated legal proceedings, interpreting enables access to legal communica tion across languages. Automatic speech recognition (ASR) captions are increasingly available in remote communication, but their value in this context remains uncertain where accuracy and completeness carry legal consequences. This study investigates whether ASR-generated live captions function as a disrupter or an enabler in remote legal interpreting, focusing on their impact across simultaneous interpreting (SI) and consecutive interpreting (CI). It uses a mixed method, simulation-based within-subject design involving 12 professional English–Chinese legal interpreters. Each participant completed a simulated Crown Court sentencing hearing in four conditions: SI with ASR, SI without ASR, CI with ASR and CI without ASR. Performance was ex amined through four evidence layers: accuracy analysis using the NTR model, ASR-impact trac ing for ASR-error handling in the interpreted output, NASA Task Load Index (NASA-TLX) for per ceived workload, and interviews on caption use, risks and coping strategies.

The findings show that ASR captions were associated with improved accuracy in both modes, but with different trade-offs. In SI, improvement was larger but less uniform, driven mainly by omission reduction and accompanied by a slight workload decrease, especially in temporal de mand. In CI, improvement was smaller but more consistent, driven mainly by substitution re duction, but coincided with higher effort and temporal demand. ASR-impact tracing showed that CI had more propagated ASR-error incidents and a higher weighted propagated burden, whereas SI had a larger non-assessable share, meaning that caption influence could not always be identified from output alone. Interview data help explain these patterns: captions were most useful when used selectively to check high-risk details, but became disruptive when interpret ers shifted into sustained reading or over-reliance. By comparing SI and CI directly and linking accuracy, ASR-error uptake, workload and user accounts, this study moves beyond whether ASR helps. It explains how and why ASR support becomes useful or risky, offering practical guidance for mode-sensitive training, interface design and calibrated reliance in remote legal interpreting.

pdf
Wangyi Tang_PhD Thesis6.35 MBDownloadView
Version of Record (ETD) Open Access CC BY-NC-SA V4.0

Metrics

3 File views/ downloads
1 Record Views

Details

Logo image

Usage Policy